A Retrospective on the Classical Gold Standard, 1821-1931 (National Bureau of Economic Research Conference Report)

A Retrospective on the Classical Gold Standard, 1821-1931

A Conference Report National Bureau of Economic Research

A Retrospective on the Classical Gold Standard,

1821-1931 Edited by

Michael D. Bordo Anna J. Schwartz

The University of Chicago Press

Chicago and London

MICHAEL D. BORDO is professor of economics at the University of South Carolina and research associate of the National Bureau of Economic Research. ANNA J. SCHWARTZ, research associate of the National Bureau of Economic Research, is coauthor, with Milton Friedman, of Monetary Trends in the United States and the United Kingdom, 1867-1975.

The University of Chicago Press, Chicago 60637 The University of Chicago Press, Ltd., London © 1984 by The National Bureau of Economic Research All rights reserved. Published 1984 Printed in the United States of America 90 89 88 87 86 85 84 5 4 3 2 1

Library of Congress Cataloging in Publication Data Main entry under title:

A Retrospective on the classical gold standard 1821-1931. (A Conference report / National Bureau of Economic Research) Papers from a conference sponsored by the National Bureau of Economic Research, held in March, 1982. Bibliography: p. Includes index. 1. Gold standard-History-Congresses. I. Bordo, Michael D. II. Schwartz, Anna Jacobson. III. National Bureau of Economic Research. IV. Series: Conference report (National Bureau of Economic Research) HG297.R44 1984 332.4'222 84-2440 ISBN 0-226-06590-1

National Bureau of Economic Research Officers Franklin A. Lindsay, Chairman Richard Rosett, Vice Chairman Eli Shapiro, President David G. Hartman, Executive Director and Corporate Secretary

Charles A. Walworth, Treasurer Sam Parker, Director of Finance and Administration

Directors at Large Moses Abramovitz George T. Conklin, Jr. Jean A. Crockett Morton Ehrlich Edward L. Ginzton David L. Grove Walter W. Heller Saul B. Klaman

Franklin A. Lindsay Roy E. Moor Geoffrey H. Moore Michael H. Moskow James J. O'Leary Peter G. Peterson Robert V. Roosa Richard N. Rosett

Bert Seidman Eli Shapiro Stephen Stamas Lazare Teper Donald S. Wasserman Marina v. N. Whitman

Directors by University Appointment Charles H. Berry, Princeton James Duesenberry, Harvard Marcus Alexis, Northwestern J. C. LaForce, California, Los Angeles Paul McCracken, Michigan Ann F. Friedlaender, Massachusetts Institute of Technology

James L. Pierce, California, Berkeley Nathan Rosenberg, Stanford James Simler, Minnesota James Tobin, Yale William S. Vickrey, Columbia John Vernon, Duke Burton A. Weisbrod, Wisconsin Arnold Zellner, Chicago

Directors by Appointment of Other Organizations Carl F. Christ, American Economic Association Robert S. Hamada, American Finance Association Gilbert Heebner, National Association of Business Economists Robert C. Holland, Committee for Economic Development Stephan F. K.aliski, Canadian Economics Association Douglass C. North, Economic History Association

Rudolph A. Oswald, American Federation of Labor and Congress of Industrial Organizations G. Edward Schuh, American Agricultural Economics Association Albert Sommers, The Conference Board Dudley Wallace, American Statistical Association Charles A. Walworth, American Institute of Certified Public Accountants

Directors Emeriti Arthur F. Burns Emilio G. Collado Solomon Fabricant Frank Fetter

Thomas D. Flynn Gottfried Haberler George B. Roberts Murray Shields

Boris Shishkin Willard L. Thorp Theodore O. Yntema

Since this volume is a record of conference proceedings, it has been exempted from the rules governing critical review of manuscripts by the Board of Directors of the National Bureau (resolution adopted 8 June 1948, as revised 21 November 1949 and 20 April 1968).

Contents

Preface Introduction Anna J. Schwartz I.

2. The Success of Purchasing-Power Parity: Historical Evidence and Its Implications for Macroeconomics Donald N. McCloskey and J. Richard Zecher Comments: Robert E. Lipsey, Milton Friedman General Discussion

23

121

TECHNICAL PROCEDURES: RULES OF THE GAME Chairman: Robert A. Mundell 3. The Bank of England and the Rules of the Game under the International Gold Standard: New Evidence John Dutton Comment: Donald E. Moggridge Reply

vii

1

THE GOLD STANDARD AS INTERPRETED IN TRADITIONAL AND REVISIONIST WORKS Chairman: Moses Abramovitz 1. The Gold Standard: The Traditional Approach Michael D. Bordo Comment: C. Knick Harley General Discussion

II.

xi

173

viii

Contents

4. Bank of England Operations, 1893-1913 John Pippenger Comment: Charles A. E. Goodhart General Discussion of Dutton and Pippenger Papers 5. The Gold Standard and the Bank of England in the Crisis of 1847 Rudiger Dornbusch and Jacob A. Frenkel Comment: J. R. T. Hughes General Discussion III.

233

INTERNATIONAL EXPERIENCE IN THE OPERATION OF THE GOLD STANDARD Chairman: Karl Brunner 6. Canada and the Interwar Gold Standard, 1920-35: Monetary Policy without a Central Bank Ronald A. Shearer and Carolyn Clark Comment: Charles Freedman General Discussion 7. Operations of the German Central Bank and the Rules of the Game, 1879-1913 Paul McGouldrick Comment.' Heywood Fleisig General Discussion 8. Swedish Experience under the Classical Gold Standard, 1873-1914 Lars Jonung Comment: Peter H. Lindert 9. Italy in the Gold Standard Period, 1861-1914 Michele Fratianni and Franco Spinelli Comment.' Richard E. Sylla General Discussion of Jonung and Fratianni-Spinelli Papers

IV.

203

INTERNATIONAL LINKAGES UNDER THE GOLD STANDARD Chairman: Allan H. Meltzer

277

311

361

405

ix

Contents

10. The Gold Standard and the Transmission of Business Cycles, 1833-1932 Wallace E. Huffman and James R. Lothian Comment: Michael Connolly 11. Real Output and the Gold Standard Years, 1830-1913 Stephen T. Easton Comment: Geoffrey E. Wood General Discussion of Huffman-Lothian and Easton Papers 12. Canada without a Central Bank: Operation of the Price-Specie-Flow Mechanism, 1872-1913 Georg Rich Comment: Peter Temin General Discussion

v.

455

513

547

THE GOLD STANDARD AS A STABILIZER OF COMMODITY PRICES

Chairman: Richard H. Timberlake, Jr. 13. War, Prices, and Interest Rates: A Martial Solution to Gibson's Paradox Daniel K. Benjamin and Levis A. Kochin Comment: Phillip Cagan General Discussion

587

14. Some Evidence on the Real Price of Gold, Its Costs of Production, and Commodity Prices 613 Hugh Rockoff Comment: Robert J. Barro Reply General Discussion 15. The Image of the Gold Standard Leland B . Yeager

651

Participants

671

Au~&fuda

~5

Subject Index

000

Preface

The conference in March 1982 at which the papers in this book were presented brought together some fifty scholars with an interest in economic history and international monetary relations to examine the current state of knowledge of the gold standard as it operated during the 110 years until 1931, when the classical gold standard may be said to have ended. Five sessions were held at which fourteen papers were presented. Comments on each paper were prepared in advance of the conference by a designated discussant. Two comments are included with the paper by Donald McCloskey and J. Richard Zecher: one by Robert Lipsey, the assigned discussant, the other by Milton Friedman, who had no specific paper assignment at the first session but who responded to issues the authors' paper raised. A brief summary of the discussion by conference participants, prepared by Barry Eichengreen, follows the papers and the prepared comments. The replies by two authors to the comments on their papers are also included. In addition to the formal papers, at the final dinner meeting Leland Yeager provided an overview of the conference proceedings. His address is the fifteenth paper in this book. We are grateful to the Earhart Foundation, the Alex C. Walker Educational and Charitable Foundation, and others for their support of the conference. All the participants express their thanks to Kirsten Foss for her efficient management of the conference arrangements. Michael D. Bordo Anna J. Schwartz

xi

Introduction Anna J. Schwartz

Britain's abandonment of the gold standard in 1931 signaled the end of the gold standard era. The new technical economic insights and improvements in statistical and computational techniques that have developed in the half century since then constitute part of the motivation for organizing a conference on the gold standard at this juncture-they should make it possible to mine new nuggets of knowledge from the historical evidence. In addition, recent experience with floating exchange rates and unstable domestic monetary policies has made that historical evidence seem highly relevant to today's problems. Much professional attention is once again focused on the merits of fixed exchange rates and constraints on domestic monetary autonomy. We date the start of the gold standard era in 1821, when Britain resumed specie payments at the parity that had pevailed before the Napoleonic Wars, indeed, from 1717 on. By the end of the era, 110 years later, the gold standard had been transmuted. In the pre-World War I period, it evolved as a system in which countries redeemed their domestic currencies in gold and in which the offsetting of gold flows by monetary authorities to maintain existing monetary conditions, though it occurred from time to time, was not regarded as proper conduct. England-the world's largest trading nation, the center of the world's commodities markets and of the world's gold market, and the world's leading creditor nation-played a central role differing from that of other countries. Though gold was the key reserve, many countries in addition held foreign exchange (largely though not exclusively in sterling), in itself an indication of confidence in the stability of exchange rates. By 1931, gold coins did not circulate in most countries on the gold standard, and paper currency could be redeemed for gold only under severe limits. Sterilization of gold flows was accepted as a desirable way to limit the internal 1

2

Anna J. Schwartz

monetary effects of gold flows. New York rivaled London as a second reserve center. The resultant destabilizing shifts of foreign-exchange reserves from the center experiencing gold losses to the other center led to a loss of confidence in the stability of exchange rates (Dam 1982, pp. 24-60). The studies undertaken for this conference were designed to deepen our understanding of the functioning of the historical gold standard. A major by-product is to affect our response to such current economic questions as the recent resurgence of interest in a gold standard as a solution to problems of inflation, high interest rates, and low productivity. The history of the gold standard may be examined from many perspectives. The conference concentrated on five: 1. What were the main themes of the traditional literature on the gold standard, spanning several centuries, compared to the themes stressed in the analysis associated with the post-World War II monetary approach to the balance of payments? 2. Did operating procedures of the Bank of England before 1914 conform to theoretical notions of the ideal functioning of the gold standard? 3. What was the experience of a sample of four countries, with and without a formal central bank (Canada, Germany, Italy, and Sweden), in the gold standard era? This evidence was designed to supplement more extensive knowledge of the operation of the gold standard in the United States and Great Britain. 4. What links integrated the international monetary system under the gold standard? 5. Did the gold standard stabilize commodity prices? The conference was enlivened by the expression of strong conflicting views on fundamental issues. One such issue was the significance of purchasing-power parity and interest-rate parity. Do independent monetary policies have any role under fixed exchange rates if purchasingpower and interest-rate parity are important? Another issue was the role of central banks under the gold standard. Did these institutions, or less formal ones that performed similar functions, impede or assist the operation of the gold standard, or were they irrelevant, so that the operation of the gold standard was automatic? The pre-World War I period was characterized by economic growth and expanding world trade. Did adherence to fixed exchange rates under the gold standard playa major role in producing growth or was it largely irrelevant to growth? The importance of accounting for the links across the Atlantic under the pre-World War I gold standard was stressed by some participants, whereas at least one participant questioned the validity of the concept of

3

Introduction

an Atlantic economy. The reality of trend movements in commodity prices and long-term interest rates under the gold standard was another subject of debate at the conference. Was it visual spurious regression that produced the appearance of trends or did economic agents apprehend the trends as actualities? The conference did not settle these issues, but future investigators will need to confront them in dealing with the international monetary system. The studies that were prepared for the conference relied on various modes of analysis. Several were based on historical nonquantitative evidence. Some adapted the National Bureau business-cycle analysis. Others used regression analysis. Several studies presented Granger-Sims tests and analyses of ARIMA techniques. One general question that arises with high-powered statistical tests applied to pre-World War I data is the validity of the results, given the questionable reliability of some of the underlying data. Economists may nevertheless welcome the findings as a starting point for further refinement of hypotheses and a spur to efforts to improve the data sets. Section 1 of this introduction summarizes the five sessions of the conference, highlighting unresolved issues. Contemporaries regarded the gold standard as a qualified success yet later observers gave a less favorable account of the era. Section 2 attempts to account for the change in views. The implications of the findings of the conference studies and of Leland Yeager's suggestive talk at a dinner session form the basis for the speculations in the concluding section 3 on the prospects for reinstating a gold standard. 0.1

0.1.1

Summary of Issues Examined

The Gold Standard as Interpreted in Traditional and Revisionist Works

Bordo surveyed six major themes developed since' the eighteenth century in the traditional approach to the gold standard: 1. Gold represented an ideal monetary standard, both domestically and internationally, because of its unique qualities as a standard of value and a medium of exchange. The evils of a depreciated money under an inconvertible fiduciary money were contrasted with the price stability that, according to the commodity theory of money, automatic operation of a gold standard yielded over the long run. 2. The essence of the gold standard was the maintenance of a fixed price of a national money in terms of gold, linking the price levels of all countries. The price-specie-flow mechanism ensured that any disturbance away from the natural distribution of gold determined by a nation's

4

Anna J. Schwartz

real income and money-holding habits would lead to an equilibrating process through arbitrage in the gold market. Gold flows, by changing a nation's money supply, would then also change its price level. 3. The law of one price ensured that through arbitrage and individual trade in commodities, prices for similar goods, taking account of transportation costs and trade impediments, would be similar. 4. Capital flows played a role in the gold standard balance-ofpayments adjustment mechanism, supporting the price-specie-flow mechanism. A decline or rise in the domestic money stock would lead to a rise or fall in short-term interest rates affecting the movement of funds from abroad. Long-term capital flows were a source of disturbance to the balance of payments but also a balancing item in the balance of payments. By raising the price level in the capital-importing country and lowering it in the exporting country, and hence producing a current-account surplus in the latter and a current-account deficit in the former, the transfer of capital resulted in a transfer of real resources. 5. The role of central banks in the adjustment mechanism was initially examined in the context of the gold standard as an automatic monetary rule, but later views shifted to the gold standard as managed by central banks to facilitate adjustment to internal and external gold flows, and finally to discussion of the extent to which central banks in fact followed the rules of the game. 6. Schemes to reform the gold standard included its management by the central bank to shield the domestic money supply from external shocks; the separation of the medium-of-exchange function from the standard of value, by adoption of a tabular standard, bimetallism, symmetallism, the compensated dollar, or a commodity standard; and finally, the creation of some form of a supernational central bank. Harley expressed disappointment at Bordo's lack of attention to the two issues he regarded as central to a better understanding of the gold standard: transfers and reparations and the "Atlantic economy." McCloskey and Zecher took issue with the traditional view that under the gold standard, prices (interest rates) in one country could be out of line with prices (interest rates) in the rest of the world for considerable periods of time, inducing first a gold flow, than a change in the country's money supply, followed by adjustment in its price level (interest rates) to bring it (them) into line. In their view, prices (interest rates) internationally never diverge (except for impediments to trade and capital movements, transportation and transactions costs) because rational economic behavior of consumers and producers (investors) links prices (interest rates) through arbitrage. Lipsey criticized the McCloskey-Zecher identification of purchasingpower parity with the law of one price as impossible to refute and empty

5

Introduction

as a theory. Using aggregate indexes of wholesale prices in different countries to test the existence of arbitrage, as the authors do, moreover, was questionable given the differences in the. construction and composition of those series. On the other hand, if all they meant by purchasingpower parity was that foreign influences on prices cannot be ignored, the proposition was unarguable. The evidence McCloskey and Zecher cite to support the view that specie flows did not activate an adjustment process is that contrary to a claim by Friedman and Schwartz (1963, p. 99), it did not so work in 1879. McCloskey and Zecher assert that U.S. price movements that year possibly anticipated gold inflows, as if arbitrage were at work, but certainly did not lag the gold inflows. Friedman's response is that the comparison McCloskey and Zecher make between price rises and inflows of gold is the wrong one. Gold flows are a proxy for the quantity of money. Comparing prices and changes in the quantity of money directly fully supports the price-specie-flow mechanism. The initial gold flows, to which McCloskey and Zecher refer, had a direct effect on the composition, rather than the rate of growth, of high-powered money. The evidence McCloskey and Zecher cite to supporting purchasingpower parity achieved by arbitrage is the rise in wholesale prices in the United States after the trough in March 1933. They name the depreciation of the dollar rather than the National Recovery Act, to which Friedman and Schwartz alluded, as the proximate cause of the domestic price rise. According to them, domestic monetary growth cannot explain movements in prices beyond what are explained by purchasing-power parity. With respect to the rise in wholesale prices after the trough in March 1933, Friedman denies that the discussion in A Monetary History was an attempt to assess the relative contribution of several sources of price enhancement. The depreciation of the dollar, the growth in the domestic money stock, and New Deal measures, including the Agricultural Adjustment Act, the Guffey Coal Act, and still others that affected wages and prices, all played their part. Yeager's conclusion that acceptance of the general validity of purchasing-power parity and interest-rate parity need not exclude the general validity of the quantity theory of money for the analysis of domestic price movements is a judicious statement. Under a gold standard, centrifugal forces had some play, but ultimately centripetal forces triumphed. The monetary approach to the balance of payments asserts that there is one world, not a collection of separate national entities. One can easily accommodate a one-world view with degrees of autonomy over limited periods for individual economies. Under fixed exchange rates, depending on the degree of autonomy, individual economies could exercise control

6

Anna J. Schwartz

over their prices and interest rates in order to extend the period for adjustment, short of cutting loose from the restraints imposed by the gold standard. Mundell pointed out that the difference between these two views was reminiscent of an earlier discussion in 1937 between D. H. Robertson and Jacob Viner on the international adjustment mechanism. Abramovitz expressed concern that findings reported in the long-swing literature had not been integrated into the discussion of the gold standard. In early writings on the gold standard, gold flows served as the principal adjustment mechanism. Subsequently, as Bordo, Dornbusch and Frenkel, and Rich note, the role of short-term capital flows in the adjustment mechanism came to be recognized. The expansion of possible modes of adjustment to disequilibrium in a country's balance of payments due to domestic or external shocks does not damage the validity of the specie-flow mechanism. It simply indicates that adjustment became possible with a broader range of technical means. In sum, the issue of whether the dominant adjustment mechanism that links economies under fixed exchange rates is purchasing-power parity and interest-rate parity or specie flows was not resolved at the conference. 0.1.2 Technical Procedures: Rules of the Game The papers by Dutton and Pippenger give conflicting interpretations of the evidence on the Bank of England's pre-1914 performance. According to Dutton, the Bank may have violated two versions of the so-called rules of the game: one, the traditional view, requiring central banks to reinforce or not counteract the effects of gold flows on domestic money supplies; the other, a view propounded by Michaely (1971), requiring central banks to refrain from countercyclical operations, limiting their objective to maintenance of convertibility. According to Pippenger, on the other hand, the Bank was sensitive to threats to convertibility posed by international capital flows or by increases in domestic income and did not accommodate changes in the level of domestic incomes. Moggridge noted that Dutton's paper does not determine how much of the offsetting by the Bank of England, in violation of the no-offsetting rule, was "an automatic reflection of the discount market being forced into the Bank and how much reflected deliberate policy" (p. 197) and whether the Bank became more inclined to violate some rules over time. Goodhart agreed that the Bank's open-market operations imparted a procyclical impulse to the monetary base, but attributed it not to profit motives, as Pippenger does, nor to the "needs of trade," as Dutton does, but rather to the Bank's concern to protect its share of the London money market that was threatened by the growth of the London clearing banks in the 1890s.

7

Introduction

The tests of the rules may not have captured the effects of short-term and long-term capital transactions and hence may have provided an inadequate basis for judgments of the operation of the system. An interest-rate variable in a regression will not necessarily reflect the impetus for capital flows. If Brinley Thomas (1973) is correct, British capital exports and American capital imports were linked more closely to domestic- and foreign-investment activity than to interest rates (Bordo's paper in this volume, app. E). Transfers on private capital account must have had a disturbing influence on the lending country of greater significance than the financing of existing commodity trade. In the London money market, in addition, issues of foreign governments created balances for the debtor that required monetary management by the Bank of England to counteract gold movements induced by international capital transactions. The importance of those transactions is indicated by the fact that before 1914, Britain invested abroad at an annual rate of 4 percent of its national income and about 30 percent of its annual savings. Working with a small gold reserve, the Bank of England nevertheless avoided catastrophe in the wake of Britain's huge capital exports and general international interests. How did it achieve that result? One suggestion is that a strategic element was involved-all the players were aware of the consequence of collectively seeking to convert sterling into gold. Perhaps the public-good aspect of the pre-World War I managed gold standard was more important than under the Bretton Woods system. Another possibility is that the Bank of England resorted to some devices that do not appear in regressions testing its performance (on the devices, see Sayers 1936; Bloomfield 1963): 1. The use of variations in the price of gold is well known. The Bank acted on the gold points by varying its price for bars and foreign coins, refusing to sell bars or giving free advances on gold imports. Bank rate did not invariably move in line with the use of gold devices. 2. Open-market operations were conducted in such a way as to shield the domestic economy to the extent possible from actions designed to accommodate international capital movements. What seems to be a violation of the rules may in fact be the adaptation of policy to Britain's international role. 3. To discourage certain loans, typically originating in foreign transactions, the Bank manipulated the rate on advances or rediscounting, setting a higher rate than the official one. 4. England held vast short-term claims on foreign debtors that responded to Bank moves. The responsiveness of short-term capital tended to equalize open-market rates in different gold standard countries. 5. By extending or restricting short-term international loans, the Bank of England exercised a degree of control over the distribution of newly mined gold.

8

Anna J. Schwartz

Accordingly, when the Bank of England intervened in the money market, it may not have been violating the rules of the game. Dornbusch and Frenkel provide another view of the nineteenthcentury gold standard in their study of the operation of the Bank in a single year, 1847, when two crises occurred in the spring and the fall-one due to an external drain, the other to both an internal and external drain. They criticize the Bank's performance in both crises and emphasize the role of international capital flows during the adjustment process. For Dornbusch and Frenkel, suspension of the Act of 1844 limiting the Bank's fiduciary issue was required for the restoration of confidence but represented collapse of the rigid gold standard rules. Hence, for them, the gold standard provided a stable financial framework facilitating financial intermediation only when there was a lender-of-Iast-resort willing to discount freely during crises. Hughes commented that the Dornbusch-Frenkel focus on the monetary liabilities of the Bank of England ignored highly volatile movements of the domestic money supply held by the public that was a multiple of the Bank's note issues. The financial crisis in his view originated in the financial system outside the Bank, and it was the "cascading deluge upon the banks and discount houses" of the private issues "that made the Bank Act of 1844 an iron lid that had to be removed by the Treasury letter" (p. 266). 0.1.3

International Experience in the Operation of the Gold Standard

The papers on the experience of a sample of four countries under the gold standard tend to dismiss the importance of monetary actions in accord with the theoretically appropriate "rules of the game" as the explanation for the operation of the fixed-exchange-rate system. Neither Sweden nor Germany before 1914 apparently observed the rules of the game, according to lonung and McGouldrick. Fratianni and Spinelli report that Italy did not even formally adhere to a gold standard for most of the period 1861-1914. In Shearer and Clark's account, Canada, without a central bank, fortuitously returned to the gold standard after World War I and shortly thereafter abandoned it. The foregoing papers provide evidence that supplements the voluminous literature on Great Britain and the United States (for example, Ford 1962 on the asymmetry of gold standard experience as between Great Britain and the peripheral countries to which it exported capital; Morgenstern 1959 on interactions among the money markets in Great Britain, Germany, France, and the United States; and other studies cited in Bordo, app. E). Freedman noted that Shearer and Clark do not provide a generalequilibrium framework for their analysis of Canadian monetary history

9

Introduction

between the wars, but direct attention to such elements as "the relationship between the price of gold and the relative costs of borrowing in New York and borrowing under the Finance Act, the use of gold devices, and the response by the authorities to the gold flows at the end of the 1920s and in the early 1930s" (p. 307). Fleisig also objected to the absence of a model of international economic interaction underlying McGouldrick's and other papers at the conference. In his view McGouldrick does not explain Germany's gold standard experience in light of current models of the international economy that dismiss both specie flow and monetary theory of the balance of payments as inappropriate. Lindert expressed doubt that Sweden's successful performance was related to the gold standard, arguing that economic growth there was attributable to the high level of human capital and abundant natural resources that would have attracted capital inflows also under flexible exchange rates. Accordingly, he suggested that Sweden remained on the gold standard because the country grew rapidly, not vice versa. Sylla found unacceptable the implication that money was a luxury good in Italy-a conclusion Fratianni and Spinelli reach from the moneydemand function they estimate. He was also skeptical that "country risk" was the explanation for deviations from purchasing-power parity that the authors calculate. Instead, he suggested that Italian prices in fact were world prices, adjusted to take account of falling internationaltransactions costs. How does one reconcile the finding that the gold standard as a fixedexchange-rate system performed well, although country after country seems not to have observed the rules required to remain on the gold standard? One possibility is that central banks violated the rules only to a limited extent. It is clear that cumulated deficits and surpluses in balance of payments occurred only when capital flows sustained them. The system did not break down as a result of such flows. Brunner suggested in discussion from the floor that in the context of well-established expectations that the gold standard would be maintained, whether required adjustment occurred primarily in the shares of traded and nontraded goods, in long- or short-term capital movements, or in substantial changes in relatives prices depended on the character of the shocks affecting individual economies-nominal or real, transitory or permanent. It is easy for present-day observers to assume that central banks before 1914 were guided by macrostabilization goals, as in the post-World War II setting, but there is no firm basis for such an assumption. Whether formal central banks existed, as in Europe, or commercial banks or the Treasury exercised central-bank functions, their behavior was successful in maintaining the standard. Was it all waste motion, because the rules of

10

Anna J. Schwartz

the game were unimportant, as the monetary-approach-to-the-balanceof-payments theorists insist? This issue the conference did not settle. Clearly, discretionary actions were taken. Whether they served to speed adjustment was again left an open question. 0.1.4

International Linkages under the Gold Standard

The papers on linkages under the international gold standard approach the question from a number of perspectives. Huffman and Lothian examine interrelations between the United States and the United Kingdom for the century from 1833 to 1932. They conclude from a historical analysis that cyclical fluctuations were transmitted from one country to the other either by gold movements or by panic-induced changes in the money multiplier. From a battery of Granger-Sims autoregressive tests, they conclude that real income in both countries was influenced by both domestic and other-country variables. In addition, they point to weak links between U.S. and U.K. price-level movements as evidence against the monetary approach to the balance of payments. Connolly, in commenting on the paper, found the evidence against the monetary approach questionable since the Granger-Sims tests the authors rely on omit contemporaneous variables that might reveal effective price arbitrage. Easton, on the other hand, using data for eight countries for 18791914, finds no evidence in bivariate relations that either real or nominal income in one country provided any information on those variables in another. Geoffrey Wood argued that Easton's finding confounds demand and supply shocks under a fixed-exchange-rate system. Demand shocks produce positive correlations, supply shocks, negative correlations between income movements across countries. By examining the period as a whole, rather than individual episodes, Easton's approach ensures that such relationships as existed would not be found. As Wood pointed out, a predictable monetary system imposes no particular systematic behavior pattern on the real economy. Thomas was troubled by the failure to test properly for the Atlantic-economy relationships that he had investigated, not only in Easton's work but also in several other papers that had been presented. Meltzer's point in discussion from the floor bore on the importance of the institutional framework under the gold standard that the papers seemed to neglect. Not only was there a predictable monetary system, but decisions in all markets could be made with firm expectations that, for example, price controls or other arbitrary measures would not be imposed. Rich finds that the price-specie-flow mechanism operated over the long run but not over the short run in pre-World War I Canada, concluding that the failure of the mechanism in the short run was an important cause

11

Introduction

of cyclical instability under the gold standard. Thus a balance-ofpayments deficit that led to a gold outflow, a decline in the monetary base, and a rise in interest rates would induce banks to reduce their reserve ratios and expand the money supply, impeding cyclical adjustment. As evidence, Rich cites lack of correspondence between Canadian and U.S. interest rates and the lower variance of Canadian-relative-toU.S. interest rates. Temin asked why U.S. and Canadian interest rates did not move together, as one would expect if asset markets were unified. Possible explanations that he offered were data problems or noncompetitive behavior by Canadian banks. The studies of international linkages only scratch the surface of the subject. Examination of effects of transmission on national product accounts unavoidably involves the use of questionable data. Annual interpolations of benchmark figures, on which most pre-World War I nominal- and real-income data are based, raise doubts about the reliability of the estimates and the statistical significance of tests utilizing those data. Apart from the reliability of the data, we still need to learn how transmission occurred. The studies do not provide a systematic investigation of the role of transmission instruments. The instruments could have been gold flows, commodity-trade flows, capital flows, interest rates, or monetary flows. Did transmission occur through nominal linkages? If so, how was the division of nominal changes between prices. and output determined in the country that was the recipient of the transmission? Did the division vary from country to country, or were there trend influences that operated jointly on all gold standard countries, such as the secular price movements from the mid-1870s to 1896 and from 1897 to 1913? 0.1.5

The Stability of Price-Level Trends under the Gold Standard

Plots of wholesale prices before 1914 seem to be characterized by well-defined, alternately declining and rising trends. The traditional explanation has focused on the commodity theory of money. A decline in the trend of the price level reflected a more rapid growth of world real output and hence in the demand for monetary gold than the growth in the world's monetary gold stock could accommodate. The movement in the price level induced a shift from nonmonetary to monetary uses of gold and ultimately led to increased gold production. A rise in the trend of the price level reflected more rapid growth in the world's monetary gold stock than in the demand for monetary gold, inducing a shift from monetary to nonmonetary uses of gold and ultimately to decreased gold production. The widely accepted view of persistent trends in the price level under

12

Anna J. Schwartz

the gold standard was subjected to strong attack by Benjamin and Kochin, who argue that U.K. prices display the characteristics of a random walk once systematic movements are eliminated by the ARIMA technique. Similarly, the yield on consols after elimination of systematic movements is essentially a random walk. They dismiss Irving Fisher's finding that a distributed lag on past inflation rates was positively correlated with current interest rates, since such a method of forecasting would be rational only if inflation rates were positively correlated serially. They report that such was not the case and conclude that the effects of war expenditures explain comovements of the price level and interest rates. Once the influence of wars is accounted for, virtually no evidence remains of a linkage between the change in the price level and the change in long-~erm interest rates. Cagan's comments on the paper provide grounds for skepticism with regard to the authors' conclusions. Under a gold standard, prices were not in fact completely stable from year to year, despite the assurance of long-run price stability that a commodity standard is said to provide. Contemporaries might not immediately have recognized that a shift of the trend in prices, say, from falling to rising, had occurred. Such recognition would gradually develop, even though year-to-year changes would be regarded by them as random. Corresponding movements in bond yields would not then be accidental but a gradual response to price trends. Rockoff examines the response of gold production to movements in the relative price of gold before 1933. Gold output responded to market incentives in line with the commodity theory of money, which posits an inverse relationship between the general price level and the opportunity cost of producing gold. The incentives affected the timing of gold discoveries and advances in mining techniques, and influenced the willingness of producers to adopt existing capital-intensive methods of production. Imposition or relaxation of governmental environmental regulations could thwart or support market incentives. Rockoff also compares the means and standard deviations of annual growth rates of the world's monetary gold stock from 1839 to 1929 with those of the U.S. monetary base from 1949 to 1979. The gold standard regime gives more stable results. As Barro notes in his comments, the monetary-gold-stock results do not carryover to broader monetary aggregates and the price level, in a comparison with the post-World War II regime. However, the more favorable results for the post-World War II period may reflect alterations in banking institutions rather than a 'shift from the gold standard. Had these alterations been implemented during the gold standard era, the year-to-year stability of broader aggregates and prices might have surpassed the results for the post-World War II era.

13

Introduction

An existing (Bordo 1981) study compares real-output stability under the gold standard in the United States and Great Britain with the corresponding measure under the managed currency systems that superseded the gold standard. A desirable addition to the conference would have been similar studies for other countries.

0.2

Changing Professional Assessments of the Gold Standard

Looking back over world experience with monetary systems in the nineteenth century and its sequel in the twentieth century, one is struck by fluctuations in the esteem with which economists have regarded a metallic standard. After the widely known examples of paper-money inflation that occurred in the closing decades of the eighteenth century and the first decade of the nineteenth century (in the U.S. colonies, in France during the Revolution, and in Britain during the Napoleonic era), the superiority of a metallic standard seemed self-evident in theory and in practice. In both established and newly created nation-states in the nineteenth century, the evolution of monetary systems usually proceeded with the displacement of silver as the monometallic standard or the consort of gold in a bimetallic standard. The norm evolved as free and unlimited coinage of gold with subsidiary coins of silver, nickel, and bronze or copper, and government fiduciary issues and bank notes freely convertible into gold. In less-developed countries, convertibility was provided by foreign-exchange reserves linked to gold. A paper standard, by contrast, came to represent fiscal imprudence and economic backwardness. Even before the end of the nineteenth century, however, popular and professional criticism of the gold standard arose. What occasioned the criticism was the secular price rise associated with the midcentury gold discoveries and the long, secular price decline that got under way in the 1870s under an expanding international gold standard. The first challenge to the virtue of the gold standard was that it did not assure price stability. In his pamphlet, "A Serious Fall in the Value of Gold Ascertained, and Its Social Effects Set Forth" ([1863] 1884), William Stanley Jevons estimated that between 1848 and 1860 the value of gold had fallen 9 percent. In 1875 he questioned the use of metallic standards of value, in view of the extreme changes in their values, and urged as a reform a tabular standard of value ([1875] 1884). Alfred Marshall ([1887] 1925) discussed "the evils of a fluctuating standard of value" (p. 189), and concluded that "the precious metals cannot afford a good standard of value" (p. 192). He dismissed bimetallism as flawed and proposed as a remedy for the fluctuating standard of value either symmetallism or a tabular standard. With the reversal of the secular price movement after 1896, concern

14

Anna J. Schwartz

shifted to the inflationary,fluctuation of the standard. The remedy that Irving Fisher (1913) proposed was the compensated dollar. The gold standard ceased to function internationally during World War I, and the question of its merits or demerits was temporarily set aside. In the aftermath, inflation in the victorious countries and hyperinflation in the vanquished, as governments financed wartime and postwar expenditures by depreciating their currencies, again revived the attraction of the gold standard. Widely reintroduced in the years 1925 to 1929 (although attenuated by the cessation of gold-coin circulation and the limitation of convertibility to bullion bars or sales of foreign exchange), the gold standard collapsed shortly thereafter, destroyed by the economic holocaust of 1929-33. This time the main professional attack was directed to fixed exchange rates through which the gold standard works, although attention also focused on specific problems that were identified as hampering the operation of the post-World War I gold standard (the maldistribution of gold, the inadequacy of world gold output, and the poorly aligned exchangerate structure that had been restored). Fixed exchange rates required the internal economy to adjust to the balance of payments. Only by cutting loose from the gold standard were countries able to escape the deflationary pressure imposed on them by the fixed-exchange-rate system. Internal adjustment to declining world prices was no longer acceptable domestic economic policy, and growing rigidity of prices and costs allegedly placed an intolerable burden of adjustment on the economy. Moreover, far from correcting externally arising disturbances, the gold standard fostered them by transmitting maladjustments from one country to another. An additional problem under the gold standard, according to its critics, was that capital movements, short-term ones in particular, did not provide a corrective mechanism but instead aggravated the underlying situation that generated the capital flows (see Bordo, appendix E, this volume). The flows, in effect, were uncontrollable. Raising the discount rate had not stopped capital flight but had intensified it; the rising rate was interpreted as a signal that further flight would lead to devaluation. At the same time, the discount-rate rise had served to heighten deflationary pressures on the domestic economy. On the other hand, a discount-rate rise that was expected to curb internal expansion instead attracted capital from abroad and promoted further expansion. Capital movements, under fixed exchange rates, induced by interest-rate changes, operated primarily on reserves and foreign exchange of the central bank but did not immediately induce changes in the current account. Fundamental adjustment, moreover, was deterred when long-term capital exports were offset by short-term capital imports. Alternatively, when long-term capital exports ceased, the capital-importing countries confronted fixed-

15

Introduction

interest charges with deflationary impact on their economies, with reflex influence on the capital-exporting, interest-receiving countries. The gold standard thus was charged with having contributed to the instability of the world economic system after 1929. Professional approbation of a paper standard that gained ground in the 1930s was tempered by the belief that unrestrained, it would encourage beggar-thy-neighbor policies. The Bretton Woods arrangements embodied the interpretation of the views and experience of the 1930s-pegged exchange rates were essential to prevent chaos in international financial and trade transactions, but national economies should be free to restrict capital flows and to resort to the expedient of devaluation in order to be relieved of the necessity to deflate when in current-account deficit. The objectionable feature of pegged rates in forcing governments to implement monetary changes that conflicted with the goals of full employment or price stability would be removed while preserving the desirable feature of providing stable conditions in foreign exchange to promote international trade. Convertibility of many European currencies was first achieved under the Bretton Woods system in 1958. For only a few years thereafter can the system be said to have functioned fairly effectively. From the mid-1960s on, it was characterized by repeated foreign-exchange crises as market participants anticipated that existing par values were unsustainable and shifted funds from a weak currency to a strong currency, exacerbating the external position for both currencies. Since the collapse of the Bretton Woods arrangements, efforts to rehabilitate the gold standard have proceeded along two lines. One, inspired by the professional development of the monetary approach to the balance of payments, argues in favor of fixed exchange rates as a way to attain the benefits of risk-pooling and the integration of commodity and factor markets on a worldwide basis. The other line is drawn from the collapse of the Bretton Woods arrangements. The lesson, on this view, is not that only a floating-rate system can accommodate inflationary policies in the reserve-center country and conflicting policies in the nonreserve countries. Rather, the lesson is that the floating-rate system has permitted enormous growth of inconvertible paper-money issues that produced unprecedented peacetime inflation rates and extraordinary levels of interest rates. Consequently, it is argued, it is essential to establish a stable international money based on gold. The advocates offer varying prescriptions. One would rely on the changing market price of gold as an indicator to the monetary authorities of the appropriate rate of increase or decrease in the growth of the money supply, with no commitment on their part to buy or sell gold or to peg its price. Although the price of gold would thus playa part in the monetary system, it would lack crucial elements of gold standards known in the

16

Anna J. Schwartz

past. Another prescription includes stabilizing the dollar price of gold, issuance of gold coins with a face value equal to the stabilized gold parity, restoration of convertibility by linking change in money bases-in the United States to gold purchases and sales in a private gold market, and in non-reserve-center countries to changes in their holdings of gold and foreign exchange-and, finally, multilateral surveillance of country balance-of-payments problems. A more radical prescription would eliminate government-issued money. The government's role would be limited to defining a monetary unit as a specific weight of gold. Private issuers would then be free to issue claims denominated in the officially defined unit. This section suggests that support for a resurrected metallic standard of whatever form would in time dissolve, as it has for all earlier standards. What the odds are for success in the restoration of a role for gold is the subject of the concluding section. 0.3

Prospects for Reinstating the Gold Standard

The conference studies deal with historical evidence-obviously necessary to our understanding of the gold standard as it once existed. That evidence also directs our attention to the possibility that the factors that permitted the gold standard to flourish are now obsolete. What were those factors? Can we now re-create them? We can distinguish at least seven objective factors that promoted the existence of an international metallic standard: 1. the essentially fixed price of gold over the century the conference studies covered 2. a link between domestic money supply and the gold reserve 3. relative stability in conditions of gold production 4. equilibrium in mint pars among gold standard countries 5. coordination of economic policies among countries adhering to the standard 6. limited role of government in economic and social affairs 7. relative absence of political upheavals exemplified by war and revolution and the role of London as the hub of the international monetary system These objective factors were stabilizing forces that made the gold standard a stable standard of value. Whether the stability was an inherent feature of the gold standard or simply the consequence of underlying stability of other institutions is the issue. In addition to the objective factors, mention must also be made of the weight of the psychological belief in the unquestioned and unquestionable obligation to adhere to the gold standard and to the specific fixed price of gold. Flouting the gold standard risked the opprobrium of one's

17

Introduction

own countrymen and of the rest of the world. Political leaders did not regard the gold standard as a policy instrument subject to manipulation in the pursuit of other goals. The hold of the gold standard as the guarantor of the domestic value of a currency and of stable international financial dealings was sacrosanct. Can we count on the stability of the objective factors in contemporary economic circumstances? 0.3.1

The Price of Gold

A fundamental problem confronting the reinstatement of the gold standard is the choice of the dollar price at which to resume. The very conception of trying to determine the correct price somehow violates the mystique of the standard. The price then becomes a political decision, the opposite of the freedom of the standard from political influence that untlerlay its mystique. For the purpose of this analysis, assume the following solution: let the inflation rate of the general price level be reduced to zero; the price of gold at that time would be the correct price at which to resume. Once a price for gold is determined, the principal central banks, it has been suggested, should proceed to peg it. To prevent the gold price from rising, sales of gold from existing stocks could be used. To support the price, countries could use their own currencies-with possible inflationary consequences. Assuming the price were "correct," the pegging operation might be successful. Arranging the responsibility for intervention in the gold market could be managed along the lines of the Gold Pool of 1961, provided exchange rates did not vary. If they did, since an exchange-rate change is a gold-price change in at least one country, speculation in gold markets would be encouraged. The pegging operation would then become more troublesome. 0.3.2 Linking the Domestic Money Supply and Gold Reserves A pegged price of gold is not a sufficient condition for a reinstatement of the gold standard. Some link between the domestic money supply and a country's gold reserves is essential. Would it be feasible to restore convertibility of paper currency into gold for domestic and foreign holders? Countries would be required to yield the discretion they currently exercise in determining the level and growth rate of their domestic money supplies and to accept the effects on money supply that changing gold reserves would dictate. Would they be willing to accept so severe a restriction on their internal monetary policies? 0.3.3

Stability of Gold Output

If a correct price of gold were achieved for resumption, the stability of the price level under the gold standard thenceforth would then depend on

18

Anna J. Schwartz

the adequacy of gold output to provide for monetary and nonmonetary demands for gold. An adequate supply of gold is essential for adequate monetary growth. The forecasts of gold output over the rest of the century in the market economies with known gold reserves are not optimistic. Whether the forecasts might be belied by discovery of new mines or mining processes and whether the inadequacy of the flow supply might be offset by changing patterns of industrial demand for gold or shifts from investment stocks still leaves the reinstatement of the standard as a measure that risks imposing long-run deflation on the economy. The fact that the bulk of current world gold output is produced by South Africa and the Soviet Union adds an element of possible instability in future gold output for political reasons. Rockoff suggests another difference between the past and the putative future performance of the gold standard related to the gold-mining industry, namely, that the public is unlikely to tolerate long and uncertain lags in the response of the gold supply to the changing demand for money. This difference possibly could be classified as psychological, but, if accurate, it clearly impinges on an objective factor. 0.3.4 Fixing Multilateral Exchange Rates Once a correct fixed price of gold were chosen, each gold standard country would adopt par rates of exchange for its currency relative to other currencies. As Yeager remarks, the mint pars under the classical gold standard expressed an equilibrium that had gradually evolved among national price levels. This time, par rates of exchange would be arbitrarily chosen. The mistakes in choice of exchange rates when Euro'pean countries resumed in the decade of the 1920s and again under the Bretton Woods arrangements are not reassuring. 0.3.5

Coordination of-National Economic Policies

The gold standard can survive in a world in which countries allow gold to move freely; gold does not accumulate in any country and gold does not drain away from any country without being allowed to exercise an expansionary or contractionary effect, respectively, on the level of prices; and major disequilibria in price levels and financial conditions among countries are not endured. The forces that caused the breakdown of the Bretton Woods system were unleashed by actions of countries with a persistent deficit or surplus in their balances of payments. Those actions were taken to delay or resist changes in prices and costs expressed in national currencies. Under fixed exchange rates, convergence of national economic policies is essential for the system to be viable. The European Monetary System presupposes such behavior. Yet since 1979 when the system was established, member countries have repeatedly preferred to alter the relation between national price and cost levels by exchange-rate

19

Introduction

changes. This is not a good augury for restoration of an international gold standard. 0.3.6

Role of Government

Under the classical gold standard, governments in peacetime did not undertake expenditures that were financed by the printing press. In some gold standard countries, government was not divorced from business and social insurance was accepted policy. Basically, however, government participation in economic activity was restrained by concern to preserve the integrity of the national currency and to maintain its domestic and external value. These concerns receded after 1929 as governments extended their. activities to finance stabilization policies in response to interest groups wielding political influence. The question then arises whether in the future governments will reverse their course, returning to a more limited role, as in the pre-World War I era. Of course, a limited role of the state is not in itself a guarantee of a viable international monetary system, since in earlier eras international monetary affairs were often in disarray, even with limited states (Dam 1982, p. 38). 0.3.7

Civil and International Peace and London's Predominance

The gold standard collapsed when countries were engulfed by war or revolution. The relative political stability of the pre-1914 era therefore contributed to the maintenance of the standard. The significance of this factor is underscored by the prewar examples of capital flows that were steered by governments for national political and strategic reasons. French investors responded to official regulation and pressure by buying Russian government loans for railroad construction-of military value, in the eyes of both governments. Germany's foreign investments also were directed to achieve national-security goals. The ensuing war destroyed not only the gold standard but also the investments. The gold standard flourished before World War I possibly because of the special position of sterling and London. That position was threatened even before the war when Paris and Berlin became important rivals of London. Thereafter, London's predominance was never reestablished. Under the Bretton Woods system, the special position was that of the dollar and the United States. As the position of the U.S. dollar crumbled, the system collapsed. Is an important aspect of the successful operation of a gold-centered monetary system an unshakable confidence that a dominant reserve-currency would always be converted into gold on demand? Which currency would be the candidate for such a role in a future gold standard? This brief survey suggests that the objective factors that served to promote the international gold standard in the past are no longer favorable to such an institution. And, as noted, the psychological factor of

20

Anna J. Schwartz

reverence for the standard has all but vanished except among a minority of faithful believers. Like Miniver Cheevy, they probably were born too late.

References Bloomfield, Arthur I. 1963. Short-term capital movements under the pre-1914 gold standard. Princeton Studies in International Finance, no. 11. Princeton: Princeton University Press. Bordo, Michael. 1981. The classical gold standard: Some lessons for today. Federal Reserve Bank of St. Louis Review 63: 1-17. Dam, Kenneth, W. 1982. The rules of the game. Chicago: University of Chicago Press. Fisher, Irving. 1913. A compensated dollar. Quarterly Journal of Economics 27 (Feb.): 213-35, 385-97. Ford, Alec G. 1962. The gold standard, 1880-1914: Britain and Argentina. Oxford: Clarendon Press. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Jevons, William Stanley. [1863] 1884. A serious fall in the value of gold ... set forth. Reprint. In Investigations in currency and finance, ed. H. S. Foxwell. London: Macmillan. - - . [1875] 1884. An ideally perfect system of currency. Reprint. In Investigations in currency and finance. See Jevons [1863] 1884. Marshall, Alfred. [1887] 1925. Memorials of Alfred Marshall. Ed. A. C. Pigou. Reprint. London: Macmillan. Michaely, Michael. 1971. The responsiveness of demand policies to balance ofpayments: Postwar patterns. Studies in International Relations, no. 5. New York: National Bureau of Economic Research. Morgenstern, Oskar. 1959. Internationalfinancial transactions and business cycles. Princeton: Princeton University Press. Sayers, Richard S. 1936. Bank of England operations, 1890-1914. London: P. S. King and Son. Thomas, Brinley. 1973. Migration and economic growth. 2d ed. Cambridge: Cambridge University Press.

PART

I.

The Gold Standard as Interpreted in Traditional and Revisionist Works

1

The Gold Standard: The Traditional Approach Michael D. Bordo

1.1

Introduction

What was the traditional approach to the gold standard? In this paper, I try to provide an answer to the question by examining the works of major writers on the subject since the eighteenth century. 1 The choice of writers and works surveyed is based on my judgment that the works encompassed a significant share of the content of the traditional approach and that the writers played a significant role in the history of economic thought. Six major themes formed the traditional approach, and five major schools of thought may be identified. 1.2 Major Themes in the Literature The first theme, which runs from Cantillon to present-day writers, was that gold (the precious metals) was an ideal monetary standard, domestically and internationally, because of its unique qualities both as a standard of value and a medium of exchange. A stable price level in the long run that an automatically operated gold standard produced, in line with the commodity theory of money, was invariably contrasted to the evils of inconvertible fiduciary money. At the hands of even well-meaning policymakers the latter would inevitably lead to depreciation of the value of money. However, most writers, following Adam Smith, emphasized the social saving from using fiduciary money instead of a commodity money Michael David Bordo is professor of economics at the University of South Carolina and a research associate of the National Bureau of Economic Research. For helpful comments and suggestions the author would like to thank Michael Connolly, John McDermott, Peter Lindert, Anna Schwartz, and Larry White; he would also like to thank Fernando Santos and Glen Vogt for able research assistance.

23

24

Michael D. Bordo

and hence were concerned with the properties of a convertible (or mixed) standard to ensure price stability. The second theme was the price-specie-flow mechanism. The essence of the gold standard was the maintenance of a fixed mint price of national money in terms of gold (achieved by specifying the weight of a nation's coinage in terms of gold). That rule ensured uniformity of the price of gold across nations (and regions) through the process of arbitrage in gold. 2 Moreover, each country's price level was determined by its stock of monetary gold, which in turn was determined (naturally distributed) by the nation's real income and money-holding habits. Consequently, the price levels of all countries were linked together under the gold standard by the fixed definition of the monetary unit in terms of gold. Any disturbance away from the natural distribution of precious metals affecting one nation's (region's) price level, and hence the market price of gold, would inevitably lead to an equilibrating process through arbitrage in the gold market. Gold flows, by changing the nation's (region's) money supply, would then also change its level of prices. For example, a gold discovery in one country would lead to an increase in its money supply, an increase in its price level, a~d a fall in the domestic market price of gold. The divergence between the domestic and world gold prices would quickly lead to a gold outflow, a contraction in the domestic money supply, and a fall of the domestic price level. An alternative way of viewing the same mechanism was to focus on the balance of trade-the rise in the domestic price level would raise prices of domestic goods and exports relative to prices of imports, leading to a balance-oftrade deficit, a gold outflow, and a contraction of the money supply. Thus the price-specie-flow mechanism was the means by which arbitrage in one commodity-gold-between nations and regions, served to keep overall national (regional) price levels in line and to maintain balance-of-payments equilibrium. Within this context, different authors stressed the pattern of adjustment of particular classes of commodities. Thus Mill focused on the behavior of the prices of tradable goods relative to those of domestic (nontradable) goods. Others focused on the secondary role of changes in the exchange rate. To the extent that gold prices between nations could differ, reflecting transportation and other costs of transferring gold (the difference between the upper and lower bounds referred to as the gold points), changes in exchange rates (the domestic relative to the foreign price of gold) would also serve to equilibrate the balance of payments without requiring a gold flow. In addition, a number of writers focused on the role of real income in the adjustment mechanism-ehanges in the quantity of money consequent upon gold flows would affect total expenditure and income in addition to, or in some cases instead of, affecting prices.

25

The Gold Standard: The Traditional Approach

The third theme, which is intimately connected to the second, was the "law of one price"-the notion that arbitrage in individual traded commodities would ensure similar prices in a common currency for similar goods, taking account of transportation costs and trade impediments. Along these lines, a distinction was made between domestic (nontraded) goods whose prices are determined primarily by domestic forces and traded goods whose prices are determined by the world mark.et. One question is how to reconcile the law of one price with the pricespecie-flow mechanism, since the latter stressed primarily consequences of arbitrage in gold, while the former stressed arbitrage in all traded commodities. For the classical economists, it was assumed that arbitrage in gold was more effective than in other commodities because of gold's special properties; moreover, since gold served as the money supply (or as the monetary base), alterations in its quantity would impinge on all prices. Ultimately, which goods serve as vehicles for arbitrage is an empirical question. The answer depends on the total costs of arbitrage, including information costs. In the eighteenth and early nineteenth centuries, gold was the commodity with the lowest arbitrage costs, hence gold flows rapidly kept gold prices in line and other goods prices followed. Later in the nineteenth century, with improvements in communications technology and the development of international securities and commodity markets, arbitrage in securities and traded commodities reduced the role for gold flows in the adjustment mechanism. The fourth theme was the role of capital flows in the gold standard balance-of-payments adjustment mechanism. The original conception of the price-specie-flow adjustment mechanism was that it operated through flows of goods and money, but by the middle of the nineteenth century, emphasis was also placed on the role of short-term capital flows as part of the equilibrating mechanism. According to the traditional approach, a decline (rise) in the domestic money stock led to a rise (fall) in short-term interest rates and consequently attracted funds from abroad. Thus in the example of a gold discovery, the increased money supply would reduce domestic interest rates relative to interest rates in other countries, producing both a short-term capital and gold outflow, thereby reducing the amount of adjustment required through changes in the domestic price level. As the nineteenth century wore on and world capital markets became more integrated, emphasis on the role of capital mobility increased to the point where it was regarded as the dominant adjustment mechanism. In addition to short-term capital flows, the role of long-term capital flows was noted as a source of disturbance to the balance of payments. Thus one element of the traditional approach was the role of long-term lending by mature countries, such as England and France, to developing

26

Michael D. Bordo

nations, such as the United States, Canada, and Argentina. Capital flows from the Old to the New World were also accompanied by gold flows, raising the price level in the capital-importing country and lowering it in the exporting country. The resultant change in relative price levels produced a current-account surplus in the capital-exporting country and a deficit in the importing country. Thus the transfer of capital resulted in a transfer of real resources. 3 The process could continue for many years, with developing (developed) countries running a persistent balance-ofpayments deficit (surplus) on current account financed by long-term capital inflows (outflows). The fifth theme, which focuses primarily on the performance of the Bank of England, was the role of central banks in helping or hindering the adjustment mechanism. This theme was a reflection of the British flavor of the gold standard literature and the key role played by the Bank of England in the analysis of the gold standard. Several aspects of the central-bank theme may be noted. One was the debate over rules versus discretion. In the early part of the nineteenth century, emphasis was placed on the advantage of combining the automatic-monetary-rule aspect of the gold standard with the benefits of low-resource-cost fiduciary money. That approach culminated in the Bank Charter Act of 1844 and the separation of the Bank of England into the Issue Department, based on a gold standard rule, and the Banking Department, based on commercial-banking principles. Second, several money-market crises and threats to convertibility in the succeeding quar.ter century led to attention in the literature to the Bank's disregard of domestic-money-market conditions in its operation as a private profit-maximizing institution following a gold standard rule. Thus the Bank, in keeping with its private role, would maintain as Iowa gold reserve as possible while using its Bank-rate weapon to protect its reserve from gold outflows. Bagehot's statement of the "responsibility doctrine" and a prescription for effective central-bank management, referred to as Bagehot's rule, emerged from the scrutiny of the Bank's behavior. A later development was the discussion of the inherent conflict between internal and external price stability under a fixed exchange rate such as the gold standard. In addition, the gold standard came to be regarded as primarily managed by central banks' use of changes in the discotmt rate to facilitate adjustment to both internal and external gold drains. Among the issues stressed were: how Bank rate was made "effective," in the sense of inducing corresponding changes in market interest rates; the use of other policy tools to protect the gold reserves; the channels by which changes in Bank rate would affect the required adjustment in the balance of payments-by inducing short-term capital flows or by changing domestic price levels, economic activity, and the terms of trade.

27

The Gold Standard: The Traditional Approach

Finally, discussion turned on the extent to which central banks followed the "rules of the game," that is, used their policy tools to speed up the adjustment mechanism to an external shock. According to the rules, the central bank of a country experiencing a gold outflow (inflow) should engage in policies to contract (expand) the domestic money supply. The sixth and final theme in the traditional approach was the advocacy of a number of proposals for reform. Many writers suggested schemes for reform of the gold standard both at the national and international levels. At the national level, a persistent theme ranging from Thornton ([1802] 1978) to Keynes ([1923] 1971) was the importance of managing the gold standard so as to reduce the conflict between external and internal stability, i.e., for the central bank to intervene and shield the domestic money supply from external shocks. Related to this theme were schemes to protect the monetary gold stock from internal currency drains, e.g., Ricardo's gold-bullion standard. Finally, schemes were designed to separate the medium-of-exchange function of gold from the store-of-value function. All these proposals attempted to rectify an important defect of the gold standard-basing a nation's money supply on one commodity subject to changing demand and supply conditions. Schemes along these lines included creation of a tabular standard, bimetallism, symmetallism, and Fisher's (1920) compensated dollar. At the international level, proposals designed to provide world price stability included schemes such as bimetallism, symmetallism, and the basing of international money on a wide commodity basket; and also, to ensure international harmony of price-level movements, they favored the creation of some form of supernational central bank. 1.3

Schools of Thought

On the basis both of common views and chronology, the five schools of thought on the gold standard are the classical school, the neoclassical school, the Harvard school, the interwar critics, and the post-World War II reinterpreters. A brief summary of the views of the leading exponents of each school follows. Detailed documentation of these views is provided in five appendixes, one for each school. 1.3.1

Classical School

Eight economists-Cantillon, Hume, Ricardo, Thornton, Mill, Cairnes, Goschen, and Bagehot-eonstituted the classical school. From the writings of these men we can distill the essence of the traditional approach. Cantillon developed the law of one price and aspects of the international adjustment mechanism. Hume is famous for the pricespecie-flow mechanism. Ricardo developed the natural distribution of precious metals and made contributions to issues related to the monetary

28

Michael D. Bordo

standard and monetary reform. Mill, perhaps the key writer of the school, covered virtually all the major themes of the traditional view, and Cairnes tested some of the theoretical implications. Finally, Goschen focused on the role of short-term capital flows, while Bagehot outlined the principles of central-bank management under the gold standard. 1.3.2 Neoclassical School Marshall, Fisher, and Wicksell of the neoclassical school extended and perfected the mechanisms analyzed by the classical school. They, however, explored some of the detrimental effects, both for individual nations and for the world, of adhering to the gold standard, and consequently the need for reform. 1.3.3 Harvard School F. W. Taussig and his students (Viner, Graham, White, Williams, and Beach) attempted to formulate and test a more comprehensive version of the traditional balance-of-payments adjustment mechanism to the external disturbance of long-term capital movements by incorporating gold flows, changes in relative price levels, short-term capital flows, and changes in discount rates. The evidence for the United States, Great Britain, France, Canada, and Argentina produced by this massive research project was largely inconclusive, and in many respects cast doubt on the traditional emphasis on relative price-level changes as the heart of the adjustment mechanism. J. W. Angell, a critic of Taussig, integrated the law of one price in the relative price-specie-flow adjustment mechanism. Despite its critical approach, his work is classified as part of the Harvard-school studies. 1.3.4 Interwar Critics After World War I, a number of writers considered the case for and against a return to the gold standard as it existed pre-World War I. Brown and Smit, accepting in the main the stylized facts of the gold standard as succinctly portrayed by the Cunliffe report (United Kingdom, Parliament [1918] 1979), assessed the gold standard as having been successful before World War I because it was a managed standardmanaged by London-and then documented the special institutional characteristics of the sterling standard. Keynes and Viner discussed the inherent policy conflict between adherence to the gold standard and domestic economic activity, and addressed a plea for more international cooperation. Whale cast doubt on the stylized facts of how the gold standard worked, suggesting that perhaps the traditional approach was incorrect.

29

The Gold Standard: The Traditional Approach

1.3.5

Post-World War II Reinterpreters

In the post-World War II period, scholars reexamined the operation of the classical gold standard on the basis of new evidence and new theoretical and statistical tools. The issues they stressed included the balance-ofpayments adjustment mechanism, capital flows, and rules of the game. The balance-of-payments adjustment mechanism under the gold standard was reexamined from a Keynesian perspective by Ford, from a modern quantity-theory perspective by Friedman and Schwartz, and from the perspective of the monetary approach to the balance of payments by Williamson, Triffin, and McCloskey and Zecher. The role of capital flows was reexamined by Morgenstern and Bloomfield. Finally the operation of central banks under the gold standard with respect to rules of the game was reconsidered by Sayers, Bloomfield, and Lindert. 1.4 A Retrospective The development of the literature on the traditional approach can be viewed from a number of perspectives. I briefly sketch out the elements of two of them: the first, that the interpretation of the gold standard by each school reflected the policy concerns of the time; the second, that the evolution of the interpretation of the gold standard has many of the characteristics of a Kuhnian scientific revolution. According to the first perspective, the development of the traditional approach by the classical economists was strongly influenced by the concern over finding the ideal monetary standard consistent with the classical principles of free enterprise and free trade. This concern thus explains the emphasis on the automatic qualities of the gold standard both as a national and an international standard, the operation of the commodity theory of money that would ensure long-run world price stability (in a stationary world), and the price-specie-flow mechanism that would ensure the natural distribution of precious metals and uniformity of price structures across the world. Behind this smoothly functioning monetary veil, real resources would be efficiently allocated to their best uses by the forces of competition between individuals and enterprises across the world. The introduction of the real-world problems of friction in the balance-of-payments adjustment mechanism and the possible conflict, at least in the short run, between the constraint of the gold standard and internal economic stability led to the development of rules of proper central-bank management of the gold standard. The neoclassical economists, writing at a time when the gold standard was the prevailing standard, accepted its rationale, but concerned themselves with removing one of its major shortcomings-specifically, the tendency for the world price level to exhibit alternating swings of defla-

30

Michael D. Bordo

tion and inflation, reflecting major shocks to the demand for and supply of gold. The Harvard economists, like the neoclassicists, writing about the heyday of the gold standard, sought a better understanding of the mechanism by which one of the most important structural changes in modern economic history took place-the transfer of real resources associated with massive lending by the mature countries of Western Europe to the developing countries of the New World. The interwar critics, writing after the collapse of the gold standard, yet strongly influenced by its heritage, were concerned with the possibility of restoring the old system. Much of their work reflected skepticism on this score because for them "special circumstances" in the prewar period made the system work: the unique interrelationship between the London gold, securities, and commodities markets that created a "sterling standard"; the commitment by major participants to maintain convertibility as their key policy goal; and relatively free trade and factor mobility. These special circumstances no longer existed. Other interwar critics focused on the negative aspects of the gold standard: the tendency for short-run price instability, the asymmetry between the adjustment mechanism of central and peripheral countries, the conflict between external and internal stability, and the tendency for economic fluctuations to be transmitted internationally by the gold standard. Hence these interwar critics doubted the wisdom of returning to the standard's iron discipline. However, in elaborating proposals for a better system, the consensus favored maintaining a fixed-exchange-rate system based on gold, with expanded national discretionary management and the establishment of a supernational central bank. Finally, in the postwar period, operating in an institutional environment far enough removed from the events before 1914, scholars of the gold standard could objectively ask how the gold standard in its many aspects worked. Armed with new theoretical and statistical tools and new compilations of data, the consensus of this work has been that the international gold standard did function smoothly in the sense of ensuring international price harmony, in allowing the international transfer of resources, and in maintaining balance-of-payments equilibria for most countries over long periods of time, but that many elements of the story-particularly the operation of the price-specie-flow mechanism and the importance of the rules of the game-were subject to doubt. According to the second perspective, the development of the gold standard literature reflected a Kuhnian scientific revolution (Kuhn 1970). Along this line, we start with the development of the classical-gold standard paradigm by the classical economists, culminating in the magnum opus of J. S. Mill. The paradigm was further extended and perfected by the neoclassical economists, especially Irving Fisher. However, anom-

31

The Gold Standard: The Traditional Approach

alies begin to appear by the end of the nineteenth century: the pricespecie-flow mechanism emphasizing the adjustment of relative price levels could not explain the actual adjustment process to international lending in a number of countries; in some cases the mechanism could be detected, in others the adjustment of price levels between countries seemed to be too rapid for the theory; and it appeared that many countries did not follow the rules of the game but engaged in extensive sterilization activities. However, the reaction to these anomalies by Taussig and others was to incorporate them into the theory as special cases. The assault on the classical paradigm began in the interwar period with the grave doubts raised by Keynes, Williams, Cassel, and others, but it was probably the 1937 article by Whale, challenging the whole classical interpretation of how the gold standard worked, that started the revolution. The further revelation of evidence inconsistent with the classical story in the postwar period added ammunition to the case presented by Williamson, Triffin, and finally McCloskey and Zecher. The last authors completely upended the classical paradigm and argued passionately that all aspects of the gold standard could be explained by the newly developed monetary approach. The scientific revolution was complete. In conclusion, we can ask: Is this the end of the gold standard story? McCloskey and Zecher, in tying together much of the unfavorable evidence against the traditional approach and then reinterpreting the facts to be consistent with the implications of the monetary approach to the balance of payments, make a strong case for a successful conclusion, except that the evidence they marshal in favor of their approach, based largely on correlation tests of commodity arbitrage, is neither extensive nor conclusive enough to end the story.

Appendix A

The Classical Economists

In this appendix the writings of eight key economists who first formulated the tenets of the traditional approach to the gold standard are summarized: 4 Cantillon, Hume, Ricardo, Thornton, Mill, Cairnes, Goschen, and Bagehot. Richard Cantillon Richard Cantillon ([1931] 1964), writing in 1755, was one of the first writers to analyze the working of a money economy. Operating within a crude quantity-theory-of-money framework ,5 Cantillon regarded the quantity of money as consisting entirely of specie-gold and silver coins. Gold and silver emerged as money commodities as a result of the evolution of natural market forces-they best satisfied the properties of

32

Michael D. Bordo

money, viz., they are "of small volume, equal goodness, easily transported, divisible without loss, convenient to keep, beautiful and brilliant in the articles made of them and durable almost to eternity" (Cantillon [1931] 1964, p. 111). Moreover, the choice between gold and silver (as well as the desired ratio in a bimetallic system) is determined by market forces-on the demand side by tastes and income, on the supply side by relative scarcity (pp. 97 and 277). In the long run, the world's monetary specie stock as well as its exchange value is determined by the foregone cost in terms of the land and labor required to extract precious metals. 6 In the short run, the two key sources of a nation's money supply are its balance-of-payments surplus and the presence of domestic gold and silver mines (bk. 2, chaps. 6, 7). Perhaps Cantillon's most important contribution was setting out a dynamic version of the quantity theory of money or what has often been referred to as monetary-disequilibrium theory. Cantillon carefully analyzed the dynamic process by which the quantity of money affected economic activity and the price level. An important element of his analysis was the international repercussions in the specie-standard world of fixed exchange rates of domestically induced monetary change: If more money continues to be drawn from the Mines all prices will owing to this abundance rise to such a point that . . . there will be a considerable profit in buying them [goods] from the foreigner who makes them much more cheaply. This will naturally induce several people to import many manufactured articles made in foreign countries, where they will be found very cheap: this will gradually ruin the Mechanics and Manufacturers of the State. (P. 165). That is, domestic inflation, by raising the prices of domestically-produced goods relative to foreign-produced goods (changing the terms of trade), will generate a balance-of-trade deficit. This deficit will induce a specie outflow, a reduction in the domestic money stock, and a reduction in domestic output and prices-the price-specie-flow mechanism. In addition to the terms-of-trade effect, the balance of payments will adjust by a direct-expenditure effect. According to this mechanism, an excess supply of money will cause domestic expenditures to exceed income; some of this expenditure will be made directly on foreignproduced goods (whose prices are determined abroad), leading to a specie outflow: It is usual in States which have acquired a considerable abundance of money to draw many things from neighbouring countries where money is rare and consequently everything is cheap: but as money must be sent for this the balance of trade will become smaller. (P. 169) Cantillon's final main contribution to the traditional view7 was a clear statement of the law of one price-eommodity arbitrage will ensure that

33

The Gold Standard: The Traditional Approach

the prices of similar traded goods will be the same across countries and across regions within countries, allowing for the influence of tariffs and transportation costs. The difference of prices in the Capital and in the Provinces must pay for the costs and risks of transport, otherwise cash will be sent to the Capital to pay the balance and this will go on till the prices in the Capital and the Provinces come to the level of these costs and risks. (P. 151) Moreover, he clearly distinguished between traded and nontraded goods on the basis of trade impediments and transportation costs. Thus the prices of traded goods that are determined abroad will be largely unaffected by domestic monetary conditions, whereas the prices of nontraded goods will respond fully, viz. In England it is always permitted to bring in corn from foreign countries, but not cattle. For this reason however great the increase of hard money may be in England the price of corn can only be raised above the price in other countries where money is scarce by the costs and risks of importing corn from these foreign countries. It is not the same with the price of Cattle, which will necessarily be proportioned to the quantity of money offered for Meat in proportion to the quantity of Meat and the number of Cattle bred there. (P. 179) This suggests that the distinction between the terms of trade and the direct-expenditure mechanisms rests on the distinction between tradedand nontraded-goods prices. The excess-money-induced expenditure falls on both traded and nontraded goods. To the extent the expenditure affects nontraded goods, their prices rise, inducing the substitution of traded goods. To the extent it affects traded goods whose prices are determined abroad, it leads to a direct specie outflow. Presumably, the effect on nontraded-goods prices will be short-lived-until substitution and the decline in the domestic money supply consequent upon the specie outflow have caused the relative prices of traded and nontraded goods to return to their initial equilibrium. David Hume In his essay, "Of the Balance of Trade" ([1752] 1955), Hume is generally believed to have originated the theory of the traditional balance-ofpayments adjustment mechanism of an international specie standard (see Viner 1937, pp. 291-92). According to Hume, a domestic monetary disturbance such as a sudden decrease in the specie stock will lead to a proportional decline in all prices and wages, a consequent decline in the prices of exports relative to the prices of imports, a balance-of-payments surplus, a specie inflow, and an increase in the domestic stock of specie.

34

Michael D. Bordo

Suppose four-fifths of all the money in Great Britain to be annihilated in one night, ... what would be the consequence? Must not the price of all labour and commodities sink in proportion . . . ? What nation could then dispute with us in any foreign market, or pretend to navigate or to sell manufactures at the same price, which to us would afford sufficient profit? In how little time, therefore, must this bring back the money which we had lost, and raise us to the level of all the neighboring nations? Where, after we have arrived, we immediately lose the advantage of the cheapness of labour and commodities; and the farther flowing in of money is stopped by our fulness and repletion. (Hume [1752] 1955, pp. 62-63) Thus, the domestic specie stock in a country (or province within a country) under a specie standard will be automatically regulated by its balance of payments. Moreover, this mechanism will ensure that each nation's (province's) price level will be consistent with adherence to the specie standard. In addition, variations in the exchange rate within the gold points will act as an additional factor to correct balance-of-payments disequilibria. There is another cause, though more limited in its operation, which checks the wrong balance of trade, to every particular nation to which the kingdom trades. When we import more goods than we export, the exchange turns against us, and this becomes a new encouragement to export; as much as the charge of carriage and insurance of the money which becomes due would amount to. For the exchange can never rise but a little higher than that sum. (P. 64n) Hume also discussed the law of one price, viz. any man who travels over Europe at this day, may see, by the prices of commodities, that money ... has brought itself nearly to a level; and that the difference between one kingdom and another is not greater in this respect, than it is often between different provinces of the same kingdom.... The only circumstances which can obstruct the exactness of these proportions [between money and real economic activity] is the expense of transporting the commodities from one place to another. (P. 66) Some writers have argued that there appeared to be an inconsistency between the law of one price, which suggests rapid adjustment of commodity prices through arbitrage, and the price-specie-flow mechanism which suggests noticeable time lags. 8 However, as is made most clear by Ricardo, the price-specie-flow mechanism is a reflection of arbitrage in the gold market, which, because of its special properties, is more rapid than arbitrage in other markets. Other prices are kept in line through the influence of changes in the quantity of gold as money. In accordance with this interpretation, Hume may have regarded the law of one price as a

35

The Gold Standard: The Traditional Approach

long-run equilibrium condition in all markets with the price-specie-flow mechanism as the means to achieve that result. David Ricardo Gold as the Standard

Ricardo ([1811] 1951, [1816] 1951), in the classical tradition, viewed the world quantity of specie as determined in the long run by cost of production (Ricardo [1811] 1951, p. 52). The quantity of precious metals used as money in each country depended "first, on its value;-secondly, on the amount or value of the payments to be made;-and, thirdly, on the degree of economy practiced in effecting those payments" (Ricardo [1816] 1951 ,po 55). Each country's share of the world specie stock and hence the natural distribution of precious metals is determined by its share of world real income and factors determining velocity (or the demand for money). The precious metals employed for circulating the commodities of the world, ... have been divided into certain proportions among the different civilized nations of the earth, according to the state of their commerce and wealth, and therefore according to the number and frequency of the payments which they had to perform. While so divided they preserved everywhere the same value, and as each country had an equal necessity for the quantity actually in use, there could be no temptation offered to either for their importation or exportation. ([1811] 1951, p. 52) The choice of gold and silver as monetary standard was "the comparative steadiness in the value of the precious metals, for periods of some duration" ([1816] 1951, p. 55). In the choice between gold and silver, gold has in its favor "its greater value under a smaller bulk" which "qualifies it for the standard in an opulent country," but coupled with the disadvantage that it is subject to "greater variations of value during periods of war or extensive commercial discredit." Silver he viewed as "much more steady in its value, in consequence of its demand and supply being more regular." The only objection to its use as a standard "is its bulk, which renders it unfit for the large payments required in a wealthy country" (p. 63). However, using both precious metals as the standard has the disadvantage that prices expressed in terms of gold and silver will vary with changing demand and supply conditions for each commodity. To avoid this instability, Ricardo suggested the substitution of paper money and "by the judicious management of the quantity, a degree of uniformity ... is secured" (pp. 57-58).9 In addition, the substitution of bank notes for specie "enables us to turn the precious metals (which, though a very

36

Michael D. Bordo

necessary part of our capital, yield no revenue) into a capital which will yield one" ([1811] 1951, p. 55). However, "the issuers of paper money should regulate their issues solely by the price of bullion" ([1816] 1951, p. 64). Indeed, for Ricardo, the key advantage of the gold standard was that adherence to the standard acted as a check against the overissue of paper money-it provided discipline (p. 78). Balance-oJ-Payments Adjustment Mechanism

Beginning with the natural distribution of precious metals, Ricardo demonstrated how this distribution would be neutral with respect to monetary changes. Any movement away from the natural distribution would be corrected by the price-specie-flow mechanism. Thus if a gold mine were discovered in one country the currency of that country would be lowered in value in consequence of the increased quantity of the precious metals brought into circulation, and would therefore no longer be of the same value as that of other countries. Gold and silver, whether in coin or in bullion, obeying the law which regulates all other commodities, would immediately become articles of exportation; they would leave the country where they were cheap, for those countries where they were dear, and would continue to do so, as long as the mine should prove productive, and till the proportion existing between capital and money in each country before the discovery of the mine, were again established, and gold and silver restored every where to one value. In return for the gold exported, commodities would be imported; and though what is usually termed the Balance of Trade would be against the country exporting money or bullion, it would be evidence that she was carrying on a most advantageous trade, exporting that which was no way useful to her, for commodities which might be employed in the extension of her manufactures, and the increase of her wealth. (Ricardo [1811] 1951, p. 54)10 Thus gold as a commodity flows to the market with the highest price and thereby maintains price uniformity between nations. As long as different countries (regions within countries) fixed the prices of their currencies in terms of gold (specified a gold weight of their coins), then arbitrage allowing for transportation costs would always keep gold prices in line. 11 This principle, referred to as the law of one price, would hold for all traded commodities, and hence in logic there was no reason why commodity arbitrage would not occur for all commodities. However, for Ricardo and other classical economists, arbitrage took place primarily in gold because of its special properties as money and because it involved the lowest arbitrage costs. Consequently since all other commodity prices were set in terms of gold-the numeraire of the system-gold flows would then keep all prices in line for countries (regions) on a gold standard. However, some prices would react more quickly to the mechanism than

37

The Gold Standard: The Traditional Approach

others, specifically prices of tradable goods, and this quick reaction probably explains the later emphasis in the literature on the role of changes in sectional prices and the terms of trade (see appendix B on Marshall). Ricardo clearly distinguished between the adjustment mechanism under the gold standard and under irredeemable paper money. An issue of convertible paper currency, e.g., Bank of England notes, will displace, through the balance of payments, a corresponding amount of specie (p. 67). However, as long as convertibility in terms of gold is maintained, the domestic price level will not be affected. 12 Once all specie is displaced and convertibility suspended, however, domestic prices will rise, and a depreciated exchange rate will be the indicator of overissue (pp. 58-59, 63-64, 72-78). In the Bullion report ([1810] 1978, pp. ccxvii-ccxxi), a distinction is made between the real exchange rate-determined by the ratio of the mint prices of gold between two gold standard countries-and the market exchange rate or the computed par. The market exchange rate includes both the influence of real factors causing a divergence from par within the gold points and the depreciation of the exchange rate (premium on the price of gold) due to a rise in the price level. Thus for Ricardo, the increase in irredeemable paper following the suspension of payments in 1797 was responsible for both a rise in all commodity prices in England with no corresponding rise in prices abroad and the depreciation of the pound (pp. ccxiv-ccxv). Proposals for Monetary Reform

As mentioned above, Ricardo viewed a properly regulated convertible paper currency as superior to a precious-metals standard. However, he believed that convertibility into gold was necessary to avoid the temptation of overissue (Ricardo [1816] 1951, p. 69). And in the Bullion report ([1810] 1978, p. ccxlvi) a strong case is made in favor of a gold standard rule and against discretionary monetary policy: The most detailed knowledge of the actual trade of the country, combined with the profound science in all the principles of money and circulation, would not enable any man or set of men to adjust, and keep always adjusted, the right proportion of circulating medium in a country to the wants of trade. When the currency consists entirely of the precious metals, or of paper convertible at will into the precious metals, the natural process of commerce, by establishing exchanges among all the different countries of the world, adjusts, in every particular country, the proportion of circulating medium to its actual occasions, according to that supply of the precious metals which the mines furnish to the general market of the world. The proportion, which is thus adjusted and maintained by the natural operation of commerce,

38

Michael D. Bordo

cannot be adjusted by any human wisdom or skill. If the natural system of currency and circulation be abandoned, and a discretionary issue of paper money substituted in its stead, it is vain to think that any rules can be devised for the exact exercise of such a discretion. As a remedy for defects of a purely metallic standard with no discretion for central bankers, he proposed a convertible banknote issue backed by bullion (Ricardo [1816] 1951, p. 66). Free export and import of bullion would be permissible. Under this scheme the costs of frequent conversions of coin into bullion would be eliminated, but the risks of attempted conver~ion of banknotes into specie in a money-market panic would not be. 13 Henry Thornton The Gold Standard

Like Ricardo, Thornton ([1802] 1978, p. 21a) viewed convertibility as a key feature of the gold standard. 14 Also, like Ricardo, he viewed the substitution of paper money for specie up to the point of convertibility as a social saving. However, he extended the analysis to consider the effects of a domestic issue of bank notes on the world price level. First, since the issue of paper money would displace specie in the domestic circulation, specie would be exported abroad, leading to an increase in the world money stock and a rise in world prices. The country displacing specie would thereby raise its capital stock (Thornton [1802] 1978, pp. 269-70). Second, to the extent the use of paper money reduced the demand for gold as money, and hence the price of bullion, this development would cause "those mines which have not yielded any rent, to be no longer worked; and the supply of gold ... to be in consequence, somewhat reduced." The process would continue until all mines will be unable to defray the charge of extracting the ore, except those which now yield the very highest rent. At this point the fall will necessarily stop . . . gold and silver must continue to bear that price, or nearly that price, ... at which they are now exchangeable for commodities. (P. 266) Balance-oj-Payments Adjustment Mechanism

Thornton clearly elucidated the price-specie-flow mechanism. The primary mechanism of adjustment following an increase in the domestic money supply (an increase in Bank of England note issue) is the effect on prices at home relative to those abroad. It is obvious, that in proportion as goods are rendered dear in Great Britain, the foreigner becomes unwilling to buy them, the commodities of other countries which come into competition with our's obtaining a

39

The Gold Standard: The Traditional Approach

preference in the foreign market; and, therefore, that in consequence of a diminution of orders from abroad, our exports will be diminished . . . . But not only will our exports lessen ... ; our imports also will increase; for the high British price of goods will tempt foreign commodities to come in nearly in the same degree in which it will discourage British articles from going out. (Thornton [1802] 1978, p. 198) The resultant deficit in the balance of payments will however be offset to a certain extent by changes in the exchange rate within the gold points. However, to the extent an unfavorable balance persists and the exchange rate falls to the specie-export point, this will lead to a specie outflow until trade balance is restored (pp. 145-47). Finally, Thornton discussed an alternative adjustment mechanismthe direct-expenditure-income mechanism: There is in the mass of the people . . . a disposition to adapt their individual expenditure to their income. Importations conducted with a view to the consumption of the country into which the articles are imported . . . are limited by the ability of the individuals of that country to pay for them out of their income.... If, therefore, ... the value of the annual income of the inhabitants of a country is diminished, either new economy on the one hand, or new exertions of individual industry on the other, fail not, after a certain time, in some measure, to restore the balance. And this equality between private expenditures and private incomes tends ultimately to produce equality between the commercial exports and imports. (Pp. 142-43) Law of One Price

According to Thornton, different prices within Great Britain for identical goods cannot exist as long as country bank notes are convertible into Bank of England notes. A very considerable advance in the price of commodities bought and sold in one quarter of this kingdom, while there was no such rise in any other, was not supposable; because the holders of the circulating medium current in the spot in which goods were . . . rendered dear, would exchange it for the circulating medium of the part in which they were assumed to be cheap, and would then buy the commodities of the latter place, and transport them to the former, for the sake of the profit on the transaction. Moreover, the law of one price can be extended from one kingdom to the whole world as long as currencies are convertible. We may ... extend our views, and conceive of Europe, and even of the world, as forming one great kingdom, over the whole of which goods pass and repass . . . nearly in the same manner in which they spread themselves through this single country.

40

Michael D. Bordo

However, prices can differ between countries within the limits of the gold points: But British paper is not exchangeable for the circulating medium of the continent, unless a discount ... be allowed. Of this fluctuating discount ... the variations in the course of exchange are the measure. (Thornton [1802] 1978, pp. 260-61) Finally, he argued that under a specie standard, one country alone can affect world (traded) goods prices only to the extent that it has monopoly power in their production. Great Britain may have this power in the short run, but in the long run the existence of substitutes will diminish the power. Policy Considerations

Thornton was one of the first to recognize the possibility of a conflict between external and internal policy goals. In the case of an unfavorable balance of trade caused by an exogenous event such as a harvest failure, the central bank could respond to the resulting gold outflow by reducing the money supply, but taking such a course of action might depress domestic activity. Hence, it would be prudent to maintain an adequate gold reserve to permit the bank to increase its loans while losing gold. IS John Stuart Mill Perhaps the clearest statement of the traditional approach to the gold standard is in J. S. Mill's Principles of Political Economy ([1865] 1961). Much of the subsequent literature is either a refinement of Mill or attempts to verify his theory. In discussing Mill, I focus on three topics: (a) gold as a commodity money; (b) the natural distribution of precious metals and the adjustment mechanism; (c) the distinction between real and nominal disturbances. Gold as a Commodity Money

Mill carefully analyzed the economics of commodity money, according to which market forces ensure a determinate money stock and price level. In the longrun, according to Mill, the exchange value of gold-what it will purchase in terms of other goods and services or the inverse of the price level-will be equal to its cost of production-"the cost in labor and expense, at the least productive sources of supply which the then existing demand makes it necessary to work" (Mill [1865] 1961, p. 502). The conformity will be maintained by deviations in gold output in response to variations in the exchange value of gold relative to its cost of production. However, because the existing stock of gold is large relative to additions to the stock, it takes a long time for full adjustment to take place.

41

The Gold Standard: The Traditional Approach

And hence the effect of all changes in the conditions of production of the precious metals are at first, and continue to be for many years, questions of quantity only, with little reference to cost of production. (P. 503) Thus, in the short run, the price level is determined by the relationship between the demand for and supply of money, and only in the long run is it determined by cost of production. Mill then compared a bimetallic standard to a single metallic standard-"There is an obvious convenience in making use of the more costly metal for larger payments and the cheaper one for smaller" (p. 507), but the arrangement only works if the ratio of the two metals is consistent with their relative costs of production. If relative values change, e.g., the value of gold rises relative to silver, this change will cause replacement of gold by silver coins and the melting of gold coins. Mill therefore preferred a limping standard. The advantage without the disadvantages of a double standard, seems to be best obtained by those nations with whom one only of the two metals is a legal tender, but the other is also coined (the more costly metal), and allowed to pass for whatever value the market assigns to it. (P. 509) Finally, Mill, like his predecessors, viewed the substitution of paper money for specie up to the point of convertibility as "a national gain," but beyond that "a form of robbery" (p. 551). Moreover, the social saving from the issue of paper money is transmitted to the rest of the world. The specie displaced through the balance of payments will initially lead to a rise in the world price level, but ultimately to a reduction in gold output. The world price level will then return to normal. 16 Natural Distribution of Precious Metals and Balance-oJ-Payments Adjustment Mechanism

In the long run, each country in the world will have that quantity of money to effect exchange consistent with keeping its value in terms of its cost of production, hence the natural distribution of precious metals across countries is determined by real forces. 17 Mill then compared the international adjustment mechanism under a barter system with that under a money system. Starting from a state of stable equilibrium where the value of exports equals the value of imports "the process by which things are brought back to this state when they happen to deviate from it, is, at least outwardly, not the same in a barter system and in a money system." Under barter, a country which wants more imports than its exports will pay for, must offer its exports at a cheaper rate, as the sole means of creating a

42

Michael D. Bordo

demand for them sufficient to reestablish the equilibrium. When money is used, the country ... takes the additional imports at the same price as before, and as she exports no equivalent, the balance of payments turn against her; the exchange becomes unfavourable, and the difference has to be paid in money. This is in appearance a very distinct operation from the former. (Mill [1865] 1961, pp. 619-20) However, this difference is only apparent; in both cases prices must adjust to restore equilibrium. In the case of a money economy: When. . . the state of prices is such that the equation of international demand cannot establish itself, the country requiring more imports than can be paid for by exports; it is a sign that the country has more of the precious metals ... than can permanently circulate, and must necessarily part with some of them before the balance can be restored. The currency is accordingly contracted: prices fall, and among the rest, the prices of exportable articles; for which, accordingly, there arises, in foreign countries, a greater demand: while imported commodities have possibly risen in price, from the influx of money into foreign countries, and at all events have not participated in the general fall. (Pp. 620-21) Thus, through the price-specie-flow mechanism the same results will be achieved as under barter, with the only difference that relative prices adjust as a consequence of changes in the quantity of money induced by specie flows rather than adjust directly. "In international, as in ordinary domestic interchanges, money is to commerce only what oil is to machinery, or railways to locomotion-a contrivance to diminish friction" (p. 622). Mill made a clear distinction between temporary and permanent disturbances to the balance of payments. When a disturbance is temporary, most of the adjustment takes place through variations in the exchange rate, within the gold points. Thus the deficit will be "soon liquidated in commodities, and the account adjusted by means of bills, without the transmission of any bullion" (pp. 617-18). In the case of a permanent disturbance to the balance of payments, the adjustment must be made by "the subtraction of actual money from the circulation of one of the countries" (p. 618).

Distinction Between Real and Nominal Disturbances Since the natural distribution of precious metals is determined by real forces, changes in that distribution will only follow from a change in real forces. Thus Mill made a clear distinction between the effects of a real disturbance, such as a remittance from one country to another, and a purely nominal disturbance, such as the discovery of a hoard of treasure. In the first case, he starts from a state of equilibrium, after the first remittance is

43

The Gold Standard: The Traditional Approach

made in money. This lowers prices in the remitting country, and raises them in the receiving. The natural effect is that more commodities are exported than before, and fewer imported, and that ... a balance of money will be constantly due from the receiving to the paying country. When the debt thus annually due to the tributary country becomes equal to the annual tribute. . . no further transmission of money takes place; the equilibrium of exports and imports will no longer exist, but that of payments will; the exchange will be at par, the two debts will be set off against one another, and the tribute or remittance will be virtually paid in goods. In addition, the terms of trade will turn against the paying country and in favor of the receiving country. "The paying country will give a higher price for all that it buys from the receiving country, while the latter. . . obtains the exportable produce of the tributary country at a lower price" (Mill [1865] 1961, pp. 627-28) In contrast, a disturbance in the money market changes the world price level with no real effects. Thus the discovery of a hoard of treasure in a country with a purely metallic currency will raise prices there, check exports and encourage imports, leading to a balance-of-payments deficit and diffusion of the new stock of money over the commercial world; consequently the country's price level will revert to its previous level. John E. Cairnes Cairnes in his essays on gold (first published in 1858-60, reprinted in 1873) used the monetary history of the Australian colonies following the gold discoveries of 1851 to test some of the principal conclusions of classical monetary and trade theory. 18 In particular, he tested the ability of the quantity theory of money to predict the comparative static effects of the gold discoveries on the money supplies and price levels of the major countries of the world, and the ability of the Hume-RicardoThornton-Mill adjustment mechanism to predict the distribution of precious metals. 19 Starting with the long-run cost-of-production theory of money, Cairnes argued that "the rate of gold earnings [is] ... the circumstance which, in the final resort, regulates the value of the metal and sets the limit beyond which depreciation cannot permanently pass" (Cairnes [1873] 1965, p. 41); and since gold earnings in Australia increased by 50 percent, i.e., the cost of gold fell by one-half, he expected the price level in Australia to double. The inflation process would spread across the world until either prices doubled or the cost of producing gold rose. Finally, as prices in the rest of the world rose, gold would become a less profitable commodity to produce and export until, in the limit, when the price of Australian imports increased by the amount of the fall in the cost of gold, Australia

44

Michael D. Bordo

would cease to have a comparative advantage in gold and would divert resources back to agriculture (p. 48). In 1872, twelve years after writing his essays on gold, Cairnes found, as he had predicted, that (a) exhaustion of the mines and the resulting rising costs of production and (b) rises in the price of imports led to a considerable shift of resources out of gold production and back into agriculture. Next Cairnes examined the factors that determine the pace of price adjustment across countries. He argued that the rise in prices would be most rapid in countries with the most advanced banking systems-the more developed the system of banking and credit, the smaller the amount of new gold required to effect a given rise in prices. Furthermore, since Australia and California conducted most of their trade with the United States and England, most of the new gold would tend to go first to these countries, and from there it would spread to the continent of Europe and to Asia (pp. 67-68).20 In general, Cairnes found the evidence agreeable to his predictionsprices increased most in Australia, followed by price increases in descending order of magnitude in Great Britain and the United States, the Continent, and finally Asia. Moreover, in a postscript, Cairnes reported that (a) world gold production doubled by 1868 and (b) most of the gold ended up in France and India, although much of it passed through England (pp. 160-65). George J. Goschen The clearest statement of the classical position on short-term capital flows in the balance-of-payments adjustment mechanism is in Goschen ([1892] 1978).

Changes in interest rates and short-term capital flows facilitate the balance-of-payments adjustment mechanism since money will be dear and scarce in the country which owes much to foreign creditors, and plentiful in that which has exported much; and, high interest will be attracting money to that quarter whence specie is flowing out in payment of foreign debts. [an] adverse balance of trade will. . . render the bills on the country which is most in debt difficult of sale, and tend to compel it to export specie; whereas the high rate of interest, which is generally contemporaneous with a drain ... of specie, will revive a demand for bills on this same country, and enhance their value in other quarters, for there will be a general desire to procure the means of remitting capital to that market where it commands the highest value. (Goschen [1892] 1978, p. 127)

Thus,

45

The Gold Standard: The Traditional Approach

where a considerable efflux of specie is taking place, the rate of interest will rise in the natural course of things. The abstraction caused by the bullion shipments will of itself tend to raise that rate. (P. 132) Moreover, a country can finance a temporary balance-of-payments deficit by borrowing abroad. Finally, arbitrage in the securities market will ensure that interest rates for a similar class of bills will be equal between financial centers, account being taken of transportation costs and exchange risks. If at any time the rate of interest here falls below that which rules on the continent, it is inevitable that the whole mass of these bills will at once be sent to London, and be discounted there at the cheaper rate, so that the proceeds may be remitted in gold to the continent to be invested there in local securities at the supposed higher rate. (P. 138) Walter Bagehot and the Responsibility of the Bank of England The pure gold standard in England-when the money supply consisted in large part of specie, and variations in its amount were determined mainly by the balance of payments-became a managed gold standard in the course of the nineteenth century-when the money supply consisted primarily of convertible notes and deposits, and variations in its amount were determined by operations of the Bank of England conforming to the external constraint of the gold standard. The evolution to a managed gold standard evoked considerable debate in the ecopomic literature. 21 In the decades after the restoration of convertibility in 1821, British monetary history was punctuated by a series of monetary crises in 1825, 1836, 1838, 1847, 1859, and 1866 (see Viner [1937] 1975, pp. 218-20). These crises occurred when the necessary contraction of the money stock consequent upon a specie outflow (an external drain) coincided with a demand by deposit and note holders for specie currency (an internal drain). In such a situation it was difficult for the Bank to maintain convertibility of its notes without resort to special measures. The Bank Charter Act of 1844 was an attempt to rectify the situation by dividing the Bank of England into an Issue Department and a Banking Department. The former was charged with the responsibility of maintaining the gold standard link by following the "currency principle": The note circulation should fluctuate one-for-one with changes in the Bank's holdings of gold. 22 The latter was to follow the principles of a profit-maximizing banking enterprise, accepting deposits and making discounts. The arrangement was criticized on two grounds: (1) the currency principle ignored the role of deposits as an increasingly important component of the money stock;23 and (2) the Banking Department in operating on a sound commercial-banking basis could not act responsibly as a central bank. In that role, it had to maintain a gold reserve large enough

46

Michael D. Bordo

to protect the rest of the banking system from the effects of both internal and external specie drains. 24 This criticism culminated in the 1860s with the formulation by Walter Bagehot, the influential editor of the Economist, of the "responsibility doctrine" and the establishment of guidelines for a central bank under a gold standard. In Lombard Street ([1873] 1969), Bagehot clearly set out the conflict between the private concern of the Banking Department to reduce the holding of specie reserves to minimize foregone interest costs and its public concern to hold larger reserves (p. 38). Bagehot argued for the "responsibility doctrine"-that the Bank of England must fulfill its special obligations as a bankers' bank and as holder of the nation's specie reserves. In consequence, he established clear guidelines for central-bank behavior in times of crisis. First, in the case of a purely external drain, the Bank of England requires the steady use of an effectual instrument. That instrument is the elevation of the rate of interest. If the interest of money is raised, it is proved by experience that money does come to Lombard Street, and theory shows it ought to come. (P. 46) The rise in Bank rate will initially lead to a short-term capital inflow and a gold inflow. And there is also a slower mercantile operation. The rise in the rate of discount acts immediately on the trade of this country. Prices fall here; in consequence imports are diminished, exports are increased, and, therefore, there is more likelihood of a balance in bullion coming to this country after the rise in the rate than there was before. (P. 47) Second, the best way for the Bank. . . to deal with a drain arising from internal discredit is to lend freely. . . . A panic. . . is a species of neuralgia, and according to the rules of science you must not starve it. The holders of the cash reserve must be ready not only to keep it for their own liabilities, but to advance it most freely for the liabilities of others. (Pp. 48, 51) In brief, the central bank has a responsibility to act as lender-of-Iastresort. Finally, in the case of both an internal and an external drain, the central bank should follow what has come to be known as Bagehot's rule. We must look first to the foreign drain, and raise the rate of interest as high as may be necessary. Unless you can stop the foreign export, you cannot allay the domestic alarm. The Bank will get poorer and poorer, and its poverty will protect or renew the apprehension. And at the rate of interest so raised, the holders ... of the final Bank reserve must

47

The Gold Standard: The Traditional Approach

lend freely. Very large loans at very high rates are the best remedy for the worst malady of the money market when a foreign drain is added to a domestic drain. (P. 56) Bagehot was aware of the limitations to the Bank's control mechanism. The Bank could only temporarily affect market interest rates. The initial effect of an increase in central-bank lending is to lower interest rates, but it also leads to an "increase of trade and increase of prices." The rise of prices and trade leads to an increase in the demand for loanable funds and also to an increase of imports and a decrease in exports. The resultant balance-of-trade deficit leads to a specie outflow and a reduction in reserves, and hence the rate of interest must be raised (pp. 12-14). In conclusion, Bagehot proposed several remedies for reform. Of key importance, "there should be a clear understanding between the Bank and the public that, since the Bank holds our ultimate banking reserve, they will recognize and act on the obligations which this implies" (p. 70); and the Bank should hold an adequate reserve to be determined by "experience" (p. 304).25 Following the publication of Lombard Street, the Bank of England's special position as both lender-of-Iast-resort and holder of the nation's reserve was recognized, but increasing stress was laid on the Bank's vulnerability in view of London's growing international liabilities (see Sayers 1957). Though Bagehot and other writers urged the Bank to meet the problem by maintaining a larger specie reserve (an approach taken by other countries), the Bank's solution was to alter its discount rate whenever its international reserves were affected, thereby primarily influencing short-term capital movements. The Bank also supplemented the use of Bank rate with other tools of monetary policy, especially open-market operations, and in addition learned to protect its reserves by using special techniques referred to as "gold devices" (see Viner [1937] 1975, p. 277; Sayers 1936, chap. 3). Thus the ultimate answer to Bagehot's concern was the use "of a powerful bank rate weapon with a 'thin film of gold' " (Sayers 1951, p. 116).

Appendix B

The Neoclassical Economists: Extension of the Traditional Approach

In this appendix, attention will center on the works of three neoclassical economists-Alfred Marshall, Irving Fisher, and Knut Wicksell-who wrote extensively on the gold standard and whose contribution can be viewed as a refinement or extension of the traditional approach.

48

Michael D. Bordo

Alfred Marshall Marshall's contribution (1923, 1926) to the traditional approach can be classified under two headings: the monetary standard and proposals for reform and the balance-of-payments adjustment mechanism.

The Monetary Standard and Proposals for Reform For Marshall, the primary purpose of a monetary standard is to ensure price stability: Violent fluctuations of prices are less distasteful to the heads of business enterprises than a gradual fall of prices. But I believe they are far more injurious both physically and morally to the community at large. (Marshall 1926, p. 20) Nevertheless, a gold or silver standard may not be the best mechanism to ensure price stability, at least in the long run. Gold and silver, separately or conjointly, can set good standards of general purchasing power, in regard to obligations and business transactions, which do not range over more than a few years; but obligations which range over long periods, call for standards that are not dependent on the hazards of mining. (Marshall 1923, p. 52) Marshall admitted that in past years, technical advances in mining precious metals had kept pace with technical advances elsewhere, so the real cost of producing the precious metals had remained constant over long periods of time, which was no accident; but he still believed that as the arts of life progress. . . man must demand a constantly increasing precision from the instruments which he uses, and from money among others: and he is beginning to doubt whether either gold or silver, or even gold and silver combined, give him a sufficiently stable standard of value for the ever widening range of space and time over which his undertakings and contracts extend . . . [indeed] gold and silver have had a less stable value, during the history of the world, than has accrued to those staple grains, which have supplied the chief means of supporting life to the great mass of the people in every age. (Pp. 53-54) Over shorter periods of time, however, gold and silver represent good monetary standards because changes in the stock are small relative to the existing stock (1926, p. 177). In addition, following the classical tradition, Marshall regarded gold as an inefficient form of money tying up scarce resources. Civilization had advanced sufficiently for an expanded role for convertible paper (p. 137). Finally, following Mill, he opposed bimetallic schemes based on a fixed ratio of gold to silver on the ground that changes in relative costs of production would lead to continuous shifts towards the lower-cost metal,

49

The Gold Standard: The Traditional Approach

thus producing more instability than reliance on one metal alone (1923, p.63). To solve the shortcomings of reliance on precious metals as a monetary standard, Marshall proposed two alternative schemes-symmetallism and a tabular standard. Under the symmetallic scheme, currency would be exchangeable for a combination of gold and silver bullion bars in fixed proportions. A gold bar of 100 grammes, together with a silver bar, say, 20 times as heavy, would be exchangeable . . . for an amount of the currency which would be calculated and fixed once for all when the scheme was introduced.... Anyone who wanted to buy or sell gold or silver alone in exchange for currency could get what he wanted by exchanging gold for silver, or silver for gold, at the market rate. Government fixing its own rates from day to day, so as to keep its reserve of the two metals in about the right proportion, might safely undertake this exchange itself, and then anyone could buy or sell either gold or silver for currency in one operation. (1926, p. 29) The scheme would provide a better monetary standard than Ricardo's gold-bullion-reserve scheme, because it causes the value of legal tender money to vary with the mean of the values of both of these metals ... and because it would be convenient both to those countries which now chiefly use gold and to those which now chiefly use silver. (P. 28). In contrast, the tabular scheme, similar to that of Fisher (see p. 53 below) would separate the standard of value from the medium of exchange. Under this scheme, long-term contracts would be tied to "an official index number, representing average movements of the prices of important commodities" (1923, p. 36). In addition, he proposed regulation of an inconvertible currency so that "the value of a unit of it is maintained at a fixed level," based on an index number (p. 50). Finally, Marshall stressed the need for an international currency or else for the international harmonization of monetary policies: There is a real, though very slow moving, tendency for national interests to overrule provincial interests, and international interests to overrule national, and I think the time will come at which it will be thought as unreasonable for any country to regulate its currency without reference to other countries as it will be to have signalling codes at sea which took no account of signalling codes at sea of other countries. (1926, p. 135) Balance-oJ-Payments Adjustment Mechanism

Following Ricardo, Marshall argued that gold flows reflect arbitrage in a widely traded commodity (gold)

50

Michael D. Bordo

[so] as to bring gold prices at the seaboards of the two countries to equality, allowance being made for carriage. If they are higher in A than in B, there will be a small temporary bounty on exportation from B to A corresponding to this difference, which must always be small. Bills drawn in B on A will multiply, and, specie point being reached, gold will go from A to B till prices in B are as high as in A. (Marshall, 1926, p. 170) However, he questioned why gold has to be the commodity used to settle payments imbalances. Thus in the case of a balance-of-payments deficit the value of gold, as a means of purchasing foreign commodities by being exported in exchange for them, will rise so much that it will be profitably exported for the purpose: ... But under these conditions merchants are likely to look around them, and see whether there is not some other thing which the country does not produce herself, and therefore does not habitually export; but which could under the circumstances be marketed profitably abroad. (1923, p. 153) For example, the balance of payments could be settled by the exportation of lead once "lead point" has been reached. The choice of which commodity is used depends partly on its portability and partly on the extent of the market which it finds in either country. The power of gold for this purpose is therefore of primary importance between two countries which have a gold currency, for gold has in each a practically unlimited market. (1926, p. 172).26 Marshall then discussed the internal adjustment mechanism to an external disturbance such as a gold outflow. The gold outflow would lower the Bank of England's reserves, the Bank would respond by raising its discount rate, and "the result would be a check to speculative investments, a diminished demand for commodities, and a fall of prices" (p. 158).27 At the same time, the rise in the discount rate, if not matched by an equal rise in the discount rate abroad, would attract gold from abroad (p. 160). Finally, he discussed the role of a force that would weaken the traditional adjustment mechanism-the international integration of markets for securities and commodities: The growing tendency of intercommunication has shown itself in the discount market more than in any other; fluctuations in the price of wheat are being held in check by the growing internationality of the wheat market; but the discount market is becoming international more rapidly even than the wheat market. (Pp. 127-28)

51

The Gold Standard: The Traditional Approach

Thus in Marshall's view, more of the adjustment to balance-of-payments disequilibrium takes place through capital flows than through the arbitrage of traded goods, with less of the burden placed on gold flows. Irving Fisher I focus on three important aspects of Fisher's (1920, [1922] 1965) treatment of the gold standard: his exposition of the operation of a commodity money standard, his discussion of the international adjustment mechanism, and his criticism of the gold standard and advocacy of a "compensated dollar. "28 The Commodity Theory of Money Fisher's exposition of the working of a commodity money standard is perhap the most lucid extension of Mill's theory. First Fisher demonstrated how, under a gold standard with unrestricted coining and melting, the price of gold bullion would conform to the price of coin, allowing for seigniorage. Thus, in the case of unrestricted coinage, if the price of gold bullion exceeds that of coin, gold users such as jewellers will melt coin into bullion; in the opposite situation, bullion owners will take bullion to the mint and have it coined. The effect of melting the coin will decrease the stock of gold coins relative to bullion, reducing the value of gold as bullion relative to gold as money, thus lowering the price level and restoring equality between bullion and money. In the opposite case, the effect of minting bullion into coin will restore equilibrium (Fisher [1922] 1965, p. 97). Next Fisher demonstrated how the world gold stock and the world monetary gold stock are influenced by the production and consumption of gold: As the stock of bullion and the stock of money influence each other, so the total stock of both is influenced by production and consumption. The production of gold consists of the output of the mines which constantly tends to add to the existing stocks both of bullion and coin. The consumption of gold consists of the use of bullion in the arts. (P.99) He then made the analogy to a reservoir, "production would be the inflow from the mines, and consumption the outflow to the arts, by destruction and loss" (p. 99). Gold production is regulated by the relationship between "the estimated marginal cost of production" and the purchasing power or exchange value of gold. He assumed that gold mining is an increasing-cost industry. Thus, gold production will always tend toward an equilibrium in which the marginal cost of production will . . . be equal to the value of the product.... If [the] purchasing power of gold is above the cost of

52

Michael D. Bordo

production in any particular mine, it will pay to work that mine.... Thus the production of gold increases or decreases with an increase or decrease in the purchasing power of gold. (Pp. 101-3). And the purchasing power of gold in turn would vary inversely with the prices of other goods. The consumption of gold "in the arts"-nonmonetary uses of gold-is related to "consumption for monetary purposes"-monetary uses of gold-by a comparison of the purchasing power of gold with the "marginal utility of what is consumed." Thus consumption of gold-the diversion of gold from monetary to nonmonetary uses-will be "stimulated by a fall in the value (purchasing power) of gold, while the production of gold is decreased" (p. 104). The two forces of production and consumption, operating in opposing directions, regulate the monetary gold stock and hence the price level. In the case of increased gold production due to the discovery of new mines, the increase in production will lead to a filling up of "the currency reservoir," and "a decrease in the purchasing power of money. This process will be checked finally by the increase in consumption. And when production and consumption become equal, an equilibrium will be established" (pp. 108-9).

International Adjustment Mechanism Fisher effectively argued that for a small open economy that is part of an international gold standard, as is the case for one state within the United States, the money supply is not an independent variable, but is determined by the need for the domestic price level to conform to foreign price levels. However, he preserved the classical quantity-theory notion of causality between the quantity of money and the price level (see Girton and Roper 1978). The price level in an outside community is an influence outside the equation of exchange of that community, and operates by affecting its money in circulation and not by directly affecting its price level. The price level outside of New York City, for instance, affects the price level in New York City only via changes in the money in New York City. Within New York City it is the money which influences the price level, and not the price level which influences the money. The price level is effect and not cause. (Fisher [1922] 1965, p. 172) Following the tradition of Ricardo and Marshall, Fisher argued that the force of arbitrage would tend to produce equality in the prices of traded commodities, but to the extent that international (and interlocal) trade does not bring about uniformity of price levels/9 "it will ... produce an

53

The Gold Standard: The Traditional Approach

adjustment of these levels toward uniformity by regulating ... the distribution of money" (p. 93). Thus gold flows because it is the most efficient international medium of exchange and because it affects the prices of all commodities. In ordinary intercourse between nations . . . there will always be a large number of commodities thus acting as outlets and inlets. And since the quantity of money itself affects prices for all sorts of commodities, the regulative effect of international trade applies, not simply to the commodities which enter into that trade, but to all others as well. (P. 93) Criticisms of the Gold Standard and Proposals for Reform

Like Marshall, Fisher criticized the basing of the monetary standard on a single commodity-gold or, for that matter, silver-because of instability in supply and demand conditions of the money metal. The commercial world has become more and more committed to the gold standard through a series of historical events having little if any connection with the fitness of that or any other metal to serve as a stable standard ... so far as the question of monetary stability is concerned, ... we have hit upon the gold standard by accident. (Fisher [1922] 1965, p. 323) Instead, in a comparison of the purchasing power of gold with that of a number of other commodities he concluded that "in terms of general purchasing power, gold is no more stable than eggs and considerably less stable than carpets" (Fisher 1920, p. 41).30 In addition, Fisher, a number of years later, criticized the gold standard because it allows price-level movements and business fluctuations to be transmitted from one country to another. 31 As a remedy for the inherent instability in the purchasing power of money under the gold standard, Fisher offered his scheme for a compensated dollar: Issue gold certificates backed by gold bullion, but vary the weight of the gold backing per dollar so as to maintain a constant purchasing power of money by tying the weight to an index number-"to mark up or down the weight of the dollar (that is, to mark down or up the price of gold bullion) in exact proportion to the deviations above or below par of the index number of prices" ([1922] 1965, p. 498). This would allow us to "keep the metal gold for the good attributes it has-portability, durability, divisibility, salability-but correct its instability, so that one dollar of it will always buy approximately [a] composite basketful of goods." It would "retain gold as a good medium [of exchange] and yet ... make it into a good standard" (1920, pp. 88-89).

54

Michael D. Bordo

Knut Wicksell Wicksell's views ([1898] 1965) on the gold standard will be discussed under three headings: the commodity theory of money, the international adjustment mechanism, and proposals for reform. Commodity Theory of Money

In contrast to Fisher, for Wicksell the stabilizing features of the commodity theory were too slow to be of consequence for price-level movements except in the very long run. Either the underlying mechanism was weak or, by the end of the nineteenth century, institutional developments had neutralized or obscured it. Thus the newly extracted gold-passes, for the most part, not into circulation, but into the stocks of cash of monetary institutions; and gold for industrial uses is mainly taken either out of these stocks or directly out of imported stocks of uncoined metal. In neither case can it be supposed that there is any direct effect on prices. (Wicksell [1898] 1965, p.31) The use of gold purely as a monetary base was utopian. In such a system the value of money would be directly exposed to the effects of every fortuitous incident on the side of the production of the precious metal and every caprice on the side of its consumption. It would undergo the same violent fluctuations as do the values of most other commodities. (P. 35) International Adjustment Mechanism

Wicksell was skeptical of the Hume price-specie-flow mechanismdisturbances affecting one country's price level relative to another's would affect the terms of trade, and the balance of payments would then be corrected by gold flows. Although the explanation was fundamentally correct, he expressed reservations: It is ... clear that international equilibrium of prices is usually reached far more rapidly and far more directly. The increase in the supply of foreign goods and the diminution in the demand for exports must themselves exert, directly and indirectly, a pressure on domestic prices which is quite independent of any simultaneous movement of precious metals. (Wicksell [1898] 1965, pp. 157-58) For Wicksell, changes in real income must be brought into the adjustment mechanism. Proposals for Reform

Like Fisher, Wicksell was concerned with price stabilization, both nationally and internationally. He favored the use of gold certificates

55

The Gold Standard: The Traditional Approach

backed by a reserve of gold bullion, with each central bank maintaining convertibility and agreeing to accept other central banks' notes and clearing them through an international clearing house (Wicksell [1898] 1965, pp. 186-87). In addition, the central bank of each nation would stabilize its internal price level by keeping the market rate of interest in line with the "natural rate of interest" following simple criteria: 32 So long as prices remain unaltered the bank's rate of interest is to remain unaltered. If prices rise, the rate of interest is to be raised; and if prices fall, the rate of interest is to be lowered; and the rate of interest is henceforth to be maintained at its new level until a further movement of prices calls for a further change in one direction or the other. To achieve international price stability, central banks would need to manipulate their gold stocks cooperatively to keep interest rates in line between nations (p. 192).

Appendix C

The Harvard Neoclassical School

F. W. Taussig of Harvard and his students-Jacob Viner, F. D. Graham, J. H. Williams, and H. D. White-and W. A. Beach (a student of Allyn Young and J. H. Williams) formulated and tested some of the main tenets of the classical Ricardo-Mill theory of the international adjustment mechanism. Taussig's reformulation of the theory and the evidence for Great Britain, the United States, France, Canada, and Argentina in the pre-World War I era are summarized here. In addition the writings of J. W. Angell, a contemporary critic of the Harvard School and a precursor of the modern monetary approach to the balance of payments are examined. F. W. Taussig In an article (1917) and a book ([1927] 1966), Taussig clearly reformulated the traditional approach in a manner suitable for empirical verification and then summarized some of the evidence. For Taussig, international borrowing was the most important disturbance to the pre-World War I international economy. He analyzed the effects of the transfer of capital on the balance of payments, money supplies, and price levels of both lender and borrower. Several of his studies of British and U. S. experience, beginning with the earliest period he covered, will be reviewed here. In an examination of the British balance of payments in the period 1853 to 1879, Taussig found that the adjustment of the merchandise balance of trade to changes in invisibles-both payments and shipping earnings-

56

Michael D. Bordo

and to capital exports was much more rapid and smooth than classical theory would lead one to expect: No signs of disturbance are to be observed such as the theoretic analysis previses; and some recurring phenomena are of a kind not contemplated by theory at all. Most noticeable ... is the circumstance that periods of active lending have been characterized by rising prices rather than by falling prices, and that the export of goods apparently has taken place, not in connection with a cheapening of goods in the lending country, but in spite of the fact that its goods have seemed to be dearer at times of great capital export. (Taussig [1927] 1966, p. 239) In addition, he found specie movements to be small relative to merchandise movements, a fact to be explained by the sensitivity of the British money supply. Moreover, he found it difficult to separate specie flows consistent with the theory from the more steady series of gold flows into Britain following the gold discoveries in the 1840s and 1850s. The general lack of conclusive evidence sympathetic to the classical mechanism for the period 1853-79 is reversed for the period 1880-1914. That period is conveniently divided into two parts: 1880-1900 and 19011914. The first period was characterized by a deficit on merchandise account financed by shipping earnings and income from abroad with no unusual capital exports. The second was dominated by massive capital exports, which were quickly reflected in a decline in the merchandise deficit. According to Taussig, the change in circumstances offered a good test of the theory. Had the trends before 1900 continued, Britain would have continued to enjoy an improvement in her terms of trade, but the terms of trade turned around after 1900. That phenomenon, he believed, was consistent with classical theory: That the gross barter terms of trade should vary as they do, becoming more favorable until 1900, thereafter less favorable, is indeed easily in accord with theory. They will naturally fluctuate in the same direction as the balance of payments. . . . More significant. . . is the fact that the net barter terms of trade move in the same direction, ... [w]e have argued that when a country has payments to receive for other items than merchandise, the direct and simple exchange of goods for goods is also affected, and is affected to the country's greater advantage. (P. 256)33 The improvement in the terms of trade raises domestic real income because money incomes rise, while the price of imported commodities falls. Taussig found the evidence that money wages rose until 1900, after which they turned down, consistent with his theory.34 However, a continuing puzzle for Taussig was how to separate the influence of equilibrating gold flows from the effects of the steady inflow

57

The Gold Standard: The Traditional Approach

of gold into Britain, the world's principal gold market. Moreover, the puzzle was complicated by the fact that commodity exports and imports [respond] with surprising promptness to the balance of international payments as a whole. The promptness is surprising because each constituent transaction . . . is purely in terms of money. . . . Yet the recorded transactions between countries show surprisingly little transfer of the only "money" that moves from one to the other-gold. It is the goods that move, and they seem to move at once; almost as if there were an automatic connection between these financial operations and the commodity exports or imports. That the flow of goods should ensue in time, perhaps even at an early date, is of course to be expected.... What is puzzling is the rapidity, almost simultaneity, of the commodity movements. The presumable intermediate stage of gold flow and price changes is hard to discern, and certainly is extremely short. (P. 261) He also examined a case under the gold standard of U.S. borrowing long-term capital from Great Britain. 35 To the extent that not all of the proceeds of the loan are spent on British goods, the increase in remittances from London to New York will cause a demand for New York exchange in London. New York exchange will rise in London, sterling exchange will fall in New York. . . . The fluctuations in foreign exchange will necessarily be confined within the gold points. (Taussig, 1917, p. 394) The next step is a gold flow from London to New York once the specie-export point is reached. Elsewhere, Taussig argued that in the case of temporary disturbances, gold rarely flows. Much of the adjustment is taken up by movements in the exchange rate within the gold points that speculators promoted. In addition, in the prewar era, sterling bills acted as a substitute for gold in important money-market centers and were transferred in lieu of gold flows. Finally short-term capital movements acted as a substitute for gold flows. 36 "In the last resort, when all expedients for adjusting and equalizing the payments between nations have been utilized and exhausted, specie will flow in payment of balances" ([1927] 1966, pp. 220-21). The gold inflow into the United States and the gold outflow from Great Britain then raises and lowers the money supplies of the two countries respectively (1917, p. 394). However, the effects of gold flows on domestic money supplies depend on institutional arrangements in each country.37 In the pre-World War I era of fractional reserve banking and convertible fiduciary money the inflow or outflow of specie. . . primarily affects the discount policy of the banks. Their discount policy in turn affects the volume of accommodation which they offer to the borrowing public, and this in turn affects the volume of notes and deposits. ([1927] 1966, p. 201)

58

Michael D. Bordo

The next step is a "train of consequences familiar to the reader of Ricardo and Mill. Prices will fall in Great Britain and will rise in the United States" (1917, p. 394). Within the United States, prices of domestic goods and exports rise relative to prices of imports, and opposite movements of sectional prices occur in Great Britain. 38 Finally, because the terms of trade have turned in favor of the United States, her citizens are better off: "Their money incomes have risen, the prices of imported commodities have fallen; as buyers of imported commodities they gain" (p. 395). The opposite takes place in Great Britain. In the case of the United States after 1879 when it was a net importer of British capital and from 1900 to 1914 when it financed a merchandise surplus with immigrant remittances and other invisibles, Taussig expected to observe gold movements and terms-of-trade effects opposite to those he had observed in the British case. As in the British case, Taussig ([1927] 1966, p. 299) found it difficult to separate the effects of equilibrating gold flows from primarily domestic gold production and consumption. However, unlike the British case, the terms of trade did not behave according to theory. It was difficult to discern a marked trend, though the net barter terms of trade were slightly less favorable before 1900 than afterwards. This, he stated is an unexpected result.... On general principles we should look for terms of trade more unfavorable in the second stage. The excess in the money volume of exports meant ... that the United States, in meeting the divers additional charges for immigrants' remittances ... sent out goods having a greater money value than the goods she bought.... The case shows an outcome different from that in Great Britain and Canada during the same period. For these countries, the actual course of events proves to be in accord with theoretical prevision. For the United States it does not. (P. 303) Taussig attempted to account for the poor results for the United States compared to those for Great Britain and Canada by reference to disturbing causes not present in the other countries. Among these were the tariff that made the terms of trade more favorable than otherwise and an exogenous increase in demand by foreigners for U.S. manufactured goods. Taussig then analyzed effects of a capital transfer under flexible exchange rates, the example relevant for the period 1862-78 when the United States had inconvertible paper money and Great Britain was on the gold standard (1917, p. 386). The initial effects of U. S. borrowing in London is to reduce the specie premium in New York (i.e., the U.S.-dollar price of gold falls). This reduction leads to a fall in the paper prices of U.S. exports (as well as domestic goods) and a rise in the paper prices of U.S. imports. As a

59

The Gold Standard: The Traditional Approach

consequence, export industries in the United States are discouraged, imports are encouraged. The opposite effects take place in Great Britain. Thus the total volume of commodities bought and sold in the United States increases, as exports are shifted to domestic consumption. Moreover, real income rises in the United States because holding the money supply constant, nominal income remains unchanged but prices have fallen. U.S. residents gain not only as purchasers of imports (they do under a gold standard as well), but also as purchasers of domestic commodities. In contrast to the gold standard case, the U.S. terms of trade improve not because U.S. residents have larger money incomes and lower prices, but because they have the same money incomes and lower prices. The opposite results occur in Great Britain. Taussig's analysis may be contrasted with that in Friedman and Schwartz (1963, pp. 84-85), of the effect on the gold premium of U.S. Treasury gold purchases abroad before resumption. The Evidence Taussig's students presented evidence for the classical adjustment mechanism based on detailed examination of the monetary history of a number of countries under the gold standard and under inconvertible paper money. Initially the case studies for Great Britain, the United States, Canada, and France under the gold standard and then for the United States and Argentina under inconvertible paper will be reported.

W. Beach In contrast to Taussig, Beach (1935) presented evidence for Great Britain for the period 1881-1913 unfavorable to the classical price-specieflow model. Beach argued that if differences in the levels of commodity prices were to dominate the adjustment mechanism, then gold would be expected to flow because the price levels of one country do not rise or fall in accordance with the levels in other countries. The movement of gold forces all countries to keep the same pace through the various phases of the business cycle. (Beach 1935, p. 170) Thus one would expect gold to flow out of a country in the upswing of the business cycle. Also, according to the classical model, one would expect capital exports to move procyclically so that other things equal, an upswing in a creditor country would induce a capital flow to a debtor country, accompanied by a gold outflow, unless offset by increased purchases of the creditor's exports. Yet Beach found that there was a tendency for gold imports to increase during the prosperity stages of business cycles and for exports to grow during depression, both in Great Britain and the United States.

60

Michael D. Bordo

Beach's explanation of this anomaly relied on several pieces of evidence. First, he observed a high correlation between business conditions and (long-term) capital exports from Great Britain. The volume of new loans, he suggested, was determined primarily by business conditions (pp. 171-73). Second, in the upswing of the business cycle in Great Britain (and the United States), there was an increased internal demand for gold currency. Third, short-term balances (loans) were sensitive to changes in discount rates, more so than domestic business and prices. Hence in the business-cycle upswing, the internal currency drain put pressure on the reserves of the banking system and the Bank of England, which led to a rise in the discount rate, a short-term capital inflow, and, ceteris paribus, a gold inflow. 39 Thus cyclical fluctuations in the movement of specie might easily be controlled by the movements of these balances. This explanation for the cyclical movements of gold found for England and the United States seems more adequate than the explanation based upon price level differences. (P. 180) Jacob Viner Viner (1924) tested the classical balance-of-payments adjustment mechanism for Canada during the period 1900-1913. 40 During that period, the great disturbing factor was the inflow of foreign capital into Canada. The adjustment to be explained was the "process whereby the Canadian borrowings, negotiated in terms of money, entered Canada in the form of goods and not in gold" (Viner 1924, p. 145). The mechanism to be tested was that of J. S. Mill. The capital inflow initially would raise the price of foreign exchange to the gold export point. This would then be followed by a gold flow from the lending to the borrowing country. Prices would rise in the borrowing country and fall in the lending country, leading to a change in imports and exports, with the borrowing country experiencing an unfavorable balance of trade and the lending country, a favorable balance. Once the unfavorable balance of the borrowing country equaled the rate of borrowing, the exchanges would return to parity, gold movements would cease, and relative prices in the two countries would stabilize at their new levels. First, examining the effects of changes in the exchange rate within the gold points, Viner argued that since transportation and insurance costs were very low between Montreal and New York, the gold points were so narrow that changes in the exchange rate were not likely to have much effect on the balance of trade (p. 155). Second, Viner found evidence that gold flows were highest in the years when Canadian borrowing was most in excess of Canadian loans to others

61

The Gold Standard: The Traditional Approach

(p. 160). At this point, Viner digressed on the role of gold in the Canadian financial system. Before World War I, gold did not circulate as hand-tohand currency but did act as a reserve asset for the chartered banks. (Canada did not have a central bank in this period.)41 The Canadian monetary base consisted of a fixed issue of government fiduciary notesDominion notes and gold reserves, largely maintained as "outside reserves" on call in New York or London, or as balances with commercial banks in those centers. According to Viner, changes in outside reserves in New York acted in a manner similar to gold flows in the classical balanceof-payments adjustment mechanism. Thus the transfer of foreign capital from London to Canada usually passed through the New York money market, raising outside reserves of Canadian banks. Then on the basis of the increased outside reserves, Canadian banks would increase their deposits (pp. 177-79). Third, the increase in the Canadian money supply would produce a rise in the Canadian price level (relative to the rest of the world) in accordance with the classical theory42 and a rise in sectional price levels. The initial effect would be on the prices of domestic goods and not on the prices of imports, which for a small open economy such as Canada are determined abroad. Some substitution away from domestic goods towards imports would follow as would also a rise in the price of exports (to the extent Canada had monopoly power in their production), leading to a reduction in exports.43 The evidence generally confirms these predictions: between 1900 and 1913 indexes of the prices of imports increased least, followed by the prices of exports, while domestic prices increased the most. Moreover, a beginning-of-period weighted index of the price of exports declined relative to an unweighted index, suggesting that commodities shifted from the export to either the domestic or import category. In addition, the ratio of domestic-goods prices to the wholesale price index increased more in Canada than in the United States. 44 Finally a decline in exports and a rise in imports completed the case for Viner in favor of the classical adjustment mechanism. 45 This theory has been verified inductively for Canada during the period 1900 to 1913. [Moreover,] a corollary of this reasoning [is] that during a period of international borrowings the terms of international exchange shift in favor of the borrowing country and against the lending country. . . . Adequate inductive verification of this proposition is supplied by the demonstration already made that export prices rose relative to import prices. (P. 295) Harry D. White White (1933) examined the French evidence for the classical pricespecie-flow explanation of the effects of capital exports. Like Britain,

62

Michael D. Bordo

France in the four decades before World War I lent large sums abroad, and as in the British case, for contemporary economists, the income from foreign investment permitted a persistent merchandise-trade deficit. In a careful reconstruction of the French balance-of-payments accounts, White demonstrated that the income from foreign investment over the 1880-1913 period was no greater than the total export of capital for the same period. Moreover he found, contrary to the official figures, that in twelve out of the thirty-four years surveyed, France had a surplus on merchandise account; and the years when the French accounts showed a deficit on merchandise trade it was paid for not by revenue from French foreign investment but by foreign-tourist expenditures in France (White 1933, p. 301). White then attempted to determine, as Taussig did for Great Britain, the adjustment mechanism by which the net capital export affected the balance of trade: Tho the totals of capital exports and net revenue from foreign investment over the period as a whole were not far apart, for anyone year during most of the period they differed considerably, thereby raising the question of the mechanism for adjustment of the disequilibrium caused by changes in the volume of capital exports. (P. 302) First, he found that in France's trade with gold standard countries, exchange-rate movements had a negligible effect on merchandise movements, but in the case of a number of countries not on the gold standard, from which France obtained one quarter of her imports, fluctuations in the exchange rate made a significant difference to French importers. Second, he concluded that in the French case, only a small portion of the sums loaned abroad were spent on French exports. Third, the movements of sectional price changes and of the volume of merchandise trade revealed a relationship in accordance with the classical sequence. Relative increases in import prices were accompanied by a decline in physical quantities and vice versa. The changes continued until there was a rough approximation between changes in the values of merchandise balances and capital exports. 46 Fourth, he could find no evidence of the linkage of gold flows to the reserves of the Bank of France and thence to the domestic money supply. According to White, the Bank of France would raise the discount rate only to offset a large gold outflow, but not to influence domestic economic activity. Moreover France's large gold reserves relative to other major gold standard countries enabled it to follow such a passive discount-rate policy. Fifth, White argued that it was possible that there existed a direct link between gold flows and price-level movements because such a large fraction of the French money supply consisted of gold currency, but

63

The Gold Standard: The Traditional Approach

no clear evidence on . . . the relationship of the quantity of money in circulation to prices is revealed by the comparison between the fluctuations in the quantity of specie and notes outside the Bank of France and the movement of prices. Moreover, even if such a relationship were revealed, the absence of correlation between the annual movements of capital and specie renders dubious the interpretation that changes in sectional price levels were induced by capital movements. (P. 304) Finally, the mixed evidence leads him to conclude that the specie-flow-specie mechanism is doubtless one of the forces, but there seems to be no justification for assuming that it is the sole or even the dominant means of adjustment. ... The neoclassical theory is not the complete explanation. The theory fails in that it explains what happens only under certain given conditions seldom found. It expounds a sequence of changes which undisturbed would in time bring about adjustment, but which seldom, if ever, operates unchecked by the frictions and rapid changes characteristic of modern economic conditions. By ignoring some of these forces and minimizing others, the neoclassical exposition exaggerates the effectiveness of gold flows and sectional price changes as a means for establishing equilibrium in international accounts. (P. 306)

Frank D. Graham Graham tested Taussig's (1917) theory of the adjustment mechanism under depreciated paper. In the period 1862 to 1878, when the United States was on an inconvertible paper standard-the greenback standard-while her principal trading partner, Great Britain, was on a gold standard, the premium on gold was a close proxy for the dollar-pound exchange rate. 47 Graham analyzed the effects of British capital flows-the major source of disturbance during this period to the balance of payments, and hence the exchange rate. The period can conveniently be divided into two episodes: 1863 to 1873, a period of heavy and continuous borrowing from London, and 1874 to 1878, a period when the borrowing dropped off. According to Taussig's theory, one would expect the period of heavy borrowing to be associated with a deficit in the balance of trade and the period of cessation with a reversal in the balance of trade. Graham's evidence showed a large annual excess of commodity imports over exports in the period of heavy borrowing, the opposite in the period of cessation (Graham 1922, pp. 231-34). Also, according to Taussig's theory, one would expect, ceteris paribus, the period of heavy borrowing to be associated with a decline in the exchange rate (the premium on gold), while the period of cessation to be associated with a rise in the exchange rate. Graham found the evidence corroborated this prediction. In comparing the price of gold with an index

64

Michael D. Bordo

number of general commodity prices, he found the quarterly average price of gold to be lower than the general price level from April 1865 through June 1876, while from July 1876 to the end of the period it was higher, with the price of gold rising relative to the overall price level after 1874. Next, according to Taussig, one would observe the following effects on sectional prices: in the first period, low paper prices of exports, gradually declining paper prices of imports, and relatively high paper prices of domestic commodities; in the second period, a gradual reversal towards higher paper prices of exports, gradually rising paper prices of imports, and relatively low prices of domestic goods. A comparison of the arithmetic means of the three different groups of commodities between the two periods provided evidence consistent with the theory. Finally, one would observe different effects on money wages (as a measure of relative prosperity) in the two periods: in the first period declining wages in the export industries relative to the domestic-goods industry, as resources are diverted from it in the face of falling prices; the opposite movement in the second period. Again, using Mitchell's wage data and classifying U.S. industries into domestic and export industries, Graham found the evidence consistent with the theory (pp. 267-70). At the same time as these effects occurred in the United States, opposite ones would occur in Great Britain, although the fluctuations in the U.S. dollar would have only limited effects on British prices since Great Britain was on a gold basis and movements of the dollar would only affect a portion of her trade. 48 Again Graham found the evidence consistent with the theory.

John H. Williams Williams (1920) tested for Argentina Taussig's theory of the adjustment mechanism under depreciated paper but unlike Graham found the evidence too inconclusive to be of more than limited support to the theory. First, according to Taussig's theory, a rising premium on gold would stimulate exports. "It does so by virtue of the fact that export prices rise more rapidly than costs, creating an extra profit or bounty for the producing and exporting classes" (Williams 1920, p. 233). Williams presented evidence indicating a rise in the price of exports concomitant with a rising premium on gold in the period 1885 to 1891. However, according to the theory, the value of exports ought to rise-but it did not. This result Williams attributed to the presence of other forces such as "the character of the Argentine exports . . . agricultural and grazing products . . . [which are] extremely susceptible to vagaries of climate" and the fact that "though [the] quantity of exports increased, the greater quantity was sold for a lower gold price per unit" (p. 234).

65

The Gold Standard: The Traditional Approach

Second, according to Taussig's theory, a rising gold premium should reduce the value of imports, which the evidence confirmed. We find first of all that the diminution of value of imports asserted by theory ... did occur: that, in fact, the diminution was very marked . . . . On comparing the course of imports with the gold premium, ... there was in every year an inverse relation between imports and the premium on gold. (P. 253) In sum, the evidence marshalled by the Harvard school in favor of the classical adjustment mechanism is mixed. Overall price levels adjust in accordance with the theory, but sectional price adjustment in accordance with the theory is limited. Little support was found for the role of gold flows, the money supply, and discount rates in the mechanism. A common finding was that the commodity trade balance adjusted rapidly to the external disturbance of capital flows, more rapidly than would be expected by a theory postulating links from price-level differences to gold flows, to changes in money supplies, to changes in sectional price levels, and then adjustment of the commodity trade balance. Explanations given by the school for the rapidity of adjustment-with no supporting evidence-included the growing integration of goods and securities markets, the role of income effects, and the sensitivity of the money-supply process. J. W. Angell

Angell ([1925] 1965), a student of Taussig's, had a distinctly different interpretation of the international adjustment mechanism under the preWorld War I gold standard. Angell focused on the relationship between individual commodity prices and national price levels. First, he argued that the classical division of the overall price level into export, import, and domestic-goods prices was "extremely misleading and may lead to erroneous conclusions." The distinction he preferred was between international or traded commodities and domestic or nontraded commodities, with the dividing line between the two types of commodities to be determined empirically: Any movable article whatsoever may enter international trade.... The first requisite for movement is that the money prices receivable, translated through a common measure, shall be higher in one country than in the other; the second, that the difference shall at least cover costs of transportation of all sorts, including tariff[s] (Angell [1925] 1965, pp. 375-76) He cited evidence for equality in the world prices of traded goods (staple commodities) at one extreme, and at the other extreme, no international competition and hence no reason for price equalization of nontraded

66

Michael D. Bordo

goods, with an in-between third category-partially traded goods. Arbitrage would ensure equality for traded-goods prices but arbitrage may not take place for many commodities because of: (a) "lack of accurate information, in each market," (b) "lack of sufficient initiative and enterprise, on the part of the manufacturers and dealers, to take advantage of the discrepancy," (c) "selling in a new market requires the prior erection of trade connections," (d) monopoly power (pp. 379-80). Second, Angell discussed evidence showing long-run similarity of national price-level movements between countries on the same metallic standard. He defined national price structures as a series of solar systems, which maintain a fairly constant relationship in their movements through space. But the component parts of the system-individual prices-are in a state of ceaseless change relative to the elements in both their own and other systems. (P. 390) However, he argued that the similarity of movement of national price levels coupled with the tendency to price equalization of internationally traded goods suggests that domestic (nontraded-goods) prices conform to the pattern set by traded-goods prices. Further, since the process of substitution between domestic and traded goods is a relatively weak one, the key mechanism that keeps prices levels in line is the classical price specie flow analysis.... No other type of explanation can adequately account for the known facts. Prices in different countries do not move together, over long periods of time, by sheer accident. There must evidently be some connecting link between them. But the influence exerted by international prices alone, on the various national price structures, is not great enough to provide this link. It is necessary to discover some condition or element that is capable of affecting the totality of prices indiscriminately, and fairly rapidly. This element is found in the mechanism by which the balance of international payments is kept in equilibrium. (P. 393) Angell offered his own version of the price-specie-flow mechanism. He downplayed the role of actual gold flows in the mechanism, arguing that neither the magnitudes nor the directions of the international flow of gold are adequate to explain those close and comparatively rapid adjustments of payment disequilibria, and of price relationships, which were witnessed before the war. (P. 400) Moreover, the character of modern banking [will not] permit the assumption of any very high degree of intimacy to the connection between a country's metallic stock and its price level. Finally, gold is among the least sensitive of the media of international payments, one of the slowest to move. (Pp. 400-401)

67

The Gold Standard: The Traditional Approach

His own version of the mechanism relies on the importance of foreign bills of exchange which act as a substitute for gold. Temporary disequilibria will be offset by movements in the exchange rate. If that proves inadequate, gold will flow, prompting a change in discount rates. Finally, if the disturbance is more than temporary, then an alteration will take place in the underlying conditions that govern the general course of international exchange itself. The volume of purchasing power in circulation in the creditor country will be built up, in consequence of the increase in the banks' holdings of bills. In the other country it will be reduced, through the decline there in bill holdings. These changes, in turn, will operate upon the corresponding general price levels. The latter effect is often, though not always, strengthened by alterations in the discount rates. The movements in prices will then influence the commodity balance of trade. . . and will continue to do so until the change in the commodity trade has become great enough to offset and correct the original disturbance in the balance of international payments. (P. 413)

Appendix D

Interwar Critics

With the outbreak of war in 1914 and Britain's suspension of convertibility, the classical gold standard expired. The gold-exchange standard existed briefly from 1925 to 1931, after which the problem of international monetary reform and the creation of a more viable international monetary system took center stage. The traditional approach to the gold standard was subjected to extensive reinterpretation and criticism, much of it derived from concern over the monetary instability of the interwar period. The reinterpretation of the gold standard began with the Cunliffe report ([1918] 1979) which succinctly restated the stylized facts of the operation of the pre-1914 gold standard and appealed for a return to the old order. Another view (Brown, Smit) of the prewar gold standard stressed that it was successful because it was a managed standard-managed by London. Followers of th~s institutional approach then documented all the many respects in which the structure of the British money, gold and commodity markets"pax Britannica"-and the astute management of the Bank of England made the system work. The key implication of the approach was that a successful gold standard could be restored if a similar institutional milieu could be re-created. A related approach (Cassel) argued that the gold standard worked well for England, and possibly for several other major countries, but not for the rest of the world. Moreover, the fact that it worked so well for England was largely an accident of history (Viner).

68

Michael D. Bordo

A recurrent theme in the interwar literature was the inherent policy conflict between internal and external stability under the gold standard fixed-exchange-rate system. According to this view (Keynes), the prewar gold standard worked because nations were willing to subsume domestic economic objectives to the maintenance of gold convertibility; but in the postwar period, a return to the harsh discipline of the prewar gold standard would be disastrous. This approach proposed the creation of a supernational monetary agency or similar means to ensure international harmonization of economic policy. The final theme in the literature was the expression of doubt about the stylized facts of how the gold standard worked. Returning to the anomalies between fact and theory discussed by the Harvard school, the critics (Whale) argued that perhaps the traditional approach itself was incorrect. The views of writers identified with each of the foregoing themes are summarized below. The Stylized Facts In the interwar period, a persistent thread in the literature was a view of the prewar gold standard as the ideal monetary standard. A classic statement of this view appeared in the Cunliffe report, but it was repeated by others including R. G. Hawtrey, T. E. Gregory, and the Interim Report of the Gold Delegation of the Financial Committee of the League of Nations ([1931] 1979). The Cunliffe report ([1918] 1979) to Parliament succinctly presented what seemed to be the salient features of the operation of the prewar gold standard in Great Britain and a series of proposals for a quick return to gold. 49 As the report documented, the money-supply process before the war was based on the Bank Charter Act of 1844. Apart from a fixed fiduciary issue, hand-to-hand currency consisted entirely of gold and subsidiary coins or of gold certificates. Gold was freely coined and there were no restrictions on the import or export of gold, so changes in the monetary base, aside from movements of gold to and from the arts, were determined by inflows from abroad and outflows. In addition, upon this base of a fixed fiduciary issue and gold and gold-backed currency rested an extensive system of checkable bank deposits, so that the pre-World War I British money supply consisted mainly of deposits. Second, the report described the operation of the balance-of-payments adjustment mechanism. A disturbance to the balance of payments led to a gold flow and a corresponding change in the money supply. Thus, e.g., when the balance of trade was unfavorable and the exchanges adverse, it became profitable to export gold, and the would-be exporter bought the gold from the Bank of England. The Banking Department in turn

69

The Gold Standard: The Traditional Approach

obtained gold from the Issue Department in exchange for notes from its reserve, with the result that liabilities to depositors and the reserve were reduced by an equal amount, and the ratio of reserves to liabilities declined. 50 The next step was a rise in the discount rate: If the process was repeated sufficiently often to reduce the ratio in a degree considered dangerous, the Bank raised its rate of discount. The raising of the discount rate had the immediate effect of retaining money here which would otherwise have been remitted abroad and of attracting remittances from abroad to take advantage of the higher rate, thus checking the outflow of gold and even reversing the stream. (Cunliffe report [1918] 1979, par. 4) Thus raising Bank rate by inducing a short-term capital inflow would be sufficient to stem a temporary balance-of-payments deficit. However, in the case of a permanent disturbance, the discount-rate rise would additionally reduce domestic credit. This description of the operation of Bank rate to facilitate the adjustment mechanism has often been referred to as the rules of the game. 51 According to Nurkse: Whenever gold flowed in, the central bank was expected to increase the national currency supply not only through the purchase of that gold but also through the acquisition of additional domestic assets; and, similarly, when gold flowed out, the central bank was supposed to contract its domestic assets also .... The chief methods to be used for changing the volume of domestic central bank assets in accordance with this principle were changes in the discount rate, designed to make borrowing from the central bank either more or less attractive, and purchases or sales of securities in the open market on the central bank's own initiative. (Nurkse [1944] 1978, pp. 66-67) Moreover, the gold standard also provided an automatic mechanism to offset an internal disturbance: When. . . credit at home threatened to become duly expanded, the old currency system tended to restrain the expansion and to prevent the consequent rise in domestic prices which ultimately causes such a drain. The expansion of credit, by forcing up prices, involves an increased demand for legal tender currency both from the banks in order to maintain their normal proportion of cash to liabilities and from the general public for the payment of wages and for retail transactions. In this case also the demand for such currency fell upon the reserve of the Bank of England, and the Bank was thereupon obliged to raise its rate of discount in order to prevent the fall in the proportion of that reserve to its liabilities. The same chain of consequences ... described [above] followed and speculative trade activity was similarly restrained. (Cunliffe report [1918] 1979, par. 6)

70

Michael D. Bordo

Thus, there was therefore an automatic machinery by which the volume of purchasing power in this country was continuously adjusted to world prices of commodities in general. ... Under these arrangements ... [the] country was provided with a complete and effective gold standard. The essence of such a standard is that notes must always stand at absolute parity with gold coins of equivalent face value, and that both notes and gold coins stand at absolute parity with gold bullion. When these conditions are fulfilled, the foreign exchange rates with all countries possessing an effective gold standard are maintained at or within the gold specie points. (Pars. 6-7) The committee recommended a restoration of the gold standard "without delay" (par. 15), to be achieved by the cessation of government borrowings and the reduction of Bank of England note issue. In addition, the committee recommended allowing the free external movement of gold and the use of Bank rate to check outflows and inflows. However, following Ricardo's "Proposal for a Secure Currency," it recommended against the use of gold coins for domestic circulation and in favor of use of all the nation's gold to be held by the Bank of England as backing for the nation's monetary base-a gold-bullion standard. Most of the proposals were adopted when Britain returned to gold in 1925. Hawtrey (1935) summarized and expanded upon the stylized facts of the Cunliffe report. He documented the domestic aspects of a gold standard. A gold-coin standard with free coinage and free export and import serves as a device to provide a limit on the supply of money, but the authorities must maintain the quality of the coin. Thus, the essence of the gold standard is that the price of gold, the value of gold in monetary units, is fixed by law and this determines the wealth value of the monetary unit itself. The use of gold coin ... provides a fairly close approximation to this ideal. (Hawtrey 1935, p. 20) A central bank acting as a bankers' bank by holding a pool of gold reserves can reduce the foregone interest cost on commercial banks' holdings of gold coins to maintain convertibility of their liabilities. However, since commercial banks keep their reserves with a central bank, it must accept the responsibility of acting as a lender-of-Iast-resort. For the banks and the public do not trouble themselves about the interchangeability of gold and credit. That is the affair of the Central Bank alone. Anyone can sell the Central Bank as much gold as he likes and can procure from it as much gold as he chooses to pay for. The Central Bank is in the gold market as both buyer and seller in unlimited quantities at a fixed price. (P. 24) The central bank has the power to control the domestic money supply by using its principal tool-its discount rate. By raising the rate, the

71

The Gold Standard: The Traditional Approach

central bank can check lending by "improvident banks" and by lowering the rate it can stimulate lending. Hawtrey then described the mechanism by which a change in Bank rate operates. The power of the central bank over the wealth value of the monetary unit ultimately depends on the deterrent effect of a high Bank rate upon the borrowing operations of the customers of the banks. Bank rate is essentially a short term rate of interest.... It is the borrowing of money for the purchase of goods that is likely to respond most promptly to a restriction or relaxation of credit, because a trader who wishes to reduce his indebtedness in respect of goods held in stock can readily do so by postponing or reducing his purchases. When traders are tending generally to do this, the effect is immediately felt by the producers of the goods in decreased orders.... . . . The installation of capital is usually financed by the raising of funds from the long-term investment market, but short-term borrowing is often resorted to in anticipation of the raising of funds from that source or for the purchase and holding of securities. If Bank rate is raised, the holding of capital assets with money temporarily borrowed is discouraged. But the effect on productive activity will be relatively slow, for the installation of capital is a prolonged process, and any such project is likely to be preceded by a long preliminary period of preparation. (Pp. 25-26) Thus, changes in Bank rate have their primary impact on the holding of inventories. Hawtrey also discussed international aspects of a gold standard. The commitment by a number of countries to fix the prices of their currencies in terms of gold establishes a fixed-exchange-rate system. The prices in anyone currency of gold in different places cannot differ by more than the cost of transporting gold between the different places-arbitrage in the gold market will ensure that outcome. To maintain convertibility of the currency-the primary responsibility of a central bank-an adequate gold reserve is essential. The threat of a loss of gold is more serious than a possible gain, since in the former case the country may be forced to leave the gold standard before the necessary adjustment can take place; in the latter case, although the central bank may temporarily lose control of the market, the stimulus to lending will eventually lead to a reversal of a gold inflow. Finally, Hawtrey, like Keynes (see below, p. 79) saw an analogy between the operation of a clearing system between banks within one country and a clearing system of central banks under the gold standard. In his analysis of the balance-of-payments adjustment mechanism following a monetary disturbance, Hawtrey incorporated elements of both the classical adjustment mechanism and the role of income changes. An increase in bank lending in an open economy by stimulating demand will

72

Michael D. Bordo

ultimately lead to an increase in production and real income, assuming less than full employment, but will initially lead to an increase in sales and reduction in inventories. Merchants and dealers will therefore seek to replenish their stocks. Those dealing in home-produced goods will do so partly by ordering fresh supplies from producers and partly by diverting to the home market goods that might have been sold abroad, while those dealing in foreign-produced goods will order fresh supplies from abroad. The diversion of production from exports and the increase in imports will create an adverse balance of payments and a gold outflow. The process continues until a new higher equilibrium level of income is reached with a lower than initial balance-of-payments deficit. Once full employment is reached, the increased spending will affect prices. However, the prices of traded goods cannot be fully raised. These, which may conveniently be called "foreign trade products" comprise not only actual imports and exports but all importable and exportable goods. The prices of foreign trade products are governed by prices in world markets and are fixed in gold. The demand for them will expand as the consumers' income expands, and as the demand expands the loss of gold grows greater and greater. (Hawtrey 1935, p. 41) In addition, some of the increased income will be diverted to the purchase of foreign securities which will worsen the current-account balance. The process can be arrested and the deficit reduced by an increase in the discount rate-quickly offsetting the capital inflow and ultimately reducing the rise in income. However, "the contraction of the consumers' income is the only substantial corrective" to the balance-of-payments deficit (p. 43). For Gregory ([1932] 1979), like Hawtrey, the essence of the gold standard is convertibility of national currency into a fixed weight of gold. One of the great advantages of the gold standard therefore is that . . . it eliminates fluctuating rates of exchange . . . [that] international trade and investment can be conducted without any fear that the sums risked in a particular trade or investment transaction will not be recovered . . . owing to changes in the relative exchange values of different moneys at the date of payment. (Gregory [1932] 1979, p. 9)52 In addition, he related the development of the international gold standard in the second half of the nineteenth century to the growth of international trade and investment in that period, and stressed the role played by gold movements in the establishment of the conditions necessary to secure equilibrium in the international balance of payments of the various countries upon the gold standard.... What the international gold standard does. . . is to force prices and incomes in different trading areas into such a relationship that the

73

The Gold Standard: The Traditional Approach

balance of payments can be adjusted without gold flows in either direction.... The international gold standard creates, not a common price level but an integrated price-and-income structure. (Pp. 11, 14-15) The gold standard can operate successfully in the context of modern banking systems and central banks (provided sterilization activity is not undertaken) and of tariffs, capital flows, and transfers. The Role of London An important theme in the literature of the gold standard, developed in the interwar period, was that the gold standard was successful primarily because it was a sterling standard. 53 Perhaps the most succinct statement of the position is in Smit (1934), although a similar viewpoint is expressed in Brown (1940). According to Smit, by 1914, all the leading money and trade centers in the world were interconnected in a triangular fashion through London, although smaller patterns, directly centered around Paris, Berlin and New York, were woven into the main pattern of the picture. (Smit 1934, p. 53) In the prewar world, sterling balances instead of gold were increasingly used by foreign financial institutions to settle international payments: The most important key to the world's foreign exchange markets lay in the sterling balances of foreign bankers kept in London.... London acted as one bank for a customer-neighborhood of bankers that comprised not only the small British island but the whole world. . . . The pound, which was internationally wanted for settling commercial and capital indebtedness with the British Isles, became more and more, as a consequence of the world-wide demand that it commanded, a conventional credit counter for settling indebtedness among all countries. (P. 54) The primary financial interconnection between countries centered on the London discount market because much of the world's foreign trade was financed by sterling bills. 54 The Bank of England by its discount rate and open-market operations was able to exercise considerable influence over this market. Thus the extraordinary effectiveness of the English official bank rate in influencing foreign exchange rates and international gold movements before the war cannot be explained unless one sees the integration that had taken place in the world credit structure (P. 55) In addition, the spread of the gold standard and integration of the international credit system were closely intertwined as "the legal guarantees of the gold standard limited the risk factor of foreign exchange

74

Michael D. Bordo

fluctuations and inspired confidence" and "London possessed the world's central gold market.... The possession of sterling balances was the surest means of getting gold when wanted" (p. 56). Finally, Smit argued that at the same time as the world moved towards a sterling standard, there was growing internationalization of commodity prices centered in British commodity exchanges. A key consequence of international integration of both commodity and money markets was that the traditional explanation of how the prewar gold standard operated placed too much emphasis on "the existence of different national monetary systems, and the quantitative relations between the separate national money and credit systems and domestic price levels" (p. 55). Even before World War I, Keynes ([1913] 1971) had noted the unique role of London in the operation of the gold standard. By World War I, England had developed a sound currency and, aided by the effective use of Bank rate, required only a small gold reserve to maintain convertibility in the face of external shocks. However, according to Keynes, most other countries were not as successful in staying on the gold standard. One key difference between Britain and other countries was that she was a net creditor in the international short-loan market, whereas most other countries were debtors. In the former case, which is that of Great Britain, it is a question of reducing the amount lent; in the latter case, it is a question of increasing the amount borrowed. A machinery which is adapted for action of the first kind may be ill-suited for action of the second. Partly as a consequence of this, partly as a consequence of the peculiar organization of the London money market, the "bank rate" policy for regulating the outflow of gold has been admirably successful in this country, and yet cannot stand elsewhere unaided by other devices. (Keynes [1913] 1971, p. 13) Most other countries in adopting the gold standard used gold as a medium of exchange but were unable to use the discount rate as an effective method to preserve the standard because, in addition to not being net international lenders, they had not established the elaborate financial network of the London money market. Consequently other countries used other mechanisms to supplement their inadequacies: they held large gold reserves "so that a substantial drain ... may be faced with equanimity"; they partially suspended payment in gold; and they kept "foreign credits and bills ... which can be drawn upon when necessary"(p. 14). Most countries (especially less-developed ones) tended to rely on the last method because it economized on the foregone interest cost of holding gold reserves. 55 Thus the gold-exchange standard evolved out of the discovery that, so long as gold is available for payments of international indebtedness at an approximately constant rate in terms

75

The Gold Standard: The Traditional Approach

of the national currency, it is a matter of comparative indifference whether it actually forms the national currency. . . . The gold exchange standard may be said to exist when gold does not circulate in a country to an appreciable extent, when the local currency is not necessarily redeemable in gold, but when the government or central bank makes arrangements for the provision of foreign remittances in gold at a fixed maximum rate in terms of the local currency, the reserves necessary to provide those remittances being kept to a considerable extent abroad. (P. 21) Cassel (1935) went even further than Keynes, arguing that the gold standard was an international standard neither in the pre- nor the postWorld War I periods. It was only a British standard. He argued that the automatic balance-of-payments adjustment mechanism of price levels adjusting to gold flows never in fact worked that way-that central banks for the sake of security maintained larger reserves than legally required, and therefore exports and imports of gold did not necessarily influence the domestic money supply or the price level. The gold supply of a country exercised such an influence only via the policy of the central bank and its regulation of the market by means of its rate of discount and its open market operations. Thus the currency necessarily became a "managed currency" whose value depended entirely on the policy of the central bank. (Cassel 1935, p. 3) In addition, capital flows hindered the automatic functioning of the international gold standard: A country normally exporting capital could compensate for a loss of gold simply by a reduction of its lending; and a country normally importing capital could compensate for a loss of gold by borrowing more. Thus it was possible to prevent gold imports or exports from having any influence on the price level of the country. (Pp. 3-4) Finally, the international gold standard did not guarantee a natural distribution of gold: Creditor countries were in a position to accumulate, if they chose to do so, disproportionate gold stocks without using them for any other purpose than for exercising political influence or merely for satisfying a national pride in the possession of gold. Debtor countries could provide gold reserves by increasing their foreign indebtedness.... The size of these reserves had very little to do with the balance of trade of the country. Nor did gold imports and exports have any distinct relation to changes in the balance of trade. (P. 4) That the international gold standard functioned so well before World War I can only be explained by the basic position that the pound sterling held in this system. Indeed, the pre-war gold standard system may not

76

Michael D. Bordo

inadequately be described as a sterling bloc held together by London's position as the world's financial clearing center and by the service of the pound sterling as a generally recognized means for international payments. (Pp. 4-5). Next, in Cassell's view, the international gold standard was fundamentally defective because it was based on the tacit assumption that the purchasing power of gold would be stable and hence that maintenance of a fixed gold parity guaranteed stability in the purchasing power of a country's currency. According to Cassel, the gold standard . . . [suffered] from an inherent and irreparable instability. This instability results partly from the instability of the value of gold itself, and partly from the insecurity of the redemption in gold of gold-standard currencies. (P. 6) On the first score, he presented evidence of considerable variability in the value of gold for the period 1850 to 1910, based on the Sauerbeck wholesale price index. This evidence he explained by "deviations of the actual gold supply from the normal gold supply"56 and "variations in the monetary demand for gold. "57 On the second score, he alleged that only Britain had a completely convertible currency in the prewar period; other countries usually put barriers in the way of the large gold exports and "eagerly watched their gold reserves." Indeed, he argued that by the end of the period most countries kept large gold reserves as a matter of national pride and consequently the key aim of policy was to protect the gold reserve rather than use the gold reserve to protect convertibility. Thus when World War I broke out, the redeemability of currency was immediately suspended in order to safeguard the gold reserve. Like Cassel, Viner (1932) argued that the prewar gold standard would. . . have been found impracticable and would have been generally abandoned ... [if not for] the development of a deliberate and centralized mechanism of control of gold movements, using central bank discount policy and credit control as its chief instruments. (Viner 1932, p. 9) Moreover, the Bank of England pioneered in the development of the technique of central-bank control. England became the manager of the international gold standard. However, the evolution of the Bank of England's effective management of the gold standard emerged as a by-product of the Bank's learning by a process of trial and error to protect its slim gold reserves. The Bank of England in the nineteenth century was primarily a profit-seeking institution and hence tried to minimize its non-interest-bearing gold holdings. However, by the close of the century, the Bank gradually began to accept responsibility for maintenance of an English gold standard: 58

77

The Gold Standard: The Traditional Approach

The Bank of England, at first as the sole issuer of paper money and the most important deposit bank, later under pressure of public opinion and in self-defense against the irresponsibility of the other English banks, partially accepted the role of a central bank with some degree of special responsibility for the mode of operation of the English gold standard and especially for the protection of the convertibility of the English paper currency. (Pp. 12-13) As a consequence, the Bank learned to makes its discount rate effective to protect its gold reserve in the face of an external drain and to hold adequate reserves to meet the exigencies of both external and internal drains. 59 Moreover, according to Viner, in the nineteenth century fluctuations in the exchange rates within the gold points and short-term capital flows aided the adjustment mechanism and reduced the size of gold flows necessary to offset a disturbance to the balance of payments. Finally, Viner made a case for the continuation of the gold standard, despite the fact that it did not produce a stable price level, because a system of inconvertible paper currencies linked by flexible exchange rates would be far worse. 60 We know too little . . . of the possibilities of stabilization to take immediately any major steps in that direction. The hostility of central bankers and the menace of political control are genuine and important factors in the situation. The gold standard is a wretched standard, but it may conceivably be the best available to us. Its past record, bad as it is, is not necessarily conclusive in this respect, as the only alternatives which have actually been tried have, on the whole, had an incomparably worse record. (P. 37) The Conflict between Internal and External Goals A major theme of the interwar period was the potential conflict between internal price stability and a fixed exchange rate under the gold standard. By fixing the price of gold in terms of domestic currency, movements in internal price levels (and real income) would be determined by external-price-Ievel (and real-income) movements. The prewar gold standard period, characterized by both price stability (in a long-run sense) and fixed exchange rates, was an accident of history, never to be repeated. At the same time, flexible exchange rates and abandonment of the gold standard rule were not embraced because of the risk of unstable exchange rates and the fear of the consequences of discretionary policy. That theme is echoed in the works of Keynes, Cassel, Viner, Nurkse, and others. To remedy the conflict between external and internal goals, various schemes were proposed to promote international harmonization of price-level movements under a managed gold standard. The views of major writers of the interwar period are surveyed briefly below.

78

Michael D. Bordo

For Keynes ([1923] 1971) the policy options facing Great Britain in the immediate postwar period were to go back to the gold standard at the prewar parity, which would involve deflation, or else to fix parity after devaluing the pound. The choice between devaluation and deflation was part of a more general dilemma-the choice between price stability and exchange-rate stability. Keynes then asked, "In the light of our answers to the first two questions, is a gold standard, however imperfect in theory, the best available method for attaining our ends in practice?" ([1923] 1971, p. 117). Because of its adverse affects on income distribution, Keynes rejected deflation. When internal and external price stability were incompatible, he chose internal price stability. Under the prewar gold standard, the choice was made in favor of fixed exchange rates and the subservience of the internal price level to external considerations. "We submitted, partly because we did not dare trust ourselves to a less automatic . . . policy, and partly because the price fluctuations experienced were in fact moderate" (p. 126). But the circumstances of the pre-1914 era were partly accidental, and it should not be presumed they would ever be repeated again. The special conditions Keynes cited for the past good performance of the gold standard were first, that progress in the discovery of gold mines roughly kept pace with progress in other directions-a correspondence which was not altogether a matter of chance, because the progress of that period, since it was characterized by the gradual opening up and exploitation of the world's surface, not unnaturally brought to light pari passu the remoter deposits of gold. But this stage of history is now almost at an end. A quarter of a century has passed by since the discovery of an important deposit. Material progress is more dependent now on the growth of scientific and technical knowledge, of which the application to gold mining may be intermittent. (P. 133) Second, the independent influences coming from the demand for gold in the arts and for hoarding purposes in Asia had a steadying influence. Third, central banks allowed their gold reserves to vary slightly, absorbing much of the additional gold produced after major discoveries and reducing some of their accumulated gold when it was relatively scarce. They thus minimized the effects on price levels. Given the special circumstances that made the gold standard successful before World War I, Keynes argued that the standard would be unlikely to work as well in the postwar period. Even if all countries adopted the gold standard-an important condition for it to be successful-the prewar system of balance-of-payments adjustment was "too slow and insensitive in its mode of operation" to handle the "large [and] sudden

79

The Gold Standard: The Traditional Approach

divergences between the price levels of different countries as have occurred lately" (pp. 128-29). Moreover, though short-term capital flows in response to interest-rate differentials helped speed up the adjustment mechanism in the prewar period, especially when the' disturbance was temporary, in the case of permanent disturbances, the adjustment "might obscure the real seriousness of the situation, and enable a country to live beyond its resources for a considerable time at the risk of ultimate default" (p. 130). This problem would be more serious in the postwar period. The case for flexible exchange rates and managed fiduciary money was that balance-of-payments adjustment to external shocks would be much more rapid under flexible than under fixed rates despite the risk of instability.61 Thus Keynes came out strongly against restoration of the classical gold standard by the United Kingdom. In truth, the gold standard is already a barbarous relic. All of us, from the Governor of the Bank of England downwards, are now primarily interested in preserving the stability of business, prices, and employment, and are not likely, when the choice is forced on us, deliberately to sacrifice these to outworn dogma, which had its value once, of £3 17s. 10V2d. per ounce. Advocates of the ancient standard do not observe how remote it now is from the spirit and the requirements of the age. A regulated nonmetallic standard has slipped in unnoticed. It exists. (P. 138) In contrast to his earlier focus on the policy dilemma facing one country alone, Keynes ([1930] 1971) concentrated on the international monetary system as a whole. The interrelationship between central banks in an international fixed-exchange-rate system such as the gold standard was ':lnalogous, he noted, to the relationship between commercial banks and the central bank within a national economy.62 Under the pre-World War I gold standard, commercial banks operated in step within one country and central banks operated in step internationally except that reserve ratios did adapt somewhat to relative scarcity or abundance of gold. Behavior of the long-run price level depended on whether new gold available for reserves was increasing faster or slower than trade of the gold standard countries, which in turn depended on the rate of discoveries and technological improvements in gold mines, the use of gold as currency, the number of countries joining the gold standard, and the growth of real per capita income. Some important differences in the relationship between central banks under the gold standard and between member banks and the central bank in a national economy, however, Keynes observed, were that central banks tended to have more variable reserve ratios;63 that more of its own money returns to the central bank than is the case for a commercial bank;

80

Michael D. Bordo

that there is a higher degree of competition for short-term capital between central banks through varying discount rates than is the case between commercial banks. 64 Thus in an international system, a central bank can pursue an independent policy-oriented primarily to domestic considerations-only within narrow limits and for short periods, with the degree of independence determined by its relative size. Thus, under a fully operative gold standard, "credit cycles have an international character" because member central banks must follow the average behavior. This implies a "real divergence of interest; and we must not expect of central banks a degree of international disinterestedness far in advance of national sentiment and of the behaviour of other organs of national government" ([1930], 1971, p. 257). The dilemma between internal-balance and external-balance considerations for one country on the gold standard is thus apparent. 65 Keynes argued that in a world of perfect capital mobility, the domestic interest rate must correspond to world interest rates: If any country tried to maintain a higher rate than its neighbours, gold would flow towards it until either it gave way or the international system broke down by its having absorbed all the gold in the world. And if it tried to maintain a lower rate, gold would flow out until either it gave way or had to leave the international system through having lost all its gold. Thus the degree of its power of independent action would have no relation to its local needs. (P. 271) The problem arises for a country if its foreign balance is inelastic, and if, at the same time, it is unable to absorb the whole of its savings in new investment at the world rate of interest. It will also tend to happen even where the foreign balance is elastic, if its money costs of production are sticky.... This, then, is the dilemma of an international monetary system-to preserve the advantages of the stability of the local currencies of the various members of the system in terms of the international standard, and to preserve at the same time an adequate local autonomy for each member over its domestic rate of interest and its volume of foreign lending. (Pp. 27172)66 As a solution to the problem, Keynes advocated a number of policies to increase the discretion of national monetary authorities while still remaining on a gold standard. One set of policies to protect a country's domestic stability in the face of "inconvenient fluctuations in the rate of foreign lending" is to maintain a large enough level of reserves: either in gold reserves at home or by holding "liquid balances in foreign centers," by arranging overdraft facilities with other central banks or by "borrowing and lending arrangements between central banks and a supernational

81

The Gold Standard: The Traditional Approach

bank" (p. 278). A second policy is to manipulate the gold points-to create an artificial spread between the official buying and selling price of gold. This can be done by direct authority or by the central bank manipulating the forward rate of exchange. 67 Finally, he suggested direct controls over capital movements. Keynes's ideal solution to the problem of combining an ideal international standard and internal equilibrium was a gold standard managed by a supernational bank. 68 The objectives of a supernational monetary agency would be to ensure long-run stability and to smooth short-run cycles around the long-run trend. One way to achieve long-run price stability would be to adopt a commodity standard based on an international aggregate of commodities. "The long-period trend in the value of gold should be so managed as to conform to a somewhat crude international tabular standard" (p. 351). To solve the problem of short-run disturbances within individual countries, Keynes advocated giving individual central banks more discretion within the fixed-exchange-rate system. The supernational bank would be established by all the world's central banks and would act as a lender-of-Iast-resort to them alone. It would hold as assets gold, securities, and advances to central banks, and its liabilities would be deposits by the central banks. These deposits, called supernational bank money (S.B.M.), would be fully convertible into gold and would serve along with gold as reserves for the member banks. The supernational central bank would then use the normal tools of monetary policy-bank rate and open-market operations-to "maintain . . . the stability of the value of gold (or S.B.M.) in terms of a tabular standard based on the principal articles of international commerce" and to avoid "general profit inflations and deflations of an international character" (p. 360). The interwar criticism of the traditional approach to the gold standard culminated in a provocative and path-breaking article by Whale (1937). He challenged both the price-specie-flow adjustment mechanism and the operation of the rules of the game. Whale referred to four pieces of puzzling evidence: (1) Taussig's finding that the adjustment of national price levels to disturbances occurred much more rapidly than the theory postulated; (2) Beach's finding that gold flows to Great Britain moved procyclically contrary to the classical prediction, and that they were more closely related to interestrate differentials than to price-level differences; (3) the finding that many prewar central banks did not follow the rules of the game, e.g., the central banks of France and Belgium rarely changed their discount rates yet remained on the gold standard; (4) the finding that price levels between regions with varying levels of economic activity moved synchronously, suggesting a linkage through arbitrage rather than adjustment with a lag to specie flows, as in the traditional theory. 69 "Might not the

82

Michael D. Bordo

national price level be similarly determined by the world system of prices?" (Whale 1937, p. 22). On the basis of this evidence, Whale suggested an alternative hypothesis to the classical mechanism: Rather than the demand for money in an open economy with a fixed exchange rate adjusting to the supply of money (and specie flows) as the classical theory predicts, specie flows and the money supply are determined by the demand for money, which in turn is determined by real income and the price level. Thus, according to Whale, an increase in real economic activity, for a given price level (determined by the world price level), would increase the demand for money, causing a balance-of-payments surplus and a gold inflow. Similarly an increase in the domestic money supply would lead to a balance-of-payments deficit, a gold outflow, and a decline in the reserves of the banking system. If the markets of the country are closely linked with foreign markets, the decline of bank reserves should lead to "an almost immediate correction" of the money supply (p. 27). Interest rates also playa different yet still important role in the alternative mechanism: What is contended is that. . . the raising of interest rates did not have the effect of producing a relative reduction of prices in certain countries. High rates in London led rather to a world fall in prices, partly because of the sympathetic movement of rates elsewhere, partly because of the effect on British entrepot trade and British long-term foreign investment. (P. 27) Two important implications follow. First, the classical transfer mechanism of Mill and Taussig must be reinterpreted. According to Whale, a transfer of capital involves a redistribution of spending power. However, rather than this process involving a change in the direction of demand and in the terms of trade, "the redistribution of spending power itself, ... apart from any change in the direction of demand and the terms of trade, may require a redistribution of money ... effected by a movement of gold" (pp. 28-29). Second, "since gold movements ... and discount rate adjustments are displaced from their central position in the process of international price adjustment, the question of 'observing the rules of the game' ... loses much of its importance" (p. 31).

Appendix E

Post-World War II Reinterpreters of the Gold Standard

In the period since World War II, economists have reexamined and reinterpreted the operation of the classical gold standard on the basis of new evidence and new theoretical and statistical tools. The principal

83

The Gold Standard: The Traditional Approach

areas of research are the adjustment mechanism, the role of capital flows, the managed gold standard, and the rules of the game. In the reconsideration of the international adjustment mechanism, several approaches can be distinguished. The Keynesian open-economymultiplier approach (Ford) explains most of the adjustment to the transfer of capital before World War I in terms of changes in economic activity rather than relative price levels. Extension of the classical price-specieflow mechanism (Friedman and Schwartz) focuses on relative price levels and interest rates in the adjustment process. The monetary approach to the balance of payments (Triffin, Williamson, and McCloskey and Zecher) integrates elements of both Keynesian and classical mechanisms and views gold flows by themselves, rather than the effects they have on price levels, incomes, and interest rates, as the equilibrating mechanism in the adjustment of the balance of payments. A major reexamination of the role of capital flows in the transmission mechanism (Morgenstern) raised serious doubts about the classical theory, though the analysis in turn was criticized (Borts). Evidence favorable. to the traditional approach was also presented (Bloomfield). Evidence on the managed gold standard amplified the view that London managed the prewar gold standard (Lindert, Sayers, Goodhart). Other studies showed that the rules of the game were largely violated before 1914 (Bloomfield) and that they were inconsequential (McCloskey and Zecher). The Adjustment Mechanism Ford (1962) downplayed the role of price-level and monetary change in the explanation of the adjustment of the balance of payments under the prewar gold standard. He stressed three themes: (1) the key element in the adjustment mechanism for Great Britain was the change in real income, working through an open-economy-multiplier process; (2) the important link between Great Britain and periphery nations via lending and exports worked primarily through changes in income; (3) there was an asymmetry between the gold standard experience of Great Britain and Argentina (an example of a periphery country). For Great Britain the gold standard mitigated the adjustment to external disturbances, for Argentina the gold standard aggravated it. These divergent experiences reflected the operation of fundamentally different gold standard financial institutions and the presence and absence of "other favorable circumstances." Ford was highly critical of the stylized facts of the traditional approach summarized in the Cunliffe report. His objections were that the report omitted the crucial role of income effects; placed weight on a link between changes in interest rates and economic activity that he believed to be tenuous; treated relative prices as the primary mechanism of adjustment in the balance of trade, based on the doubtful assumption of

84

Michael D. Bordo

elastic demands for imports and exports; inappropriately ignored the feedback effects of changes in income of the major trading nation-Great Britain-on other countries; and made no mention of the important role of foreign lending. Instead, Ford emphasized the key role of changes in income in the balance-of-payments adjustment mechanism.70 A disturbance to the balance of payments such as the exogenous decline in exports would reduce much of the initial balance-of-payments deficit by a decline in real income, working via the multiplier, and would reduce the demand for imports without any change in price levels. 71 The extent to which the balance of payments was equilibrated, without requiring a gold outflow and changes in the money supply, depended on the relative sizes of the marginal propensities to import and to save. 72 In addition, feedback effects of changes in other countries' income due to the fall in British demand for their products would (depending on the size of the other country) also facilitate the adjustment. The second adjustment mechanism was change in the money supply. To the extent the balance of payments was not equilibrated by income change, the resultant gold outflow would reduce the domestic money supply, which would raise interest rates and lower domestic economic activity, imports, and the balance-of-payments deficit. The rise in interest rates would also induce a short-term capital inflow, providing a temporary cushion for the balance of payments. A rise in the central bank's discount rate would further stimulate the reduction in domestic activity and encourage a short-term capital inflow. However, Ford did not regard the Bank of England's playing by the rules of the game as the key element in the process, and he downplayed the accommodating role of centralbank policy in other countries. 73 Finally, Ford presented evidence he regarded as consistent with the key role of income in the adjustment process for Great Britain. Comparing deviations from a nine-year moving average of exports, imports, and income from 1870 to 1914, he found that the parallelism between movements of deviations from nine year moving averages of exports, imports and national income is marked, with some slight tendency for imports to lag behind exports, ... these movements provide powerful evidence both for the exports-incomeimport automatic adjustment mechanism and for the view that in most British booms and slumps variations in export values were a vital factor. (Ford 1962, p. 61)74 According to Ford, British loans to developing nations such as Argentina, Australia, and Canada were primarily transferred through changes in real income, in contrast to the classical approach which emphasized changes in sectoral prices. A British loan to a country like Argentina

85

The Gold Standard: The Traditional Approach

would lead to an increase in imports from Britain and an increase in debt service. When the investment projects in the borrowing country matured, production of exportables would increase, some of which would be imported by Britain. The resultant rise in Argentine incomes would then generate the revenues necessary to service the foreign debt as well as increase demand for British goods. The arrangement worked well in the long run, but in the short run problems could arise for the borrowing country. If the amount lent abroad temporarily exceeded Britain's current-account surplus, a gold outflow from Britain would occur, and the Bank of England would raise Bank rate. As a result, the borrowing country would experience a gold outflow. Important investment projects might be halted before completion leading to a balance-of-payments crisis if the borrower were unable to meet the debt-service obligations. Ford examined in detail the cyclical adjustment mechanism in Argentina under the gold standard (and under inconvertible paper). The typical cycle can be described as follows: A rise in the price of Argentina's primary staple exports, generated by an increase in foreign demand, would produce a current-account surplus, a rise in economic activity, and a gold inflow. The gold inflow would lead to a rise in the domestic money supply. Some of the increased expenditure would go to the production of additional exports, some to the purchase of imports, and some into the nontraded sector-primarily for the purchase of land. A land boom would develop, followed by further foreign investment and further expansion. Ultimately, the boom would be choked off, either by a reduction in foreign economic activity that reduced the demand for Argentinian exports, or by a rise in the discount rate by the Bank of England to offset a gold outflow. Furthermore, were speculation in land to get out of hand, there would ultimately be a collapse in land prices and a domestic liquidity crisis associated with bank failures. Foreign lending would be discouraged. If the internal drain were accompanied by an external drain, then domestic gold reserves would be insufficient to withstand the onslaught and the country would suspend convertibility. This was the sequence of events in 1885 and 1913. Ford argued that the gold standard experience of periphery countries such as Argentina was considerably less favorable than that of center countries such as Britain. For two reasons the cycle in British economic activity was dampened by the gold standard whereas the cycle in Argentina was aggravated. First, Britain's position as the center of the world's money market meant that in the face of an external drain, a rise in Bank rate would draw on short-term capital and sterling balances in London without loss of gold, while Argentina lacked the cushioning financial institutions. 75 Second, it was easier for a creditor nation to obtain immediate relief from external pressure by reduced foreign lending, "whereas

86

Michael D. Bordo

. . . in a period of stringency it was difficult or even impossible for a debtor country (with a past history of depreciation) to offset gold exports by increasing its borrowing abroad" (p. 182). In their monumental study of U.S. monetary history, Friedman and Schwartz (1963) treated the gold standard in a traditional way in their analysis of the role of monetary forces and of the price-specie-flow mechanism to explain balance-of-payments adjustment under the gold standard; in their application of the commodity theory of money to explain secular price movements; and finally, in their discussion of the role of central-bank management under the gold standard. They described the role of the money supply under an international specie standard that the United States adhered to from 1879 to 1933, as follows: The amount of money in anyone country must be whatever is necessary to maintain international balance with other countries on the same standard, and the amount of high-powered money will alter through imports and exports of specie in order to produce this result . . . the amount of high-powered money is a dependent rather than an independent variable, and is not subject to governmental determination. (Friedman and Schwartz 1963, pp. 51-52) Moreover, for a country which is an economically minor part of the gold standard ... the major channel of influence is from fixed rates of exchange with other currencies through the balance of payments to the stock of money, thence to the level of internal prices that is consistent with these exchange rates.... [However] the links have much play in them, so that domestic policies can produce sizable short-term deviations in the stock of money from the level dictated by external influences. (Pp. 89-90) Under a flexible-exchange-rate regime, by contrast, such as the greenback period from 1862 to 1879, "the amount of high-powered money is determined by governmental action" (p. 51). Adjustment to both external and internal disturbances was facilitated by the classical relative-price-Ievel adjustment mechanism and capital flows. The events of the period 1879-82 are analyzed in these terms. Good harvests in 188Q-81 led to an increase in U.S. exports. The resulting increase in demand for dollars implied a relatively higher U.S. price level consistent with balance-of-payments equilibrium. 76 Pending the rise in prices, a gold inflow ensued that led to a rise in the money stock and the price level. At the same time, a gold outflow from Great Britain led to a monetary contraction and a decline in the price level. As a consequence the Bank of England raised Bank rate, reversing the gold flow to the United States.

87

The Gold Standard: The Traditional Approach

The gold inflow was a passive reaction which temporarily filled the gap in payments. In its absence, there would have had to be an appreciation of the dollar relative to other currencies-a solution ruled out by the fixed exchange rate under the specie standard-or a more rapid rise in internal U.S. prices. At the same time, the gold inflow provided the basis and stimulus for an expansion in the stock of money and thereby a rise in internal prices at home and downward pressure on the stock of money and prices abroad sufficient to bring an end to the necessity for large gold inflows. It would be hard to find a much neater example in history of the classical gold standard mechanism in operation. (P. 99) Other episodes are treated similarly: prices and incomes, aided by capital flows, adjust to maintain external balance. 77 The brunt of the adjustment was sustained in some cases by changes in the money stock, in others by changes in velocity or real output. Economically, these were the channels whereby a necessary adjustment was worked out. They were not the forces determining what adjustment was necessary.... The discipline of the balance of payments under the gold standard enforced that adjustment and determined its size. (P. 101 )78 Friedman and Schwartz applied the classical commodity theory of money to explain secular price-level movements: Under a specie standard confined to a single country, or for the world as a whole under an international standard, the existing amount of specie is determined by the available physical stock plus the relative demand for monetary and other uses; and changes in the amount of specie, by relative costs of production of specie and other goods and services. (P. 52) They explained the secular deflationary episode of 1879-96 by [a] combination of events, including a slowing of the rate of increase of the world's stock of gold, the adoption of the gold standard by a widening circle of countries, and a rapid increase in aggregate economic output, ... despite the rapid extension of commercial banking and of other devices for erecting an ever larger stock of money on a given gold base. (P. 91) The subsequent turnaround in prices and secular-inflation episode is explained by fresh discoveries of gold in South Africa, Alaska, and Colorado combined with the development of improved methods of mining and refining, especially the introduction of the cyanide process. These occurred during a period when there were few further important extensions of the gold standard yet a continued development of devices for "economizing" gold. (P. 91)

88

Michael D. Bordo

Moreover, the period of secular deflation "was an important factor in stimulating the search for gold and for economical techniques for extracting gold from low-grade ore" (p. 188). Finally, in their discussion of the sterilization of gold inflows in the 1920s by the Federal Reserve system, Friedman and Schwartz explain how violations of the rules of the game weakened the adjustment mechanism of the gold standard. The sterilization of gold could be justified as a means of insulating internal monetary conditions from external changes. Its international effect, however, was to render the maintenance of the international gold standard more difficult. Suppose all countries linked by a gold standard were to sterilize gold flows. Gold flows would then set in train no forces tending to bring them to a halt or to reverse them. The system could last only as long as the flows resulted from purely temporary imbalances of sufficiently small magnitude to right themselves before draining the countries losing gold of their reserves. The effect would be to insulate the countries from minor adjustments at the cost of letting them accumulate into major ones. (P. 283) Friedman and Schwartz attributed the Federal Reserve system's failure to stem the banking crisis of 1931 to its failure to follow the classical medicine prescribed by Bagehot for central-bank operations in the face of both an external and an internal drain-to lend freely but at a high discount rate (p. 395). Emphasis on the "stylized facts" of the gold standard on intercountry adjustment via specie flows that produced relative price-level adjustments, aided by capital flows and by central banks following the rules of the game, failed, according to Triffin (1964), "to bring out the broader forces influencing the overall pace of monetary expansion on which individual countries were forced to align themselves" (p. 2). Evidence damaging to the traditional story included: "enormous degree of parallelism-rather than divergent movements-between export and import fluctuations for anyone country, and in the general trend of foreign-trade movements for the various trading countries. . . from 1880 to 1960" (p. 3); and "overall parallelism-rather than divergence-of price movements, expressed in the same unit of measurement, between the various trading countries maintaining a minimum degree of freedom of trade and exchange in their international transactions" (p. 4); downward wage rigidity among countries that maintained exchange stability; Bloomfield's (1959) evidence of the failure of most central banks to play by the rules of the game; the ineffectiveness of changes in discount rates in many countries to stem capital flows or change relative prices; and the important role of long-term capital flows in maintaining enduring balance-of-payments disequilibrium without relative price adjustment.

89

The Gold Standard: The Traditional Approach

Triffin accordingly painted a different picture of how the gold standard worked. The most important aspect of the gold standard was exchange-rate stability maintained by "pressures for international harmonization of the pace of monetary and credit expansion [between central banks] similar . . . to those which. . . limit divergent rates of expansion among private banks within each national monetary area" (p. 11), enforced by the constraint of convertibility into gold of fiduciary money. 79 Given stable exchange rates, then national export prices remained strongly bound together among all competing countries, by the mere existence of an international market not broken down by any large or frequent changes in trade or exchange restrictions. . . . National price and wage levels also remained closely together internationally, even in the face of divergent rates of monetary and credit expansion, as import and export competition constituted a powerful brake on the emergence of any large disparity between internal and external price and cost levels. (P. 10) As a consequence, monetary expansion generating inflationary pressures could not be contained within the domestic market, but spilled out directly . .. into balance of payments deficits rather than into uncontrolled rises of internal prices, costs, and wage levels. These deficits led, in turn, to corresponding monetary transfers from the domestic banking system to foreign banks, weakening the cash position of domestic banks and their ability to pursue expansionary credit policies. (Pp. 10-11) Central banks could only temporarily slow down the adjustment process by engaging in offsetting open-market operations or using other tools of monetary policy, because ultimately their international reserve ratios would decline. Thus, in the gold standard world of fixed exchange rates, price levels were closely linked together and the balance-of-payments deficit (surplus) reflected both money-market disequilibrium and the method by which it was eliminated. In an approach similar to that of Triffin, Williamson (1961, 1963) reinterpreted U.S. experience under the gold standard. He urged analysis of the balance of payments in a general-equilibrium context, with the long-swing cycle in the growth of real output determining specie and capital flows. 80 Increased real growth would lead to both an excess demand for goods (a balance-of-trade deficit), an excess supply of bonds (a capital inflow), and an excess demand for real balances. The excess demand for money would be satisfied by a specie inflow, with little change in the price level, accompanied by a long swing in capital inflows (Williamson 1961, p. 379). The external balance was both a cause as well as a

90

Michael D. Bordo

reflection of the long swing, since in the 1830s it was British demand for U.S. cotton that was the key source of the long swing in output, which in turn induced British investment in railroads and canals. 81 In an important article applying the recently developed monetary approach to the balance of payments to the operation of the classical gold standard from 1880 to 1914, McCloskey and Zecher (1976) extended the challenge to the traditional approach intimated by Marshall in the 1880s, endorsed by Whale in the 1930s, and repeated by Triffin and Williamson. The monetary approach states that for an open economy with fixed exchange rates, the national stock of money, rather than prices, adjusts to changes in the public's demand for money (see Frenkel 1971; Johnson 1976; Mundell 1971). Contrary to the Hume price-specie-flow mechanism-which postulates significant lags in the adjustment of price-because of instant arbitrage, according to the monetary theory, no lags are observed in the adjustment of world prices. In the most rigid version of the theory, an increase in the demand for money ca~not reduce prices because prices of internationally traded goods are determined in world markets and kept comparable in different countries by international arbitrage, and prices of domestic goods and services are kept in line with prices of internationally traded goods by domestic arbitrage. The reduction in the public's demand for goods and securities leads to reduced imports and expanded exports on the goods side and to higher interest rates and capital imports on the securities side. The current account or the capital account or both move into surplus. To prevent appreciation of the currency, the monetary authority buys foreign exchange from its nationals, paying out newly created high-powered money. The increase in high-powered money leads to a multiple expansion of the domestic quantity of money which continues until the public's demand is satisfied. In open economies on a fixed exchange rate, a once-for-all increase in the quantity of money in one country and a decrease in another would produce a balance-of-payments deficit in the first and surplus in the second and lead to a flow of money to the second until equilibrium was reestablished. If a monetary authority in one country alone increased high-powered money, that would be equivalent to an increase in the world money supply. That country would experience a temporary balance-of-payments deficit until the world money supply was redistributed in proportion to the size of the country of issue, and the world price level would rise accordingly. In the long run, domestic monetary policy in a small country has a negligible influence on international prices, although in the short run the monetary authority can affect its price and income level by open-market sales (purchases) equal to its balance-of-payments surplus (deficit) that will maintain the national money stock below (above) its equilibrium value. The closer the Ijnks among world commodity markets, the higher the degree of capital mobility, the less scope for

91

The Gold Standard: The Traditional Approach

independent monetary policy in the short run. The greater the elasticity of substitution between traded and nontraded goods, the less successful will such policy be. In the long run, however, independent monetary policy is inconsistent with fixed exchange rates. McCloskey and Zecher tested the key assumptions of commodity arbitrage by examining correlations among price changes between countries and between regions within countries. For traded goods such as wheat, they found synchronous correlations equally high between regions as between nations, unlike the case of nontraded goods such as labor services and bricks. For overall price indexes, they found a significant correlation between the U.S. and U.K. wholesale price indexes, less so for GNP deflators, and even less for consumer price indexes. The higher correlation for the wholesale price than for the other index undoubtedly reflects the larger share of traded goods in the former. 82 They conclude: What has been established here is that there is a reasonable case . . . for the postulate of integrated commodity markets between the British and American economies in the late nineteenth century, vindicating the monetary theory. There appears to be little reason to treat these two countries on the gold standard differently in their monetary transactions from any two regions within each country. (McCloskey and Zecher 1976, pp. 379-80) They also cite less conclusive evidence in favor of capital-market arbitrage. They tested their model by comparing gold flows-predicted by a simple demand for money function less the money supply produced by domestic credit expansion-with actual gold flows and found a close relationship. In their view, we have established at least a prima facie case for viewing the world of the nineteenth century gold standard as a world of unified markets, in which flows of gold represented the routine satisfaction of demands for money. (P. 385) Capital Movements Both short-term and long-term capital movements play an important role in the traditional approach to the gold standard. Short-term capital flows were to act as an equilibrating mechanism, to economize on gold flows and to reduce the burden of adjustment by changes in relative price levels. Private short-term capital movements, assumed to be highly responsive to interest-rate differentials (induced by changes in the discount rate), would act to equate interest rates (adjusted for exchange risk) in different money markets. Morgenstern examined the evidence in favor of the "solidarity hypothesis," that arbitrage would ensure uniformity of interest-rate differentials to exchange risk (measured by the difference of

92

Michael D. Bordo

the exchange rate from the gold points), and found it to be inconsistent with the principles of the gold standard. Borts, in criticizing both his methodology and data, put Morgenstern's conclusions into serious doubt. Bloomfield found evidence that private short-term capital flows responded to interest differentials and expected exchange-rate changes as predicted. Long-term capital flows, a key disturbing force in the traditional approach, were believed to have been well accommodated by the gold standard balance-of-payments adjustment mechanism. They allowed both developing debtor nations and mature creditor nations to run persistent balance-of-payments disequilibria without requiring adjustment in the balance of trade. Bloomfield's study supported the integral role of long-term capital flows in the development of the "Atlantic economy." Morgenstern (1959) examined short-term-interest-rate data of different maturities, exchange rates, and the gold points for four key gold standard countries: Great Britain, France, Germany, and the United States in the periods 187~1914 and 1925-38, subjecting the "assumption of international solidarity of money markets" to three tests. 83 The first test compared derived exchange rates, based on cross rates in third markets, with the actual exchange-rate series. If arbitrage were effective, differentials between the series would vanish. Morgenstern found evidence of differentials persisting in both periods, but more so after World War I, with the greatest deviations occurring in periods of crisis or disturbance such as the 1890 Baring crisis. The second test identified deviations of exchange rates beyond the median gold import and export points. Violations persisted for long periods, with greater frequency of deviations beyond the gold export point than beyond the gold import point. The· third test compared market-interest-rate differentials with the "absolute maximum permissible" differential determined by the distance between the gold points. Violation of the principle occurred in both prewar and postwar periods. Morgenstern concluded that we ought to view the period of the classical gold standard as inadequately described by the typical mechanism at least in one respect: the interaction between two and more money markets via exchange rates and interest rates is not nearly as precise and rigid as postulated (Morgenstern 1959, p. 569) He explained these results by "friction" and central-bank intervention, suggesting the replacement of the interpretation of the gold standard as a mechanism by the notion that central banks and other market participants engage in game strategy in a struggle for gold. Borts (1964) criticized both Morgenstern's methodology and his data. Borts attributed the results of the first test comparing cross with own exchange rates to the

93

The Gold Standard: The Traditional Approach

methods of making exchange quotations. Morgenstern's data were either exchange quotations in the form of single prices at which brokers cleared the market or the average of bid-ask spreads. For Borts the bid-ask spreads were likely measures of transactions cost; hence "what appears to be an opportunity for arbitrage profit could be the difference between the clearing price and the prevailing practice of giving quotations" (Borts 1964, p. 225). Borts found the results of the second test comparing the spot exchange rate with the gold points ambiguous. In deriving the gold points, Morgenstern assumed circumstances favoring his results since he took into account neither different methods of covering exchange risk nor frequent operations by the monetary authorities on the gold points. These ambiguities could be sorted out, according to Borts, by directly examining gold movements "in an effort to confirm the position of the exchanges with regard to the gold points" (p. 227).84 Finally, Morgenstern's third test comparing interest differentials to the maximum exchange risk on uncovered funds was faulty, in Borts's view, because (1) "[he converted] the percent movement in the exchange rate into a percent difference on one-year paper. The exchange risk. . . has no time dimension and exists no matter what the maturity of the paper held" (p. 227); (2) he did not account for possible forward cover. 85 Bloomfield examined the role of both short-term (1963) and long-term (1968) capital flows under the pre-1914 gold standard. He suggested that private short-term capital movements served to equilibrate the balance of payments in the short run by tending to reduce the size of gold flows or acting as substitutes for changes in official exchange holdings (Bloomfield 1963, p. 44). To test the latter hypothesis, he compared the signs of the first differences of annual changes in the stock of net foreign short-term assets of commercial banks with annual changes in the stock of centralbank gold and foreign-exchange reserves of the Scandinavian countries and found the postulated negative relationship (p. 58).86 Long-term capital flows in the classical gold standard period in the form of portfolio investment came mainly from Britain and France, followed by Germany; the bulk of the funds went to the developing countries of the new world to finance the development of infrastructure and production and exportation of primary products. Following the work of Williamson (1964), Cairncross (1953), O'Leary and Lewis (1955), and Thomas (1973), Bloomfield (1968, pp. 18-34) found evidence of a longswing cycle in long-term capital movements. Consistent with their theories, he found for debtor countries positive correlations between capital imports and indicators of domestic investment such as domestic building, and between capital imports and net immigration; for creditor countries a negative correlation between capital exports and domestic investment, and a positive one between capital exports and net emigration. 87 These

94

Michael D. Bordo

results, plus the evidence that similar long-swing movements in the United States and Canada in turning points were inversely related to swings in British building and economic activity in general, gave support to Thomas's (1973) thesis of an Atlantic economy. Finally, contrary to theory, interest rates did not explain movements in annual data of British capital exports and U.S. capital imports. In the multiple regressions that Bloomfield estimated, domestic and foreign investment activity were the significant regressors. The Managed Gold Standard The mainstream view of the classical gold standard that emerged from the interwar period (appendix D) was that it evolved into a sterling standard. The concentration of world capital, commodity, and gold markets in London made the pound sterling an attractive reserve asset in addition to gold and made it easier for the Bank of England to control its gold-reserve ratio by altering Bank rate, in the process affecting the policies of other central banks and influencing economic conditions both at home and abroad. 88 In the post-World War II period, the degree of management of the prewar standard was further explored. Key currencies other than sterling were shown to have been important in the pre-World War I period, though sterling's role was still predominant, and the Bank of England's use of the Bank-rate weapon was deemed to be less effective than traditionally believed (Lindert 1969). The ways in which the Bank of England managed the gold standard were described (Sayers 1936, 1957; Goodhart 1972). Following Bloomfield (1963), Lindert (1969, pp. 13-27) found that holdings of several major currencies~the pound, the franc, and the mark-represented an important and growing fraction of the international reserves held by many countries in the period 1900-1913. 89 As expected, London was the primary reserve center for the world but francs and marks were popular on the Continent. According to Lindert, key currencies were held for the interest income they earned; they involved lower transaction and transportation costs than gold, and maintenance of balances in a foreign currency such as sterling often gave easier access to credit in the London money market. Perhaps the key reason these currencies were held was that their good brand name guaranteed with certainty that they could be converted into gold on demand. In the case of the pound, the location in London of the international money market and the world's gold market likely enhanced its brand name. Because their currencies were widely held, the central-reserve countries could run larger balance-of-payments deficits than otherwise and

95

The Gold Standard: The Traditional Approach

could maintain them longer. However each central bank was sensitive to its gold-reserve ratio and when it declined the bank would react by raising its discount rate. 90 Lindert found that when Great Britain raised Bank rate, other central banks on the Continent would respond, but London had stronger "pulling" power and could always attract short-term funds from the Continent. 91 A flow of short-term funds proceeded from peripheral European nations, running balance-of-payments surpluses to Paris and Berlin, and then to London. The asymmetry in discount-rate drawing power may be explained by the fact that for the center countries, short-term foreign assets tended to be less liquid than short-term liabilities; "since tighter monetary policy tends to stimulate shifts toward liquid assets, banks would react by seeking greater key-currency balances at the expense of bills on lesser centers" (Lindert 1969, p. 78). Thus, the indisputable position of London as the dominant financial center during the prewar years meant that "other countries, had, therefore, to adjust their conditions to hers."92 After the publication of Lombard Street (Bagehot [1873]1969), the Bank of England began to take seriously its responsibilities for both maintaining convertibility and preventing domestic monetary instability, doing so not by increasing its gold reserves, as Bagehot suggested, but by altering its discount rate whenever its gold reserves were threatened (Sayers 1951, pp. 109-10). Sayers described the prewar techniques used to make Bank rate "effective" in the sense of linking it tightly to short-term market rates. The methods included open-market operations, eligibility requirements, and the switch to a penalty rate in 1878. 93 In addition, under special circumstances, when it feared the internal repercussions of raising Bank rate, the Bank would protect its gold reserve by using "gold devices"-direct operations in the gold market. It is generally agreed however that the Bank achieved full control over its reserves after 1890 (Presnell 1968). According to the traditional approach, a rise in Bank rate would equilibrate the balance of payments via two principal channels: by inducing a short-term capital inflow (reducing an outflow) and by checking domestic economic activity, the domestic price level, and the price of imports. Lindert (1969, pp. 43-44), Bloomfield (1959, p. 42), and Goodhart (1972, chap. 15) evaluated the evidence for the domestic channel as indicating at best a weak and protracted adjustment with the case for the link via capital flows sacrosanct. The case made against the domestic channel was twofold: the limited response of the domestic money supply to changes in the Bank of England's gold reserve, and the limited response of domestic economic activity to changes in the interest rate. With respect to the money supply, Goodhart (1972, p. 208) was unable

96

Michael D. Bordo

to detect any close, positive association between the cash base (reserves) of the commercial banks, represented by bankers' balances at the Bank of England, and the Bank's gold reserves. According to him, the direction of causation was the reverse of the traditional one. An increase in economic activity would lead to an increase in commercial-bank lending and deposits, and the increase in bank reserves required to maintain stable reserve ratios would be supplied by the Bank of England at the expense of its other discounts, thus producing both a lower gold-reserve ratio at the Bank of England (the proportion) and a higher discount rate. The rise in the discount rate would then lead to a gold inflow restoring the Bank's proportion. 94 Goodhart concluded (1972, p. 219): Indeed, on this view, the great years of the gold standard (189~1914) were remarkable, not because the system enforced discipline and fundamental international equilibrium on this country by causing variations in the money supply, but because the system allowed for the development of such large-scale, stabilising and equilibrating, shortterm, international capital flows, that autonomous domestic expansion was rarely disrupted by monetary or balance of payments disturbances. With respect to domestic economic activity, Tinbergen (1950, p. 133) found that the influence of interest rates on the course of investment activitywhich is the chief influence interest rates exert, according to our results-is only moderate. A rise in interest rates depresses investment activity, but only to a modest extent. and Pesmazoglu (1951, p. 61) that variations ... in the long-term rate of interest did not have an important influence on fluctuations of British home investment between 1870 and 1913. 95 The Rules of the Game According to the traditional approach, the key objective of monetary policy was to maintain convertibility into gold and to use monetary policy, specifically the discount rate, to facilitate internal adjustment to external disequilibria. However, Bloomfield (1959, pp. 25-26) found that while central banks were primarily concerned with maintaining convertibility, their policy actions were discretionary, not automatic: Not only did central banking authorities ... not consistently follow any simple or single rule or criterion of policy, or focus exclusively on considerations of convertibility, but they were constantly called upon to exercise, and did exercise, their judgment on such matters as whether or not to act in any given situation and, if so, at what point of

97

The Gold Standard: The Traditional Approach

time to act, the kind and extent of action to take, and the instrument or instruments of policy to use. . . . It does indicate that discretionary judgment and action were an integral part of central banking before 1914, even if monetary management was not oriented toward stability of economic activity and prices in the broader modern sense. In a test of the extent to which central banks under the pre-1914 gold standard played by the rules of the game, Bloomfield interpreted the rules as meaning' 'that central banks were supposed to reinforce the effect of these flows [gold flows] on commercial bank reserves, not merely not to neutralize them. This implied ... that central banks were supposed to lower their discount rates in the face of persisting gains of gold. . . and to raise them when there were persisting losses." Such a policy would have the effect of "increasing central bank holdings of domestic earning assets when holdings of external reserves rose, and of reducing domestic assets when reserves fell" (Bloomfield 1959, p. 47). Following Nurkse's approach in his examination of central-bank behavior in the 1929-38 period, Bloomfield compared year-to-year changes in international and domestic assets for twelve central banks in the 1880-1914 period and found that, in the case of every central bank, the changes in the two classes of assets were more often than not in the opposite direction (wit!1 the Bank of England coming close to being the exception to the rule). Thus, Far from responding invariably in a mechanical way, and in accord with some simple or unique rule, to movements of gold ... , central banks were constantly called upon to exercise, and did exercise, discretion and judgment in a wide variety of ways. Clearly the pre-1914 gold standard system was a managed and not a quasi-automatic one from the viewpoint of the leading individual countries. (Bloomfield 1959, p.60) Based on their reinterpretation of the classical gold standard according to the monetary approach to the balance of payments, McCloskey and Zecher (1976) denied that the Bank of England could have "acted as conductor of the international orchestra" as in Keynes's ([1930] 1971) description, but "was no more than the second violinist, not to say the triangle player, in the world's orchestra" (McCloskey and Zecher 1976, pp. 358-59). The monetary theory holds that the world's economy is unified by arbitrage and that the world's price level is determined by the world's money supply. Then the Bank of England's potential influence on prices (and perhaps through prices on interest rates) depended simply on its power to accumulate or disburse gold and other reserves available to support the world's supply of money.

98

Michael D. Bordo

. . . Only by decreasing the securities and increasing the gold it held . . . could the Bank exert a net effect on the world reserves . . . Had the Bank in 1913 sold off all the securities held in its banking department it would have decreased world reserves by only 0.6 percent; had it sold off all the gold in its issue department, it would have increased world reserves by only 0.5 percent. (McCloskey and Zecher 1976, p. 359) Finally, according to McCloskey and Zecher, the central banks of the world ignored the rules of the game-stipulating that a deficit in the balance of payments be accompanied by contractionary monetary policy, a surplus by expansionary policy-because the rules were "inconsequential." According to the monetary theory "neither gold flows nor domestic deflation have effects on prevailing prices, interest rates, and incomes" (p. 361) since the central bank of a country adhering to the gold standard could only control the composition of the monetary base as between international reserves and domestic credit, not its total amount.

Notes 1. Some of the material covered is drawn from, and may overlap, earlier surveys by Viner (1937), Fetter (1965), and McCloskey and Zecher (1976). 2. The meaning of the price of gold is its relative or real price or the purchasing power of gold. This meaning is not explicitly stated by all writers but presumably it is what they intended. It is the fixed mint price of gold in terms of national currency divided by some commodity price index. 3. For some writers changes in incomes rather than changes in relative price levels produced the adjustments. 4. Some would argue that Adam Smith ([1776] 1976) should be included in this list. Smith subscribed to most of the views expressed by the other classical economists. He viewed the world specie stock and its exchange value in the long run as determined by the richness of gold and silver mines. He also stressed the social saving of using paper money for specie up to the point of convertibility. However, he did not discuss the price-specie-ftow mechanism and his belief in the real-bills doctrine (see Mints 1945) has resulted in the downgrading of his contribution to the traditional view. Recently, however, Girton and Roper (1978) and Laidler (1981) have argued that Smith may have been correct after all, if interpreted according to the recent monetary approach to the balance of payments. According to these authors, Smith viewed a country such as contemporary Scotland as a small open economy on a fixed exchange rate with an exogenously determined price level. Under such circumstances, the quantity of money would adjust to the demand for money-a result consistent with both the real-bills approach and the absence of any change in the terms of trade. 5. This discussion draws heavily on Bordo 1983. 6. Cantillon [1931] 1964, bk. 2, chap. 16. Indeed the intrinsic value of the precious metals and hence long-run supply is determined by the cost of production of the least productive mine (p. 101), Le., by marginal cost. Temporary variations of the exchange

99

The Gold Standard: The Traditional Approach

value of money from its intrinsic value can be caused by changes in the demand for precious metals (for nonmonetary uses), but in the long run exchange value will equal intrinsic value (p. 97). 7. Other aspects of the traditional view mentioned in Cantillon's Essai include a discussion on capital mobility (pp. 191-93), the gold points (pp. 253,255,257,261), the use of bills of exchange to settle international balances (pp. 229,245,247), and the operation of the forward exchange market (p. 259). 8. Viner (1937, pp. 316,319) argued that the distinction between the law of one price as pertaining to the equality of prices stated in a common currency of identical traded goods, allowance being made for transport costs, and the price-specie-flow mechanism, which involved changes in the relative prices of import and export goods (the terms of trade), was held by all the classical writers. Samuelson (1971) viewed the "inconsistency" as an error of interpretation by Viner and others. According to him, when prices in each country are measured relative to wages, the equality of identical-traded-goods prices will hold after a gold discovery initially affects the money supply and prices in one country. 9. Also paper has the advantage that it is flexible and can be supplied quickly in periods of crisis (Ricardo [1816] 1951, p. 58). 10. In addition the issue of bank notes will produce the same effect as a gold discovery (Ricardo [1811] 1951, p. 55). 11. Ricardo also described the force of arbitrage in maintaining equality between prices (the value of money) in the country and in London ([1811] 1951, p. 87). 12. Because it was legal to export bullion, but illegal to export coin, the price of bullion would initially rise above the mint price. Eventually, however, people would evade the prohibition and melt coin into bullion (Ricardo [1811] 1951, p. 64n). 13. According to Sayers (1953), Ricardo perceived that the essential condition of a gold standard is not gold coinage, but convertibility into gold for international transactions. 14. Chronologically, Thornton preceded Ricardo by several years, but it is convenient to present his views following those of Ricardo. 15. The size of the gold reserve should be determined by the degree of confidence "between independent countries" and the "largeness of the balance between the independent places" (Thornton [1802] 1978, pp. 155-56). 16. There may, however, be distribution effects between gold mining and other countries since the reduction in the use of gold as money will reduce the price levels of mining countries relative to those of the rest of the world. In addition, there will be first-round effects depending on how the money is issued-by private bankers or the government. Money issued by bankers would lead to a fall in the interest rate, a capital outflow, and a gold outflow, with no effect on the price level. If issued by government or by private manufacturers, the initial effect would be on domestic prices, leading to a current-account deficit and a gold outflow. 17. Here Mill ([1865] 1961, p. 625) cited Ricardo (Principles, 3rd ed., p. 143). "Gold and silver having been chosen for the general medium of circulation, they are, by the competition of commerce, distributed in such proportions amongst the different countries of the world as to accommodate themselves to the natural traffic which would take place if no such metals existed, and the trade between countries were purely a trade of barter." 18. This section is based on Bordo 1975. Note that Jevons ([1884] 1964) considered the same issue as Cairnes and in the course of his investigation constructed a price index to measure the extent of depreciation of the value of gold caused by the discoveries. 19. Cairnes also tested Ricardo's theory of comparative advantage in predicting Australia's switch from being a net exporter to a net importer of agricultural products, and the Cantillon transmission mechanism to predict the dispersion of price changes between different commodity groupings. 20. Cairnes argued that England and the United States, because of their efficient banking and credit systems, should gain relative to France and the rest of the Continent

100

Michael D. Bordo

since they would require less gold to finance the necessary price rise. The gold flowing out of England into France would serve to displace silver, since France was on a silver standard. The displacement of silver would for a time act as a parachute preventing French prices from rising until gold completely replaced silver; at the same time the released silver would flow eastward and into the silver currencies of Asia, augmenting gold flowing there directly. At this point in an argument similar to that of Jevons ([1884] 1964), Cairnes pointed out that the "parachute effect" would not be as important as Chevalier (1859) maintained in preventing a rise in French prices, since gold and silver were substitutes, so that as gold currency substituted for silver, the price of silver would tend to fall along with the price of gold. 21. See Viner 1937 and Fetter 1965 for a complete history of the debates. 22. That was the recommendation of the currency school which argued that a mixed currency--one consisting of both specie and notes-should be made to operate as if it were a pure specie standard. See Fetter 1965, p. 130. 23. This was the banking school's position. For them convertibility into gold and free competition in banking were sufficient to maintain an adequate money supply consistent with both internal and external balance. See Viner 1937, pp. 222-24. 24. See Viner 1937, pp. 264-70, and White (1981) who has reformulated the currencybanking schools debate as turning on the question whether to centralize the right of note issue in a single institution or to allow competition ("free banking"). 25. In addition, Bagehot was in favor of publication of the accounts of the Banking Department of the Bank of England and employing more professionals and fewer amateurs in the government of the Bank ([1873]) 1969, pp. 302, 72). 26. Also Marshall stated that "a person who had to bring home the returns of any sales in a country had to elect what commodity he would bring, and the question whether he should bring lead or tin was governed ... by exactly the same conditions as whether he should bring lead or gold. If after allowing for expenses of carriage you get a little more by bringing home the lead and selling it than by bringing home the tin, he would choose the lead; if he would get a little more by bringing home gold and selling it, he would bring home the gold" (1926, p. 121). 27. However this is only a temporary effect; in the long run changes in gold have no effect on the rate of interest, which is determined by "the average profitableness of different business" (Marshall 1926, p. 130). 28. The views on the gold standard of Fisher's contemporary, J. Laurence Laughlin (1903), according to Girton and Roper (1978), anticipated the monetary approach to the balance of payments. A critic of the traditional approach, Laughlin disagreed with the Hume price-specie-flow mechanism which postulated lengthy lags until relative-price-Ievel differences led to corrective gold flows. He argued (in a manner similar to Angell, see p. 66) that commodity arbitrage tended to keep price levels of gold standard countries always in line. In addition, in the tradition of Adam Smith, he reversed the causation of money and prices of the classical quantity theory. According to Laughlin, for an open economy, the supply of money adjusted through the balance of payments to the demand for money, determined in turn by the "needs of trade." Thus gold did not flow to equilibrate price levels but to satisfy an excess demand for (supply of) money. 29. Fisher cites a number of reasons why prices may not be equal. "Distance, ignorance as to where the best markets are to be found, tariffs, and costs of transport help to maintain price differences.... Practically, a commodity will not be exported at a price which would not at least be equal to the price in the country of origin, plus the freight" ([1922] 1965, p.92). 30. Fisher also considered the case of bimetallism and rejected it on grounds similar to those noted by other classical writers-it tends to degenerate into monometallism whenever the market ratio of gold to silver diverges from the official ratio ([1922] 1965, pp. 123,325). He also argued against an irredeemable paper standard because of the inevitable tendency of governments to overissue (p. 131), against Marshall's symmetallism scheme because it

101

The Gold Standard: The Traditional Approach

bases the standard on too narrow a base of commodities (p. 328), and against a tabular standard: how to express money to conform to that standard was a problem (p. 335). 31. On the basis of a statistical investigation in 1933 of price-level movements of twenty-seven countries, Fisher found that the price levels of gold standard countries tended to move together, those of silver standard countries moved together, and the average price level of each group varied with changes in the relative prices of gold and silver. Moreover, he found evidence for a short-run tradeoff between price-level changes and changes in trade and employment within countries. Finally, evidence that countries not on the gold standard during the Great Depression, e.g., Spain and China, avoided the deflation suffered by gold standard countries and the concomitant contraction in output and employment, convinced him that "depressions travel internationally ... the infection is carried chiefly via the monetary standard" (1935, pp. 15-16). 32. See lonung 1979 for a discussion of Wicksell's theory of price-level movements. Basically Wicksell argued that price levels will rise cumulatively if the market rate of interest, determined in the loan market, diverged from the natural rate of interest, determined by the forces of thrift and productivity. If the market rate were below the natural rate, prices would rise cumulatively, the price rise only being arrested by a gold outflow that would reduce the banking system's reserves, causing banks to raise their loan rate. When the market rate of interest was above the natural rate, a cumulative deflation would occur. Wicksell explained periods of secular inflation and deflation in the nineteenth century using this approach. In contrast to Wicksell, the Swedish economist Cassel applied classical doctrine, explaining episodes of world inflation and deflation by the growth of the world's gold supply relative to the growth in demand for gold, the former influenced primarily by the production of new gold, the latter by the growth of real income. 33. Several years later A. G. Silverman (1931) tested the classical theory that capital exports, ceteris paribus, would lead to a rise in the price of imports relative to the price of exports. Using British data over the period 188~1913, he compared "year to year percentwith year to year absolute age changes ... for the ratio of import to export prices differences for Hobson's indirect estimates of capital exports expressed in terms of its average deviation" and found that over the whole period "an annual increase or decrease in capital exports is more often than not accompanied by an opposite change in the ratio of import to export prices." Thus he concluded that "the orthodox analysis ... does not seem to be borne out. For most of the period under consideration 'net barter terms of trade' in their yearly variations become more favorable with an increase in capital exports, and vice versa" (p. 124). His explanation for this result was that "an increase in British demand for foreign securities was apparently offset by an increased foreign demand for English goods" (p. 124). 34. In Canada, the primary recipient of the new capital, the opposite took place after 190{}-an improvement in the terms of trade and rising money wages. See the discussion on Viner, pp. 6~61. 35. According to the theory, if the proceeds of the loan were spent in the lending country then the price-specie-flow mechanism would not operate (Taussig [1927] 1966, p. 230). 36. "Still another equalizing factor is the movement of securities that have an international market. They are sold between the great financial centers in a way that replaces or lessens the transmission of gold.... In any given financial center, a tight money market and a high discount rate tend to lower the prices of . . . international securities among them.... An inflow of gold, which might be expected to take place toward the country of tight money, is replaced by an outward movement of securities" (Taussig [1927] 1966, pp. 218-19). 37. In a description of the monetary system of Great Britain, the United States, Canada, and France, Taussig demonstrated how different the response mechanism to gold flows can be, ranging from the sluggish response of the French system with its high specie-money ratio to the rapid response of the British monetary system with its low gold-reserve ratio and

102

Michael D. Bordo

loaned-up banking system. The Canadian system, with gold reserves held abroad in New York and London, gave the impression that deposits and notes increased before the gold inflow (Taussig [1927] 1966, pp. 201-7). Thus "in all countries using deposits and checks freely, the looseness of the connection between bank reserves and bank deposits leads ... to a chronological order different from that assumed in the Ricardian reasoning. An inflow of specie may follow, not precede, an enlargement of the circulating medium and a rise in prices. So it may be, at least, for a short time, even for a period of many months. Indeed, if there be further forces at work than those merely monetary, it may remain so for years" (pp. 207-8). 38. In his analysis of the massive capital inflows to Canada, Taussig stated, "If the world level of prices had remained unchanged, we should have expected in Canada a fall in the prices of imported goods, and a rise in the prices of domestic goods. Exported goods in the long run would have shown a movement similar to that of the domestic, but with a lag which would for some time keep their prices either on the same low level as the imported, or in a position intermediate between that of the imported and the domestic articles" ([1927] 1966, p. 228). 39. Beach 1935, p. 180. In an appendix, he presented evidence for the United States, supportive of this explanation, that agrees with an earlier study by A. P. Andrew (1907). 40. Viner's contribution is summarized in Taussig [1927] 1966, chap. 19. For the literature critical of Viner, see Meier 1953 and Dick 1981. 41. Until 1935, although the largest commercial bank, the Bank of Montreal, performed many of the functions of a central bank. See Rich 1978. 42. Viner disputed Laughlin's view that all price levels are tied together via arbitrage, citing large differences in the price of gold among countries (Viner 1924, p. 206). He distinguished between traded goods, whose prices are closely linked internationally, and domestic goods, whose prices are only affected indirectly. 43. However, prices of most Canadian exports were determined internationally, hence the rise in the price of international goods produced in Canada by the increased price of domestic goods and services would cause a decrease in exports. 44. Also an index of the price of services rose relative to the overall price index, as did an index of money wages relative to those in the United States and Great Britain, confirming the relative price adjustment mechanism (Viner 1924, pp. 241, 248). 45. Much of the reduced exports came from the diversion of raw materials to domestic use, while a large share of the increased imports consisted of capital goods, largely from the United States-both forces conducive to economic development. The fact that most of the proceeds of the loan were not spent in the lending country, Great Britain, is given as further verification of the classical mechanism, which otherwise would not come into play. See Taussig [1927] 1966, pp. 230 and 259, where he states, "It was to be expected that Canada, getting a growing excess of imports over exports in terms of money, should also get more imported commodities in proportion to her commodities exported. But for the verification of theory it is particularly significant that the net barter terms also become more favorable. The Canadians not only got more of physical goods in proportion to the goods they sent out, but they got, on better terms, those imported goods which may be regarded as coming in payment for their own exported goods, and which had no relation to the borrowings." 46. White 1933, p. 303. Nevertheless, he concluded that "the influence of sectional price changes as a force in the adjustment does not in the case of France appear to have played so prominent a role as is presupposed by the neo-classical doctrine .... Shifts in demand schedules were doubtless a more effective medium. No substantiation of this view could be found in the French trade statistics, but actual substantiation would in any case be impossible because fluctuations in prices as a causal factor in merchandise movements could not be excluded; it would be impossible to determine what proportion of the changes in the volume of merchandise imports and exports was due to changes in demand schedules and what proportion to changes in sectional prices."

103

The Gold Standard: The Traditional Approach

47. See Kindahl 1961 and Friedman and Schwartz 1963 for further discussion. 48. According to Graham, in the first period, exporters to the United States will obtain for a time practically the same paper prices as before the lending, and these translated into gold will be higher than before. At the same time import prices in Great Britain will rise because the gold obtained by American sellers for their products when translated into paper yields less than before the depreciation of gold. Unless the British buyers can import from some other country, the U. S. sellers will gradually be able to raise their prices. The ultimate effect will be a rise in the British price level which will appear as a relative increase in the price of domestic goods. The opposite forces will occur in the second period (Graham 1922, pp. 259-60). 49. According to Nurkse ([1944] 1978), the general picture of the adjustment process and the role of central-bank policy presented in the Cunliffe report "was one which during much of that [interwar] period dominated men's ideas both as to the actual working of the gold standard before 1914 and as to the way the gold standard should be made to work after its restoration" (p. 67). Thus the prewar gold standard was portrayed in similar terms in both the Interim Report of the Gold Delegation of the Financial Committee (League of Nations [1931] 1979) and the Macmillan report (1931). 50. The report ignored the role of fluctuations in the exchange rate within the gold points and the temporary sterilization of gold flows. 51. The rules were never formally spelled out. According to the Macmillan report (1931), "the management of an international standard is an art and not a science, and no one would suggest that it is possible to draw up a formal code of action admitting of no exceptions and qualifications, and adherence to which is obligatory, on peril of wrecking the whole structure" (par. 47). 52. Moreover Gregory doubted the ability of forward exchange markets to cover exchange risk because he felt that "just when the relative values of currencies are most uncertain and when, therefore, the advantages to be derived from the organization of a forward exchange market would be greatest, the difficulties of organizing it ... are greatest also" ([1932] 1979, p. 10). 53. This theme also appeared in the prewar literature in Keynes [1913] 1971. See the discussion below. It has played an important role in the postwar explanation of the classical gold standard's success. See, e.g., Triffin 1964, Lindert 1969, D. Williams 1968, and Palyi 1972. 54. Behind this elaborate network of financial flows was the real process of the transfer of capital to developing countries from the developed countries and the real flow of goods through the current account. According to Williams (1947, p. 155), "England's creditor position in the nineteenth century had developed gradually, along with the development of a world economy involving the division of productive effort between the older industrialized areas and the younger agricultural areas and the flow of accumulated savings from the former to the latter. The same circumstances which assigned to England the leading role in capital export made London the international money market and the Bank of England the administrator of the gold standard." 55. Important pre-1914 European examples were Austria-Hungary and Russia. In Asia, India and the Philippines represented the classic successful examples of the operation of the gold-exchange standard. 56. Calculated on the basis of the underlying trend growth rate of real income. See Cassell in the Interim Report . .. of the Financial Committee (League of Nations [1931] 1979). 57. Among the key sources of variation in monetary demand in the pre-World War I era were "the large accumulation of gold in the United States in preparation for the introduction of the gold standard" in 1879 (Cassell 1935, p. 11), and the competition by central banks to strengthen their gold reserves: "the orthodox use of gold reserves for ironing out temporary deficits in the balance of trade fell into the background and often lost all

104

Michael D. Bordo

importance in comparison with the gold movements caused by competition for gold and ultimately traceable to the artificial position given to gold in the world's monetary system" (p. 13). 58. Viner (1937) described how the conflict between private and public interests of the Bank was resolved in the period 1844 to 1870. Also see the discussion above on Bagehot (p. 46). 59. Not however without courting disaster on numerous occasions (Viner 1932, p. 16); also, Viner 1937, pp. 259-74. 60. Viner's (1937) chronicle of the history of nineteenth-century debates over the monetary standard centered on the role of discretionary management under a gold standard rule. "Although most present day writers seem to believe either that the non-automatic character of the modern gold standard is a discovery of the postwar period or that it was only in the postwar period that the gold standard lost its automatic character, currency controversy during the entire nineteenth century concerned itself largely with the problems resulting from the discretionary or management elements in the prevailing currency systems. The bullion controversy ... turned largely on the difference in the mode of operation in the international mechanism of a managed paper standard currency, on the one hand, and of a convertible paper currency, on the other, with the latter treated generally, but not universally, as if it were automatic. Later, the adherents of both the currency and the banking schools distinguished carefully between the way in which a supposedly automatic 'purely metallic' currency (which, in addition to specie, included bank deposits but not bank notes) would operate and the way in which the Bank of England was actually operating a 'mixed' currency (which, in addition to specie and bank deposits, included bank notes)" (pp. 388-89). 61. "This means that relative prices can be knocked about by the most fleeting influence of politics and of sentiment, and by the periodic pressure of seasonal trades. But it also means that the postwar method [of flexible exchange rates] is a most rapid and powerful corrective of real disequilibria in the balance of international payments arising from whatever causes, and a wonderful preventive in the way of countries which are inclined to spend abroad beyond their resources" (Keynes [1923] 1971, p. 130). 62. The analogy holds perfectly if we assume that each central bank has a rigid goldreserve ratio so that the aggregate quantity of central-bank money is determined by the aggregate gold reserves of the central banks; that no gold is used as currency; and hence that variations in the world monetary gold stock are determined by the difference between the amount of new gold mined and the amount consumed in the arts (Keynes [1930] 1971, p. 250). 63. This reflects two factors: a central bank cannot turn to a "lender of last resort" if its reserves are deficient; a central bank does not maximize profits and thus may keep higher reserves than otherwise (Keynes [1930] 1971, p. 252). 64. In the British banking system there is little interest-rate competition (Keynes [1930] 1971, p. 254). 65. J. H. Williams, "Monetary Stability and the Gold Standard" (1932) in Williams 1947 covered much the same ground. 66. Keynes argued that this dilemma did not present itself to Great Britain before 1914 because "the influence of London on credit conditions throughout the world was so predominant that the Bank of England could almost have claimed to be the conductor of the international orchestra. By modifying the terms on which she was prepared to lend, aided by her own readiness to vary the volume of her gold reserves and the unreadiness of other central banks to vary the volume of theirs, she could to a large extent determine the credit conditions prevailing elsewhere" (Keynes [1930] 1971, p. 274). However since World War I, the decline of Great Britain's influence on world credit conditions meant that she now faced the dilemma.

105

The Gold Standard: The Traditional Approach

67. See Sayers 1936 and p. 95 for a discussion of how this was frequently done on a de facto basis by the Bank of England in the 1890-1914 period. 68. As in the Tract, Keynes considered the case for a managed money standard with flexible rates but ultimately rejected it because the uncertainty associated with exchangerate fluctuations would impede long-term capital mobility. "If we ... desire that there should be a high degree of mobility for international lending, both for long and for short periods, then this is, admittedly, a strong argument for a fixed rate of exchange and a rigid international standard" ([1930] 1971, p. 299). Also, see Williams (1947), who recommended maintenance of the gold standard as a restraint, setting the "limits to which monetary variation can be carried" but widening the role of discretionary monetary policy (pp. 187-88). Ultimately "the logical end of the evolution of credit management, and the only real hope of solution of the conflict between external and internal stability, would be closer cooperation of central banks looking toward some form or degree of supernational management" (p. 190). 69. Whale cited evidence of rapid price adjustment between Lancashire and the rest of England, and between England and Scotland. 70. Ford argued that accounting for real-income changes would also explain Taussig's puzzle, "that periods of active lending have been characterized by rising prices rather than falling prices and that the export of goods apparently has taken place, not in conjunction with a cheapening of goods in the lending country, but in spite of the fact that its goods have seemed dearer at times of great capital export" (Taussig [1927] 1966, p. 219). 71. Ford does not entirely dismiss the role of price changes, but doubts that the elasticities are high enough or prices flexible enough to carry the full burden (1962, p. 12; 1977, p. 17). 72. Ford estimated the marginal propensity to import at about 0.3 and the marginal propensity to save at 0.1 to 0.2. This produces an open-economy multiplier of2 to 2.5 which is not sufficient to equilibrate the balance of payments (1962, p. 54). 73. Following Keynes ([1923] 1971) and Bloomfield (1959), Ford argued that the rules of the game were followed only by creditor countries, and not even by all of them (1962, p. 16). Britain's playing by the rules was facilitated by the location in London of the world's principal capital and gold markets (pp. 11-12). 74. This evidence is contrasted to Ford's finding, unfavorable to the classical relativeprice mechanism, that "the cyclical behavior of the net barter terms of trade shows no such consistent pattern" (1962, p. 76). 75. Favorable circumstances for Britain other than the institutional environment stressed by Ford were: confidence in the convertibility of sterling, the alternating pattern of trends in home and foreign investment, and the sensitivity of British exports to British overseas lending (1962, p. 190). 76. "The price level in the U.S. relative to that in Britain rose from 89.1 to 91.1" (Friedman and Schwartz 1963, p. 98). 77. However one conspicuous example, discussed by Friedman and Schwartz, when the classical approach could not fully explain the adjustment mechanism, was the period 1896-1901. 78. A study by Macesich (1960), using a similar approach, demonstrated that the monetary instability in the period 1834-45 was not caused primarily by the Bank war and Jacksonian policy, as traditionally believed, but rather was produced by external events. Given that the United States was part of the international specie standard, the author argued, internal prices had to adjust to external prices, and how they did so did not matter. Macesich isolated the different determinants of monetary change and found that changes in the ratio of the public's holdings of deposits-plus-notes to specie and the ratio of the bank's liabilities to specie explained most of the change in the money supply reflecting uncertainty engendered by the Bank war. The approach of Temin (1969), based on different data

106

Michael D. Bordo

sources, was similar to Macesich's, but attached greater importance to changes in highpowered money, and less to the ratios, in explaining monetary movements. 79. Following Keynes ([1923] 1971, [1930] 1971) and Cassel (1935), -Triffin doubted the strength of the long-term equilibrating forces of the commodity theory of money and termed the gold discoveries of the nineteenth century favorable accidents temporarily reversing a tendency to secular deflation. Instead, "the reconciliation of high rates of economic growth with exchange-rate and gold-price stability was made possible by the rapid growth and proper management of bank money, and could hardly have been achieved under the purely, or predominantly, metallic systems of money creation characteristic of the previous centuries" (Triffin 1964, p. 15). Triffin (1960) foresaw that a gold-exchange standard based on key currencies, such as that which dominated the interwar and the post-World War II periods, would ultimately fail because of a growing threat to the convertibility of the key currencies as their use as international reserves increased. 80. See similar approaches by Abramovitz (1973) and Thomas (1973). 81. In the postbellum period it was the foreign demand for wheat that was the key source of the long swing. Williamson's interpretation of the 1830s differs markedly from that of Macesich and Temin (n. 78 above) who each stressed the role of external monetary forces as the key disturbing factor. 82. However, the evidence to date on commodity arbitrage, based on more recent evidence, is far from conclusive, with the majority of studies casting doubt on its effectiveness for other than internationally traded commodities (see Kravis and Lipsey 1978). 83. "When two (or more) countries are on the gold standard then there exist definite limits for the absolute differences between their short-term interest rates. The actual differences at a given moment depend on the absolute stand of the exchange rates at the same moment, which in turn can vary only between the gold points of the currencies.... When the interest rates of two countries conform, then we say that their money markets are in a state of solidarity; when the differentials do not conform with the respective absolute positions of the exchange rates, i.e., when they exceed the respective "permissible limits," then we say that they violate that solidarity ... the principles of the gold standard" (Morgenstern 1959, pp. 166-68). 84. Morgenstern explicitly rejected the available data on gold movements because his study (1955) found them to be unreliable. In that study, using official data of imports and exports of gold coin and bullion for the United States, Great Britain, Germany, France, and Canada and a sample of four years, 1900, 1907, 1928, and 1935, official gold exports from one country did not square with gold imports for another. 85. "Without examining the future exchange prices, the author cannot make a case that the maximum permissible interest rate differentials were in fact violated. For the market will respond to the best opportunity. If New York interest went to a 1% premium over London, Morgenstern would say the market was not operating perfectly. Yet with a premium on spot exchange, it would be perfectly consistent with the operations of a competitive market. The question of interest differentials which exceed the maximum exchange risk then involves the interest parities of forward exchange rates. Morgenstern did not examine this at all" (Borts 1964, p. 227). 86. However, a similar test for Canada revealed no correlation (Bloomfield 1963, p. 65). 87. For Canada and Sweden, merchandise exports in real terms also showed long swings that tended to lead those in other variables. 88. Scammel (1965) echoed this view by stating "it is, in the writer's view, arguable that the gold standard was in fact quasi-organizational, being operated by a team of central bankers cooperating under the leadership of the Bank of England on behalf of the world business community" (p. 34). 89. Foreign-exchange reserves accounted for 19 percent of total world reserves in 1913. Japan, Russia, and India held the largest fraction of their reserves in sterling.

107

The Gold Standard: The Traditional Approach

90. Of four effects of a rise in discount rate to correct the balance of payments, Lindert ruled out the effect on aggregate demand as involving too lengthy a lag to account "for the remarkable smoothness and rapidity with which exchange rates, international gold flows, and gold reserves of central banks seem to have been altered." Moreover he found the lag between changes in Bank rate and import prices also to be too long to be relevant, and the evidence on the effect on new issues of long-term foreign securities to be unclear. Hence only the effect on short-term funds operated as the key channel of influence of Bank rate (1969, pp. 43-47). 91. As a test of the relative "pulling" power of the different currencies, Lindert (1969, p. 50) regressed each exchange rate on the other two exchange rates, on its own discount rate, and the second center's rate. He found a hierarchy of dominance running from London to Paris to Berlin. 92. Quotation from the Macmillan report (1931, p. 125) in Lindert 1969, p. 49. The fact that London and the other centers could maintain "deficits without tears" in the pre-World War I period ultimately led to a weakening of the balance-of-payments adjustment mechanism because the longer the process continued, the more difficult it became "to undertake the contractionary measures that would have been required to restore payments 'equilibrium' " (Lindert 1969, p. 79). Also see Triffin 1960. 93. Additional methods used were the outright sale or purchase of securities, selling consols spot and buying for the account, borrowing in the market, borrowing from clearing banks, borrowing from special depositors, and moral suasion. After 1878, the Bank lent to its own customers at the market rate of interest, while at the same time it charged discount houses a penalty rate above the market rate. "The position of the penal rate was ordinarily a matter of daily concern and therefore influential over the market rate itself" (Sayers 1951, p. 115). 94. Goodhart's analysis follows closely that of Whale (1937). Both Goodhart's and Whale's results can be reinterpreted as consistent with the monetary approach to the balance of payments. According to that approach, a rise in economic activity in an open economy such as Great Britain would generate an excess demand for money that would be satisfied in part by a gold inflow. Indeed Mills and Wood (1978), applying the Granger-Sims causality test to the pre-World War I U.K. money supply and national-income data, found that income caused money, evidence that they considered sympathetic to the monetary approach. 95. Also see A. G. Ford (1977, p. 42), who cites similar evidence on investment activity.

References Abramovitz, Moses. 1973. The monetary side of long swings in U.S. economic growth. Memorandum no. 146, Stanford University Center for Research on Economic Growth. Mimeo. Andrew, A. P. 1907. The Treasury and the banks under Secretary Shaw. Quarterly Journal of Economics 21 (Aug.): 519-68. Angell, J. W. [1925] 1965. The theory of international prices. Reprint. New York: Augustus M. Kelley. Bagehot, w. [1873] 1969. Lombard Street. Reprint of the 1915 edition. New York: Arno Press.

108

Michael D. Bordo

Beach, W. 1935. British international gold movements and banking policy, 1881-1913. Cambridge: Harvard University Press. Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard. New York: Federal Reserve Bank of New York. - - - . 1963. Short-term capital movements under the pre-1914 gold standard. Princeton Studies in International Finance, no. 11. Princeton: Princeton University Press. - - - . 1968. Patterns of fluctuation in international investment before 1914. Princeton Studies in International Finance, no. 21. Princeton: Princeton University Press. Bordo, M. D. 1975. John E. Cairnes on the effects of the Australian gold discoveries, 1851-73: An early application of the methodology of positive economics. History of Political Economy 7 (no. 3): 337-59. - - - . 1983. Some aspects of the monetary economics of Richard Cantillon. Journal of Monetary Economics 12 (Aug.): 235-58. Borts, G. H. 1964. Review of International financial transactions and business cycles, by O. Morgenstern. Journal ofthe American Statistical Association 59 (Mar.): 223-28. Brown, William A. 1940. The international gold standard reinterpreted, 1914-1934. New York: National Bureau of Economic Research. Bullion report. 1810. See Report . .. on the high price of bullion [1810] 1978. Cairncross, Alec K. 1953. Home and foreign investment, 1870-1913. Cambridge: Cambridge University Press. Cairnes, J. E. [1873] 1965. Essays in political economy: Theoretical and applied. Reprint. New York: Augustus M. Kelley. Cantillon, R. [1931] 1964. Essai sur la nature du commerce en general. Ed. H. Higgs. Reprint. New York: Augustus M. Kelley. (First published 1755.) Cassel, Gustav. 1935. The downfall of the gold standard. Oxford: Clarendon Press. Chevalier, M. 1859. On the probable fall in the value of gold! The commercial and social consequences which may ensue, and the measures which it invites. Translated from the French by Richard Cobden. 3d ed. Manchester: A. Ireland. Cunliffe report. [1918] 1979. See United Kingdom, Parliament [1918] 1979. Dick, Trevor. 1981. Canadian balance of payments, 1896-1913: Mechanisms of adjustment. Mimeo. Fetter, Frank W. 1965. The development of British monetary orthodoxy, 1717-1875. Cambridge: Harvard University Press. - - - . 1953. The Bullion Report re-examined. In Papers in English monetary history, ed. T. S. Ashton and R. S. Sayers. Oxford: Clarendon Press.

109

The Gold Standard: The Traditional Approach

Fisher, I. 1920. Stabilizing the dollar. New York: Macmillan. - - . [1922] 1965. The purchasing power of money. Reprint. New York: Augustus M. Kelley. - - - . 1935. Are booms and depressions transmitted internationally through monetary standards? Bulletin of the International Statistical Institute 28 (no. 1): 1-29. Ford, A. G. 1962. The gold standard, 1880-1914: Britain and Argentina. Oxford: Clarendon Press. - - - . 1977. International financial policy and the gold standard, 18701914. Mimeo. Frenkel, J. 1971. A theory of money, trade, and the balance of payments in a model of accumulation. Journal ofInternational Economics (May): 159-87. Friedman, M., and A. J. Schwartz. 1963. A monetary history of the United States, 1867-1960. Princeton: Princeton University Press. Girton, L., and D. Roper. 1978. J. Laurence Laughlin and the quantity theory of money. Journal of Political Economy 86 (Aug.): 599-625. Goodhart, C. A. E. 1972. The business ofbanking, 1891-1914. London: Weidenfield and Nicolson. Goschen, G. J. [1892] 1978. The theory of the foreign exchanges. Reprint. New York: Arno Press. Graham, F. D. 1922. International trade under depreciated paper: The United States, 1862-79. Quarterly Journal of Economics 36 (Feb.): 220-73. Gregory, T. E. [1932] 1979. The gold standard and its future. Reprint. New York: Arno Press. Hawtrey, R. G. 1935. The gold standard in theory and practice. 5th ed. London: Longmans Green. Hume, D. [1752] 1955. Of the balance of trade. Reprint. In Writings on economics, ed. E. Rotwein. Madison: University of Wisconsin Press. Jevons, W. S. [1884] 1964. Investigations in currency and finance. Reprint. New York: Augustus M. Kelley. Johnson, H. G. 1976. The monetary approach to balance of payments theory. In The monetary approach to the balance of payments, ed. J. Frenkel and H. G. Johnson. Toronto: University of Toronto Press. Jonung, Lars. 1979. Knut Wicksell and Gustav Cassel on secular movements in prices. Journal ofMoney, Credit, and Banking (May): 165-81. Keynes, J. M. [1913] 1971. Indian currency and finance. Vol. 1 of The collected writings ofJohn Maynard Keynes. Reprint. London: Macmillan and New York: Cambridge University Press for the Royal Economic Society. - - . [1923] 1971. A tract on monetary reform. Vol. 4 of The collected writings. See Keynes [1913] 1971.

110

Michael D. Bordo

- - . [1930] 1971. The applied theory of money: A treatise on money. Vol. 6 of The collected writings. See Keynes [1913] 1971. Kindahl, J. K. 1961. Economic factors in specie resumption. Journal of Political Economy 69 (Feb.): 30-48. Kravis, Irving B., and Robert E. Lipsey. 1978. Price behavior in the light of balance of payments theories. Journal of International Economics 8 (May): 193-246. Kuhn, T. 1970. The structure of scientific revolutions. 2d ed. Chicago: University of Chicago Press. Laidler, D. 1981. Adam Smith as a monetary economist. Canadian Journal of Economics 14 (May): 185-200. Laughlin, J. L. 1903. The principles of money. New York: Scribner's. League of Nations. [1931] 1979. Reports of the gold delegation: Interim report of the gold delegation of the financial committee. Reprint. New York: Arno Press. Lindert, Peter H. 1969. Key currencies and gold, 1900-1913. Princeton Studies in International Finance, no. 24. Princeton: Princeton University Press. Macesich, G. 1960. Sources of monetary disturbances in the U.S., 183445. Journal of Economic History 20 (Sept.): 407-34. Macmillan report. 1931. See United Kingdom, Parliament 1931. Marshall, Alfred. 1923. Money, credit, and commerce. London: Macmillan. - - - . 1926. Official papers. London: Macmillan. McCloskey, D. N., and J. R. Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance of payments. See Johnson 1976. Meier, Gerald M. 1953. Economic development and the transfer mechanism: Canada, 1895-1913. Canadian Journal of Economics and Political Science 19 (Feb.): 1-19. Mill, J. S. [1865] 1961. Principles of political economy. Reprint. New York: Augustus M. Kelley. Mills, T. C., and C. E. Wood. 1978. Money-income relationships and the exchange rate regime. Federal Reserve Bank ofSt. Louis Review 60 (Aug.): 22-27. Mints, L. 1945. A history of banking theory in Great Britain and the United States. Chicago: University of Chicago Press. Morgenstern,Oskar. 1955. The validity of international gold movement statistics. Special Papers in International Economics, no. 2. Princeton: Princeton University Press. - - - . 1959. International financial transactions and business cycles. Princeton. Princeton University Press. Mundell, R. 1971. Monetary theory. Pacific Palisades, Calif.: Goodyear.

111

The Gold Standard: The Traditional Approach

Nurkse, R. [1944] 1978. International currency experience. Reprint. New York: Arno Press. (First published by the League of Nations.) O'Leary, P. J., and W. Arthur Lewis. 1955. Secular swings in production and trade, 1870-1914. The Manchester School 23 (May): 118-25. Palyi, Malchior. 1972. The twilight of gold, 1914-1936: Myths and realities. Chicago: Henry Regnery. Pesmazoglu, J. S. 1951. A note on the cyclical fluctuations of British home investment, 1870-1913. Oxford Economic Papers 3 (Feb.): 3961. Presnell, L. S. 1968. Gold reserves, banking reserves, and the Baring crisis of 1890. In Essays in money and banking in honour ofR. s. Sayers, ed. c. R. Whittlesey and J. S. C. Wilson. Oxford: Clarendon Press. Report from the select committee on the high price ofbullion. [1810] 1978. New York: Arno Press. Ricardo, D. [1811] 1951. High price of bullion: A proof of the depreciation of bank notes. In The works and correspondence of David Ricardo, ed. Piero Sraffa, 3. Cambridge: Cambridge University Press. - - - . [1816]. 1951. Proposals for an economical and secure currency; with observations on the profits of the Bank of England as they regard the public and the proprietors of Bank stock. In The works and correspondence of David Ricardo. See Ricardo [1811] 1951. Rich, G. 1978. The cross of gold: Money and the Canadian business cycle, 1867-1913. Mimeo. Samuelson, P. 1971. An exact Hume-Ricardo-Marshall model of international trade. Journal of International Economics 1 (Feb.): 1-11. Sayers, R. S. 1936. Bank of England operations, 1890-1914. London: P. S. King and Son. - - - . 1957. Central banking after Bagehot. Oxford: Clarendon Press. - - - . 1951. The development of central banking after Bagehot. Economic History Review, 2d ser., 4 (no. 1): 109-16. - - - . 1953. Ricardo's views on monetary questions. In Papers in English monetary history. See Fetter 1953. Scammel, W. M. 1965. The working of the gold standard. Yorkshire Bulletin of Economic and Social Research 17 (May): 32-45. Silverman, A. G. 1931. Some international trade factors for Great Britain, 1880-1913. Review of Economics and Statistics 13 (Aug.): 114-24. Smit, J. C. 1934. The pre-war gold standard. Proceedings ofAcademy of Political Science 13 (Apr.): 53-61. Smith, A. [1776] 1976. An inquiry into the nature and causes ofthe wealth of nations. Reprint. Chicago: University of Chicago Press. Taussig, F. W. 1917. International trade under depreciated paper, a contribution to theory. Quarterly Journal of Economics 21 (May): 380-403.

112

Michael D. Bordo

Taussig, F. W. [1927] 1966. International trade. New York: Augustus M. Kelley. Temin, P. 1969. The Jacksonian economy. New York: W. W. Norton. Thomas, B. 1973. Migration and economic growth. 2d edt Cambridge: Cambridge University Press. Thornton, H. [1802] 1978. An inquiry into the nature and effects of the paper credit of Great Britain. Fairfield, N.J.: Augustus M. Kelley. Tinbergen, J. 1950. Business cycles in the United Kingdom, 1870-1914. 2d edt Amsterdam: North-Holland. Triffin, Robert. 1960. Gold and the dollar crisis. New Haven: Yale University Press. - - - . 1964. The evolution ofthe international monetary system: Historical reappraisal and future perspectives. Princeton Studies in International Finance, no. 12. Princeton: Princeton University Press. United Kingdom. Parliament. [1918] 1979. First interim report of the committee on currency and foreign exchanges after the war. Cmnd. 9182. Reprint. New York: Arno Press. - - - . 1931. Report of the committee on finance and industry. (Macmillan report). Cmnd. 3897. London: HSMO. Viner, Jacob. 1924. Canada's balance of international indebtedness, 1900-1913. Cambridge: Harvard University Press. - - - . [1937] 1975. Studies in the theory ofinternational trade. Reprint. New York: Augustus M. Kelley. - - - . 1932. International aspects of the gold standard. In Gold and monetary stabilization, edt Q. Wright. Chicago: University of Chicago Press. Whale, P. Barrett. 1937. The working of the pre-war gold standard. Economica 4 (Feb.): 18-32. White, H. D. 1933. The French international accounts, 1880-1913. Cambridge: Harvard University Press. White, L. H. 1981. Free banking in Britain: Theory, experience, and debate, 1900-1945. Ph.D. diss., University of California at Los Angeles. Wicksell, Knut. [1898] 1965. Interest and prices. Reprint. New York: Augustus M. Kelley. Williams, D. 1968. The evolution of the sterling system. In Essays in money and banking in honour of R. S. Sayers. See Presnell 1968. Williams, John. 1920. Argentine international trade under inconvertible paper money, 1800-1913. Cambridge: Harvard University Press. - - - . 1947. Gold and monetary stabilization. In Postwar monetary plans and other essays. New York: Knopf. Williamson, Jeffrey G. 1964. American growth and the balance of payments, 1820-1913. Chapel Hill: University of North Carolina Press.

113

The Gold Standard: The Traditional Approach

- - - . 1961. International trade and u.s. economic development 1827-1843. Journal of Economic History 21 (Sept.): 372-83. - - - . 1963. Real growth, monetary disturbances, and the transfer process: The U.S., 1879-1900. Southern Economic Journal 29 (Jan.): 167-80.

Comment

C. Knick Harley

Michael Bordo has certainly presented an extensive review of some two centuries of thought on the workings of a currency either consisting of specie or based on notes freely convertible into specie. I have learned a great deal about the evolution of ideas from the 160 pages and 446 footnotes in the version of the paper I have worked with. Certainly I do not have the qualifications, the time, nor the inclination to discus individual economists or even individual schools of thought in comparable detail. Rather I would like to highlight some major themes that run through the literature. In particular, I would like to try to shift the emphasis of the discussion toward the last fifty or a hundred years and stress what Viner called "the international mechanism" partially in its theoretical context but especially within the context of the process of foreign lending in the late nineteenth century. Two issues have always dominated discussion of the gold standard. The first considers the determinants of the value of the standard in terms of other commodities and the stability of that value over time. The second considers the nature of international equilibrium and the relationships among the prices of various commodities in various locations or countries that equilibrium requires. Central to this discussion is the comparison of equilibria under differing underlying conditions and also investigation of dynamics of adjustment when underlying conditions alter. I intend to add nothing to the discussion of the value of the standard, but will state the obvious: The fluctuations of the value of the standard over the last two-thirds of a century make the concerns of Marshall, Fisher, and their contemporaries seem rather trivial. To my mind the most interesting literature relating to the gold standard deals with what can broadly be called, after Viner, "the international mechanism." Unfortunately, I am disappointed by Michael Bordo's treatment of the literature that has grown up over the past half century. In particular I am surprised that so little attention has been paid to Viner's C. Knick Harley is professor of economics at the University of Western Ontario, London, Ontario, Canada.

114

Michael D. Bordo

Studies in the Theory ofInternational Trade (six footnote references, most apparently as an afterthought) which has always seemed to me to have been the most thorough and considered statement of the traditional position as it emerged between the wars. Along the same lines I miss any discussion of the theoretical literature on transfers and reparations (Keynes [1929] 1949; Ohlin [1929] 1949; Samuelson 1952,1954). Finally, more should be said about the literature on the "Atlantic economy" that has appeared since the Second World War. I would, therefore, like to use the few minutes at my disposal to present what I see to have been the issues raised by these strands of literature and thus highlight what I believe to be major issues that should remain in our research agenda. The literature on the "international mechanism" contained two strands to which I have already referred. First, there is the real theory of comparative-static comparison of international equilibria. Second, there is discussion of what we would now call macroeconomic processes involved in moving between equilibria. Now these two issues were often not clearly separated in the literature and Ohlin ([1929] 1949, p. 179, n. 3) was moved to remark (and was echoed by Samuelson), "Professor Taussig seems to me to present two different and incompatible theories: (1) the barter theory of Mill; (2) a theory of the monetary and price mechanism." Ohlin, of course, proceeded to explain the real (barter) theory in a manner that was quickly recognized to be correct. At the same time he left the "theory of the monetary and price mechanism" to "Professor Viner, with whom I am in substantial agreement." Ohlin demonstrated that the older supposition, which underlay much of the work by Taussig and his followers, that the real transfer implied a shift in terms of trade against the paying country, was incorrect. The comparative statics of the transfer problem under a considerable range of conditions is now firmly understood and is presented with his characteristic lucidity by Samuelson (1952, 1954). Even if errors in identifying the conditions of new equilibrium are avoided, issues of the (macroeconomic) adjustment to the new equilibrium in a complex monetary economy remain. First, the transfer mechanism involves a reduction of "total buying power" in the sending country and an increase in the receiving country. In addition new equilibrium will normally require relative price adjustments, although as Ohlin demonstrated, the adjustment in the relative prices of traded goods is smaller and less predictable than earlier writers imagined. Furthermore, in a world where transportation costs result in a considerable portion of national product being untraded, resources wil have to be redirected toward (or away) from traded goods in the sending (receiving) economy, presumably through the mechanism of relative price changes. The final equilibrium will require a lower money stock in the sending country and a higher stock in the receiver. Discussion of the monetary aspects of the

115

The Gold Standard: The Traditional Approach

adjustment mechanism consists of attempts to discover the processes that are involved in these adjustments. Furthermore, most investigators seem to believe that alteration of relative prices is not frictionless and can be expected to affect both employment and the general price level during a period of adjustment. Thus Keynes ([1929] 1949, p. 167) characterized those who see no such adjustment prlblems as "applying the theory of liquids to what is, if not a solid, at least a sticky mass with strong internal resistance." Discussion of the operation of the gold standard in the short run must now proceed among the other battles in the arena of short-term macroeconomic adjustment where heroes of flexibility confront gladiators of rigidity. Historical fact-as well perhaps as historical myth-maintains scholars' interest in the late-nineteenth-century gold standard. The international economy seems to have experienced less short-run instability than either theoretical models or twentieth-century experience would suggest. Britain, in particular, seems to have succeeded in maintaining enviable short-run stability despite recurring real shocks in the form of large periodic fluctuations in foreign investments that reflected the cycles in expansion in areas of recent settlement. For example, in 1903 foreign investment was about 2.5 percent of British GNP, but by 1913 it had grown to some 7 percent. Now the remarkable feature of these cycles to Taussig and his students when they discovered them and to subsequent investigators who believe in "sticky mass" theories of short-term adjustment was both the near absence of fluctuations in unemployment and inflation and the absence of monetary crises despite Britain's fractional reserve banking system. What happened factually is clear enough. First, the balance of payments adjusted rapidly, primarily through fluctuations in exports, so that an increase in foreign lending from about £50 million in 1903 to well over £200 million in 1913 was transferred abroad without significant gold flows and thus without placing any noticeable strain on the Bank of England's gold holdings that were always well under £40 million. Thus loans were transferred without monetary strain. Second, as Cairncross (1953) demonstrated, fluctuations in employment and the domestic price level were largely avoided because domestic investment, particularly building, declined as exports rose in response to foreign investment and then displaced exports as the latter declined when foreign investment declined. Now it is obvious to anyone who examines the evidence that the international system could not have worked in the manner postulated by the simplest version of the gold standard adjustment mechanism. That is to say, the increase in lending could not possibly have been initially transferred in gold and the real transfer then been effected by gradual price adjustments. In the first place there was insufficient gold in Britain, and in the second the adjustment seems too rapid.

116

Michael D. Bordo

The historical literature, it seems to me, is best seen as trying to understand this adjustment. One strand has paid attention to short-run capital flows and the use of British balances as a substitute for gold as the mechanism that allowed fluctuating capital flows in the face of very moderate gold movements. That explanation, however, does not seem to offer any strong mechanism acting in Britain to effect the real transfer, nor does it help to explain how sterling in its role as a reserve currency avoided the problems of sterling since the First World War or of the dollar since 1960. A second major strand in the literature has maintained that the capital movements were not exogenous shocks but part of a wider process of growth in the Atlantic economy as a whole. The working out of that process of growth involved simultaneously changes in the composition of production in the borrowing and lending country necessary to effect the real transfer and the capital flow in financial terms. In this tradition I would obviously include Brinley Thomas's and Jeffrey Williamson's work. I would also draw attention to a very interesting theoretical article by George Borts (1964). Others like A. G. Ford argue that while long-run forces in the expanding international economy aided adjustment, it is probably somewhat misleading to focus exclusively on Britain's stability. Their argument would run that there were inflationary periods alternating with financial crises and deflation, but the international market for short-term funds insulated Britain and forced the instability onto the periphery. I remain agnostic about the mechanism that smoothly allocated resources alternately between building houses in Oldham and Oklahoma over at least a half century. Until we understand that mechanism, we must take care in drawing conclusions about the classical gold standard. Perhaps, as H. J. Habakkuk (1962) has argued, there was not a mechanism but rather chance. It seems more likely to me that there were elements of the systematic relationships suggested in the literature, but these may well have been peculiar to their own historical situation. It certainly seems that the international transfer of capital within the late nineteenth century has been and should remain central to our study and understanding of the workings of the gold standard. More central, I would argue, than Michael Bordo's summary allows. Finally, I cannot concur with Bordo's vision of McCloskey and Zecher's (1976) article as a Copernican revolution in our thinking about the gold standard. Certainly they have made some useful arguments about what sorts of disequilibria are possible. But they have not considered what is to my mind the major issue of the adjustment mechanism in the late-nineteenth-century gold standard that I have just discussed at length. Finally, I have been unable to resist reading a few sentences of Viner (1937, pp. 316--17) to them:

117

The Gold Standard: The Traditional Approach

The classical school and its important followers all held the same views on this point: after allowance for transportation costs, the market prices of identical transportable commodities must everywhere be equal or tend to be equal when expressed in or converted to a common currency. When, therefore, critics of the classical theory have taken it to task on the grounds that it explained the adjustment of international . balances by the influence on the course of trade of divergent market prices in different markets of identical transportable commodities. . . they have misinterpreted the classical doctrine. References Borts, George H. 1964. A theory of long-run international capital movements. Journal 0/ Political Economy 72 (Aug.): 341-59. Habakkuk, H. J. 1962. Fluctuations in house-building in Britain and the United States in the nineteenth century. Journal 0/ Economic History 22 (June): 198-239. Keynes, J. M. [1929] 1949. The German transfer problem. Reprint. In Readings in the theory o/international trade, ed. H. S. Ellis and L. S. Metzler. Philadelphia: Blakiston. (First published in Economic Journal39 [Mar.]: 1-7.) McCloskey, D. N., and j. R. Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance o/payments, ed. J. Frenkel and H. G. Johnson. Toronto: University of Toronto Press. Ohlin, B. [1929] 1949. Transfer difficulties, real and imagined. Reprint. In Readings in the theory o/international trade, ed. H. S. Ellis and L. S. Metzler. Philadelphia: Blakiston. (First published as The reparation problem: A discussion, in Economic Journal 39 [June]: 172-78.) Samuelson, Paul A. 1952. The transfer problem and transport costs: The terms of trade when impediments are absent. Economic Journal 62 (June): 278-304. - - - . 1954. The transfer problem and transport costs II: Analysis of effects of trade impediments. Economic Journal 64 (June): 264--89. Viner, Jacob. [1937] 1975. Studies in the theory 0/ international trade. Reprint. New York: August M. Kelley.

General Discussion ABRAMOVITZ emphasized and endorsed the view that a central mystery concerning the adjustment mechanism under the international gold standard lies in reconciling the relative freedom enjoyed by Great Britain from the long swings in· aggregate economic activity that characterized

118

Michael D. Bordo

the United States with the magnitude of international capital movements in the nineteenth century. FRIEDMAN elaborated upon the issue raised by Abramovitz by referring to the findings of a study that Anna Schwartz and he recently completed on monetary trends in the United States and the United Kingdom. One of their most surprising findings was the independence of phase-cycle-tophase-cycle movements in real income in Great Britain from events in the United States. Friedman and Schwartz had been unable to find any systematic factors explaining real-income fluctuations in Britain. Friedman suggested that this phenomenon was precisely the one cited by Abramovitz. MCCLOSKEY responded to a point raised by Harley. He noted in several of the papers at the conference a definition of equilibrium in the international affairs of nations that is both artificial and a potential source of confusion. According to that definition, a country's international affairs are in equilibrium only when the balance of payments is zero or, in a more extreme version, when the current-account balance is zero. In fact, countries can maintain surpluses or deficits in their balance of payments for periods of decades or longer. This is evident in the recent experience of Japan, for example. Similarly, countries sometimes go for centuries with surpluses or deficits in their trade balances. Thus, a zero trade balance or zero current-account balance may not be a useful definition of equilibrium. ZECHER, commenting on David Hume's methodology, pointed out that though Hume may have believed in the law of one price as always in effect, yet he might still ask what would happen if prices did differ between countries or regions. For exan:tple, he used this method in explaining the law of gravity by starting with a situation where water did not find its own level. KOCHIN noted Bordo's exclusion of a notable authority on the gold standard, Adam Smith. He noted also another issue omitted from Bordo's paper, namely, Hume's recognition that one of the purposes of the gold standard is to possess a treasure against the contingency of war. FRENKEL, commenting on Harley's remarks, was skeptical that one should require a discussion of the gold standard to bring to the forefront the terms of the trade adjustment consequent on the transfer. Frenkel suggested that as far as the gold standard is concerned, the terms-of-trade effect is probably not a central issue, since much of the adjustment mechanism operates through the income-expenditure mechanism a la Ohlin. Frenkel also asked discussants to define precisely what they meant when referring to the law of one price. In his own discussion of the law of one price, Hume had stated merely that any man who traveled in Europe in his day observed that differences in the prices of commodities between

119

The Gold Standard: The Traditional Approach

one kingdom and another are no greater than such differences between provinces in the same kingdom. DORNBUSCH expressed surprise at Frenkel's statement that the transfer issue is not central to the analysis of the gold standard. It is central, he suggested, if we think that relative price changes are part of the gold standard adjustment mechanism. WHITE offered a supplement to Bordo's discussion of the classical period in which Bordo considers Ricardo and Mill but omits the currency-banking school controversy. White suggested that Bordo's account of the Act of 1844 relies too heavily on the arguments of its advocates. It may be misleading to say that the currency school wished to combine the automaticity of the gold standard with the capital savings of a fiduciary currency, since in fact the Act of 1844 was not required in order to accomplish that goal. Simply permitting the existence of an unregulated financial system with fractional reserve banking would have been sufficient to achieve that goal. White argued that the currency school's principal complaint was that the gold standard was insufficiently automatic; they sought to eliminate what they saw as a slippage between exports of gold and reductions of the quantity of money. White went on to argue that the so-called banking school was made up of two distinct sets of opponents of the Act of 1844. Bordo indicates that the opponents of the act favored free competition, but in fact certain members of the banking school, such as Thomas Tooke, were quite hostile to free competition in banking and sympathetic with the Bank of England's monopoly. At the same time, there were other opponents of the Bank Act, best referred to as the free banking school, who advocated free competition in banking. The free banking school had a monetary theory of the trade cycle and blamed the Bank of England for the economy's cyclical instability. Tooke, on the other hand, had a nonmonetary theory and placed little of the blame on the Bank of England. BORDO concurred with Frenkel's view that the transfer problem was not central to the gold standard story. He accepted White's interpretation of the existence of divisions within the banking school.

2

The Success of Purchasing Power Parity: Historical Evidence and Its Implications for Macroeconomics Donald N. McCloskey and J. Richard Zecher

2.1

Two Views of the International Economy and Why They Matter

A model of the economic life of nations that emphasizes the mutual interdependence of the nations is easy to believe in a period of relative tranquility, such as the heyday of the gold standard, 1880-1914, or of the gold-dollar standard, 1945-71. In less tranquil periods, such as the thirty years of war and depression from 1914 to 1945 or the decade just past, one might suppose that the history is less favorable to the model. We propose to show that this supposition is misleading, and that interdependence was strong. The strength of interdependence depends on the strength of purchasing-power parity. And purchasing-power parity is stronger than it looks. Purchasing-power parity has recently been much in the scholarly news. Some of the new interest in an old idea is attributable to the recent turbulence of international finances, giving practical reasons for wanting Donald N. McCloskey is chairman of the department of economics and professor of history at the University of Iowa, Iowa City. J. Richard Zecher is senior vice-president and chief economist of the Chase Manhattan Bank, New York, New York. The authors are grateful to Francisco Comprido, Daniel Vesper, and Benjamin Russo for valuable research assistance. Part of section 2.4 draws on a paper in preparation by McCloskey and John Lewis, "The Anomalous Price Rise of 1933-34." Earlier versions have been presented at the Macroeconomics Seminar at the University of Iowa, the Economic History Seminar at Northwestern University, the Southern Economics Meetings in New Orleans, a seminar at the University of California at Davis, a conference on the gold standard at the University of Southern California, and seminars at Ohio State University and the Chase Manhattan Economics Group. The participants in these various sessions were tough but open-minded in their skepticism, for which thanks especially to Fischer Black, Eric Gustafson, Steven Kohlagen, Peter Lindert, Robert E. Lipsey, Joseph Vinso, Bob Slighton, Bill Dewald, Sykes Wilford, Blu Putnam, and Dayle Nattress.

121

122

Donald N. McCloskey and J. Richard Zecher

to know when exchange rates or prices are in equilibrium. But some interest is a result of autonomous intellectual change by itself. The chief of these changes was the monetary approach to the balance of payments developed in the early 1970s by Robert Mundell and his students, with involvement by Harry Johnson, Ronald McKinnon, Jeffrey Williamson, and various others. The monetary approach can do without the law of one price. 1 One can approach the balance of payments as a monetary phenomenon-that is to say, not primarily the real phenomenon that the elasticities approach believes it to be-without committing oneself to any particular view of the working of international arbitrage. The monetary approach, after all, merely notes that the balance of payments is by definition a balance on monetary account (just as the current balance is a balance on commodity account and the capital account is a balance on an account of future claims), then makes the innocent-sounding suggestion that its explanation might focus on the excess supply of and demand for money. It is not essential to adopt purchasing-power parity to believe that there may be merit in this view. Nonetheless, purchasing-power parity was in fact commonly invoked in the early theorizing in the monetary approach. Those present at the creation in Chicago in the late 1960s and early 1970s felt they were merely appropriating for use in the study of monetary affairs an assumption that was a necessary commonplace in the study of real affairs. It is hard to see how the real theory of international trade could have gotten far without postulating enough rationality on the part of economic actors to arbitrage away for each commodity any price differences outside the gold pointsthe term "gold points" defined to include all the risks and other costs of transportation. And, to go further, if the gold points were as wide for many commodities as is often implied in criticisms of purchasing-power parity, it is hard to see what usefulness there could have been in the standard propositions in the real theory, such as a tendency to factor price equalization or a tendency to satisfy the Heckscher-Ohlin theory of exports or indeed any tendency to equilibrium. The pioneers of the monetary approach felt they were simply bringing to international finance the intellectual habits formed in the study of the real theory, especially the intellectual habit of supposing that people exploit opportunities for profit. It is no accident that the second generation of leaders in bringing rigor to the study of international trade-the Mundells and Johnsons, following on the first generation of Samuelsons and Meadeswere the inventors of the monetary approach and were inveterate users of the assumption of purchasing-power parity.l We believe that the doctrine of purchasing-power parity, brought into theorizing about the monetary approach by the back door, should be the guest of honor. It is a more radical proposition than the one that the

123

The Success of Purchasing-Power Parity

nation's liquidity balance (or, still more generally, the asset balance) probably has much to do with the amount of money it imports or exports. It challenges in fact the whole way of doing macroeconomics. The usual way of doing macroeconomics might be called the Martian approach. A national economy, usually the United States, is taken in its relations with the rest of the world to be similar to Mars. The price level on Mars, obviously, is determined by Martian demand and supply curves (whether for money or for aggregate goods is not important); likewise, the interest rate. An occasional spaceship might land from Earth bearing gold or Federal Reserve notes, thereby driving up the price level on Mars (or increasing speculative balances in the presence of a liquidity trapagain, the rest of one's economic ideology is irrelevant to the point at issue). The arrival of the spaceship might even be occasioned by events on Mars, but only in a very long run, since it is a long way from Earth to Mars. Mars is a closed economy that has to adjust to the money supply or aggregate demand or expectations that Martians have, period. The Martian approach characterizes 90 percent of the articles and books on macroeconomics, written mostly by Americans. The theoreticians among them can always argue that it is unimportant to them whether or not any actual economy matches their models, for they are concerned with higher things. So much the worse for theory, one might say. A floor or two down the ivory tower the more empirically concerned theorists can argue, quite correctly, that their models might well apply to the whole world even if they are inappropriate for one part of the world. It is strange then to include institutions (such as a central bank with a national policy) in the models that have no worldwide equivalents, one might say. Empiricists, living with the computer down in the basement, can and do argue wearily that they are working on a still larger model (with 10,001 sectors) and will perhaps be able to fit the international sector into one of these. The rest of the world is to them merely another, rather small, sector of the American economy, similar in importance, say, to office equipment. They view an appeal to include the international milieu as a tiresome request to further complicate an already complex model by inserting an office-equipment sector. At the most they are willing to consider project LINK, with its plan to cure the maladies of misspecified national models by putting all the models into the same hospital. It is notable where the other 10 percent of the books and articles on macroeconomics originate. They originate from small open economies. Cassel was Swedish. His countryman, Knut Wicksell (1918), had no difficulty believing purchasing-power parity, debating with the American Taussig (1918) on the matter.3 The assumption of thoroughgoing arbitrage between regions and countries pervades the work of Heckscher and Ohlin. Canadians were the heirs to the Swedes in producing dispro-

124

Donald N. McCloskey and J. Richard Zecher

portionate numbers of excellent economists. They too, living with the great bear of the United States as the Swedes lived with the great bear of the German Empire, have found it difficult to think in Martian terms about their homeland, and even about their new American home. Johnson, McKinnon, and Mundell were raised in Canada, as were many of their students. Martian thinkers are accustomed to dismissing such cases with the remark, "precisely: they came from small economies; the United States is large." The remark is irrelevant, reflecting a common notion that recent developments in balance-of-payments theory depend somehow on the assumption that we are dealing with small economies. The psychological disposition to recognize the existence of an international milieu may be smaller in a big country, but the milieu is still there. The United States may be so big, to be sure, that it can significantly alter the world's price level or interest rate (at least so the finance ministers of other countries believe). But the American price level and interest rate are no less the world's on this account. That General Motors is big does not put it in a different market for automobiles than British Leyland or Simca. America's money supply may well in some periods act as the world's high-powered money, with multiple effects (although the usual accounts of monetarism do not talk this way); America's policy in some periods may well affect expectations abroad (although the usual accounts of rational expectations do not talk this way). Adopting such arguments would constitute a radical break with Martianism. And in any case it would entail recognizing that prices and interest rates were world, not national, phenomena, which is quite another way of doing macroeconomics. The other way of doing macroeconomics may be called the Iowa City approach. No one doubts that Iowa City has virtually no control over its price level and its interest rates. In very short order an attempt by Iowa City bankers to raise interest rates on loans to twice the market rate would empty the loan offices of the banks. In rather longer order (a month, say), an attempt by Iowa City grocers to raise prices to twice the market would empty the grocery stores. In still longer order (a couple of years, say), an attempt by house owners to raise rents above the cost and value of housing determined by the substitutability of housing for other goods in production and consumption would empty the city. Likewise, no one would believe there was a useful sense in which Iowa City could have a monetary policy. It could impose tariffs or price controls, to be sure. But the more usual and subtle instruments of monetary policy would be blunt in the hands of the First National Bank. If Iowa City had its own money supply (and under the free-banking legislation before the Civil War it, like many American cities, in fact did), increasing the money supply would have no effect on the Iowa City price

125

The Success of Purchasing-Power Parity

level or interest rate, so long as the exchange rate was fixed. No economist would place any credence in a model of the economy of Iowa City that allowed Iowa City's prices and interest rates to be determined wholly or even largely within Iowa City by the forces of aggregate demand and supply. The choice is one between a world in which purchasing-power parity does and does not work well enough to be a good description. Is the price level of the United States (when the exchange rate is fixed) substantially or importantly free to move independent of the price level of the rest of the world? Is the United States (or the United Kingdom or whatever) more like Mars or is it like Iowa City? The choice between the Martian and the Iowa City approaches is an empirical one. The United States is not literally either Mars or Iowa City. The question is which approach is closer to the truth, or if you prefer, which mix of the two is true. In particular, it is not enough to remark blandly that both approaches apply to some degree and then proceed to use one or the other to buttress some conclusion on policy or history. For much of the period 1880 to the present the major economic powers were on literally or virtually fixed exchange rates, and it is to such a case that the argument applies most easily. But it is not true, as some think, that a regime of flexible exchange rates completely unhinges an economy from the world market. With a correction for the exchange rate, purchasing-power parity might still apply (though one would expect uninsurable exchange risk to make the gold points wider). And if purchasing-power parity does apply, then the central bank can have only a neutral effect on the economy. The bank would be free to push the general price level up or down (and could just as well make the exchange rate the policy instrument as the money supply), but could not alter relative prices, pegged by world markets. Relative prices-for instance price of investment goods relative to consumer goods-are commonly objects of monetary policy. The common objectives are unattainable if purchasingpower parity works well. And, to repeat, if governments bind themselves to a fixed exchange rate, they cannot even have a neutral influence on prices. Another red herring sometimes drawn across the trail should be avoided as well: purchasing-power parity is assured in the very long run by the price-specie-flow mechanism. Therefore, the argument goes, the monetary approach, which assumes that markets operate very quickly among nations, is merely another way of expressing conventional monetarism, which assumes that markets do not operate quickly among nations.4 The argument is misleading. The price-specie-flow mechanism is a disequilibrium model. It is essentially that two economies that for some reason develop a divergence in their purchasing-power parities will

126

Donald N. McCloskey and J. Richard Zecher

generate flows of gold that will realign the parities. By contrast, the monetary approach, subspecies purchasing-power parity, is an equilibrium model. It claims, nonetheless, by virtue of an alleged quickness with which price divergences among countries are arbitraged away, to be relevant to a much shorter run than could reasonably be supposed for the other. That in one respect (namely, purchasing-power parity) the models happen to have the same outcome in the long run should not be allowed to obscure that the two exhibit radically different behavior in most other ways. In particular, monetary policy does work in the price-specie-flow model (at least in some short run and at least if the model does not belie itself by introducing a rapid price-specie flow) but does not work in the purchasing-power-parity model except by way of influences on the world money supply. The monetary approach takes much from monetarism but, in the end, differs importantly from the monetarist aproach to national monetary policy. We should point out that the historical record contains little evidence that the price-specie-flow mechanism actually happened. Economists, accustomed to thinking of the Facts of History, may be surprised; it will not surprise historians, hardened to the ubiquity of the Myths of History. The intellectual status of the mechanism is similar to the kinked demand curve of oligopolists: it does not work empirically and is unreasonable besides (for instance, it would provide opportunities for speculative profit). In an earlier paper (1976, p. 367) we reviewed the empirical anomalies in the price-specie-flow mechanism. For instance, we argued that Milton Friedman and Anna Schwartz misapplied the mechanism to an episode in American history. The United States went back on the gold standard in January 1879 at the pre-Civil War parity. The American price level was too low for the parity, allegedly setting the mechanism in motion. Over the next three years, Friedman and Schwartz argued from annual figures, gold flowed in and the price level rose just as Hume would have had it. They conclude (1963, p. 99) that "it would be hard to find a much neater example in history of the classical gold-standard mechanism in operation." On the contrary, however, we believe it seems much more like an example of purchasing-power parity and the monetary approach than of the Humean mechanism. In the monthly statistics (Friedman and Schwartz confined themselves to annual data), there is no tendency for price rises to follow inflows of gold, as they should in the price-specie-flow mechanism; if anything, there is a slight tendency for price rises to precede inflows of gold, as they would if arbitrage were shortcutting the mechanism and leaving Americans with higher prices directly and a higher demand for gold. Whether or not the episode is a good example of the monetary theory, it is a poor example of the price-specie-flow mechanism .5

127

The Success of Purchasing-Power Parity

The price-specie-flow mechanism, therefore, is not a good way to harmonize closed-economy monetarism with the fact that we live in a world of many economies. What should be clear by now is that if purchasing-power parity is found to be a useful characterization of the world, then closed-economy theorizing and empirical work in macroeconomics should be changed to allow for the direct effects of international price arbitrage. Whether monetarists or Keynesians or rational expectationists, economists should begin thinking and measuring in global terms. 2.2

The Root Definition of Purchasing-Power Parity

"Come, come," the representative Martian will say, "don't waste my time-we know that prices diverge. Purchasing-power parity fails." That prices are not identical everywhere is not an important failure of purchasing-power parity. A minor reply is that prices can be different in level but related in their changes, a distinction made in the usual statistical tests of parity. The main reply is that purchasing-power parity is a consequence of rationality in arbitraging. If all the opportunities for riskless (or insured) arbitrage among countries that are profitable at existing interest rates and other costs of arbitrage have taken place, then the price level of the world may be said to have exhausted its ability to determine the price level of one country. Now it may have exhausted it, yet be trivial. To take the single commodity case for illustration (and only for illustration, it being a major theme below that commodity-by-commodity arguments do not suffice), the gold points might be so wide that even though they are not violated they are not useful as describing a constraint on the economy. Wheat and boomerangs in Rumania and Tasmania in 1682 (or in 1982?) may have offered no opportunities for profit by arbitrage ex ante, yet their prices were surely free to move within wide limits independent of each other. Our hypothesis is that in the modern world among the main trading nations the forces of rationality in arbitrage were powerful enough to fix anyone general price level and interest rate ceteris paribus, in terms of the others. The usual statistical question is whether or not the result is a unit elasticity of one price level with respect to another. We shall see in a moment whether the statistical question is the right one. But the price level in one country could be determined by the world in the sense that it was fixed ceteris paribus by the action of arbitrageurs-even if the elasticity were not unity. For this reason arguments that ratios of purchasingpower parities tend to drift are irrelevant.6 They do drift, just as demand curves drift. To say that the law of demand fails because in an uncontrolled experiment the observed price does not correlate well inversely

128

Donald N. McCloskey and J. Richard Zecher

with the quantities would be an error. So too here. The arbitrageurs could be making the American price level exogenous to, say, American monetary policy even if changing technologies of traded as against nontraded goods, changing compositions of market baskets, and changing errors in the underlying statistics caused the American price level to be poorly correlated with other price levels. It is easy to construct examples that illustrate the point. Imagine that electrical equipment produced in the United States sells for 20 percent more than that produced in Germany, after translating the prices at the exchange rate, because of differences in the energy efficiency of the equipment. Imagine further that deviations as little as plus or minus one percentage point from the 20-percent differential would create opportunities for profitable arbitrage, exploited instantly. Now suppose the equilibrium price ratio rises (that is, the ratio dividing profitable from unprofitable arbitrage) suddenly to 30 percent from 20 percent, because of changes in energy prices (worldwide). Suppose finally that the same plus-or-minus-one-percentage-point band exists at the 30-percent differential as at the old 20-percent differential. The economist stumbling on such data might conclude that parity fails: the drift from a 20-percent to a 30-percent differential would be interpreted by him as indicating the poorness of correlation between prices in one place and in another. Yet in the sense relevant here of ceteris paribus the prices in Germany and the United States are mutually linked (plus or minus one percentage point) just as strongly at 30 percent as at 20 percent. A monetary policy in the United States that had as one intended result a rise in electrical-equipment prices in the United States relative to (exchange-adjusted) German prices would fail. We wish, then, to appropriate for purchasing-power parity the prestige of the postulate of rationality. It surpasses belief that many opportunities to make easy money buying low and selling high persist long enough to be observed in economic data. Yet much of the opposition to purchasingpower parity seems to believe that it is so. When specialists in finance such as Richard Roll (1979) think about international markets, they assume with hardly a comment that all opportunities for arbitrage are exhausted in a matter of weeks. Roll remarks that "in the monetary approach ... prices and exchange rates tend toward equilibrium ... in the very long run, say a year or more" (italics added; p. 135).1 His criticisms of the monetary approach come from a novel direction: instead of criticizing the approach for supposing that the long run is as short as a year, he criticizes it for supposing that the long run is as long as a year. The view from finance is refreshing and highly relevant. We consider purchasing-power parity to be a proposition similar in more ways than one to the efficient-markets hypothesis, to be demonstrated on similar grounds, namely, on the ground of the shared belief of economists in

129

The Success of Purchasing-Power Parity

rationality and on grounds of whatever evidence can be adduced to confront the belief. The belief is that there must be nothing systematic or predictable in the future of prices in one country relative to present prices in another (except, indeed, instantaneous equality), or else there is money to be made in exploiting the prediction. If differences in price levels take years to be eliminated by trade, then trade right now-in one commodity or in the CPI-is profitable. The result, we believe, is a "theorem" linking the pursuit of profit and the exogeneity of the general price level. If opportunities for arbitrage are exploited (allowing fully for the cost of transport and information), then the price level of one country is fixed by the rest of the world, even in the very short run. The argument is not a test by itself, but merely a theorem, a higher-order proposition about the relation between equilibrium and the exogeneity of prices. The theorem is a curious product, to be sure, for it is not merely a logical proposition. It requires the world to be arranged in a certain way to be true. We arrive again at an empirical question. Three points of logic nonetheless may make the fundamental theorem of the Iowa City approach more palatable. First, arbitrage does not need to occur commodity by commodity. All prices in an economy are connected to each other. Prices of bricks in New York and London are held together not only by thp. direct forces of arbitrage in the market for bricks itself, weak as they are, but by the indirect forces of arbitrage in related markets-the markets for brick-making labor, say, or the market for lumber for which bricks are a substitute. That nontraded goods exist is sometimes thought to be a rebuttal to purchasing-power parity. Not so, at least if the nontraded goods are provided in markets sensitive to costs (military bases, for instance, may be an exception). It is also thought to be a rebuttal to note that the law of one price need not hold for such-andsuch commodity if that commodity's market is obstructed. Not so, at least if the commodity in question is related in production and consumption to other goods. Commodity-by-commodity thinking can be misleading. The simplest form of the argument is the Walrasian point that one absolute price (the numeraire) serves to set all other prices in the economy, given resources, technology, and tastes. At the extreme, then, if Mars were connected to Earth by the market in chewing gum alone, the two price levels would nonetheless be fixed in relation to each other. How much tighter, one might suppose, would be the relation between two economies connected by the prices of thousands of goods and services. A second point of logic, seldom recognized, is that arbitrage across space down to the extent of the transport-cost wedge is reinforced, for storable commodities (housing, wheat, automobiles, cement), by arbitrage across time down to the wedge of storage costs. Suppose the price of cement rises in the United States because demand has risen. One might say that cement will not immediately flow from Canada or Spain to the

130

Donald N. McCloskey and J. Richard Zecher

United States, that it takes time to reorganize the direction of transport. But if the cement price is expected eventually to take up its usual relation with foreign prices, then the present relation is constrained-it cannot deviate from the long-run relation by more than the cost of "transporting" cement from the long into the short run. The argument applies even to nontradable goods: if the forces of general equilibrium would eventually bring even housing prices, say, into a rough parity, then the storability of housing will enforce the parity earlier. A third point is that a properly measured price index would be an index of characteristics, not named goods and services. The imperfections of the usual price indexes should not be used as evidence for the failure of purchasing-power parity, whether the imperfection is a sheer error in reporting or, as here, an error of concept. It is commonly argued that goods and services, especially manufactured goods, are not perfect substitutes across countries, that the category "vacuum cleaners" contains Panasonic (Matsushita Electric) model Me-881 and Sears Power-Mate model 20 A 2099 with different characteristics, and therefore that departures from the gold points are rational. The argument is that competition between the two models of vacuum cleaners is not perfect. The response must be, for one thing, that the degree of monopoly in international trade is beside the point. Matsushita Electric might well be discriminating in the prices it sets for vacuum cleaners in Japan and in the United States, yet it would still be true that there was a stable relationship between the Japanese and the American price determined, say, by relative elasticities of demand. For another thing, the characteristics making up the good may well have perfect markets. Vacuuming power, ease of use, reliability of service, and the like are separately measurable, at least in the consumer's mind, and each is perfectly substitutable across brand names. The bundle of characteristics called a Panasonic vacuum cleaner may not be exactly duplicated in any other vacuum cleaner, but the price of vacuuming power may be set on a competitive market. Another way of making the point is to think of the prices of named goods and services as being composed, speaking statistically, of many factors in principal components, shared with many other goods. This is a statistical way of stating the general equilibrium point: what must be arbitraged is a relatively small number of characteristics, not each of millions of named goods and services one by one. The degree of identity in products across countries is no more relevant than is the degree of identity of other measures of the composition of consumption. 2.3

Tests of the Efficacy of Arbitrage

Thus armed against irrelevant doubt, we turn to the empirical questions. One can cast light on the degree of thoroughness of equilibrium (or

131

The Success of Purchasing-Power Parity

"arbitrage") by measuring it directly. The measurement is extremely difficult, essentially because a measurement entails second-guessing people in the business. If the right amount of resources is being used for arbitrage, then prices in one place are not free to move independent of prices elsewhere. To deny purchasing-power parity in this behavioral sense is to deny that the right amount of resources is being used for arbitrage. Often unrecognized by critics of purchasing-power parity is that their conclusion that it has failed usually implies an ability to make money. Anyone who knew that purchasing-power parity was true in some long run but not yet true in the present would have a rosy financial future. The divergences from purchasing-power parity detected in the literature are so gross and the statistics purporting to show the divergences so easy to collect that the opportunities for profit are large. Go thee and prosper. The only direct test of the rationality of arbitrageurs, then, is literal second-guessing. That is, one assembles the facts on prices and transport costs, being careful to allow for such subtleties as the cost-of-exchange risk between the time the arbitrage opportunity arises and the time it is exploited, and does the calculation that the arbitrageuse presumably did, or should have done. She herself would use list prices only after ascertaining that they reflected prices at which she could actually transact, allowing for delivery lags, credit terms, and risk of default. If she missed a profitable opportunity, one can either doubt the completeness of one's calculations or doubt her rationality, depending on the strength of one's devotion to the working hypothesis of rationality. The test is very difficult to perform, though it has the compensating merit of being most relevant to the question at issue. We urge others to attempt it.8 Another, more practical test is to examine whether the shipments of goods implied by the supposed opportunities for arbitrage in fact occurred. Enthusiasts of purchasing-power parity would argue that the mere threat of such flows suffices to bring prices speedily back to unprofitable levels of divergence. Doubters of purchasing-power parity would argue that only the flows themselves suffice, and these only over a considerable period. Taking the doubters' view, then, the flows themselves should be observable. A country that exhibits divergences from purchasing-power parity convincing to the doubters should also exhibit lowered exports and increased imports, a trade deficit: a place with "high" prices would have a hard time selling and an easy time buying. The reasoning is, of course, the first step in the elasticities approach to the balance of payments. In other words, the balance of trade should become more negative for a country that was above its trend of purchasing-power parity. Nothing of the sort shows through in the U.S. statistics. One would expect excessive U.S. inflation to hurt U.S. exports. That is, one would expect a rise in the deviation from purchasing-power parity, expressed as

132

Donald N. McCloskey and J. Richard Zecher

a rise in the difference between the U.S. price index and the U.K. or Canadian price index in U.S. dollars, to cause a fall in the trade balance. The equation for the United Kingdom meets the expectation: the sign on d (Pus - ePUK ) is indeed negative (standard errors in parentheses): Change in U.S. trade] = 5.3 - 2.8[d(Pus - ePUK )]. [ balance with U.K. (19.3) (2.5) The Durbin-Watson is 1.59, in the indeterminant range, but of course much better than the result of running the levels instead of the differences. The slope coefficient, however, is insignificant at conventional levels (notice that the insignificance of the constant indicates that there is no linear trend to worry about). It is unclear what insignificance might mean with observations that are not a sample but the universe of the relevant U.S. and British variables in the period. The R 2 of the equation is a mere .02. It is clear what that means. The equation for Canada, which in view of its contiguity with the United States should prove a better test, is worse: Change in U.S. trade] = 3.5 + 0.61[d(Pus - ePCan )]. [ balance with Canada (6.1) (0.73) The sign is perverse, the R 2 only .01, and again the coefficients are insignificant .9 One could certainly raise the R 2 ,s here and perhaps correct the sign by embarking on a search through all possible specifications of lags and functional forms and periods. For instance, from 1880 to 1912 the United States deflates faster (down to 1896, the usual turning point for price series in the period), then inflates faster than the United Kingdom and Canada. True to expectations, the trade balance with the United Kingdom (though not with Canada) exhibits the same U, but inverted. U.K. clamor about German and U.S. competition becomes great in the 1890s and diminishes in the Edwardian boom of exports and investment abroad that followed. But such evidence would have to contend with a markedly parallel movement of net U.S. inflation and net U.S. exports-the reverse of expectations-relative to both the United Kingdom and Canada during the 1920s and 1930s. In other words, the apparent deviations from purchasing-power parity appear not to represent unexploited opportunities for arbitrage. Another way to say it is that the deviations were only apparent. The "deviations" could be, for instance, the result of peculiarities in the price indexesU.S. price indexes are always more volatile than the U.K. ones in the late nineteenth century, which may well say more about how the indexes were constructed than about the underlying real character of price formation. The rise and fall of net exports to the United Kingdom, then, could be a result of matters wholly internal to the countries, such as building booms

133

The Success of Purchasing-Power Parity

in the face of industrial maturity. Alternatively, the "deviations" could be once-for-all (or for-the-duration) changes in the cost of moving goods from one place to another, as it appears are the violent leaps up and down of parity during the Great War. Changes in tariffs or in other legal costs of movement can permit a change in the relative position of two price levels without in any way belying the assertion that the price levels are not free to move where they will. Such a tariff inflation appears to have taken place in Germany during the 1880s (although we have not as yet been able to find out how powerful the explanation is). While prices fell elsewhere, led by agricultural prices responding to competition from Russia and the United States, prices in Germany held up, as Bismarck forged a tariff politics of rye and iron. Further tests of the hypothesis that profitable opportunities for arbitrage arise when measured prices diverge might proceed commodity by commodity, but the aggregate results at least are unpromising for the hypothesis. 2.4 The Irrelevance of Price Regressions Unadorned The representative Martian will by now be close to apoplexy. "Innumerable regressions of one country's prices on another have been performed recently, and half of them show that purchasing-power parity fails. Surely such tests are conclusive." No, they are not. There are two points here. The first, the more fundamental, is that the regressions are not useful tests of our hypothesis, which has to do with the rationality of arbitrageurs, not with how closely one price correlates with another. Our hypothesis says that prices are linked and therefore insensitive to internal forces such as monetary policy (as we shall show presently), not that the prices are linked by some linear relationship having such-and-such a slope. The second point is that even the hypothesis of linear relationship, which unlike ours is not based on a foundation of individual rationality, has been tested inadequately. The tests take the form of regressing, say, the U.S. GNP deflator on the U.K. GNP deflator multiplied by the exchange rate of dollars for pounds.

Pus = a + J3[(e)(PUK )] + error term. If the slope coefficient J3 is significantly different from 1.0 at conventional levels, then purchasing-power parity is said to fail. A number of criticisms can be and have been leveled at such equations, although none is the main point here. It can be argued, for instance, that allowing for simultaneity bias brings the J3 coefficient closer to 1.0 (Krugman 1978). It can be argued that the equation is misspecified, not allowing properly for lags or for secular trends. It can be argued that "significance" is beside the point for a nonsample. It can be argued that the prices, especially the

134

Donald N. McCloskey and J. Richard Zecher

foreign prices on the right-hand side, are measured in error, as they are for instance if the United Kingdom is taken to stand for the world in the calculation of prices, biasing the coefficient towards zero. It can be argued that the slope coefficient could differ from 1.0 if such a difference were implied by the Walrasian general equilibrium. It can be argued that the whole procedure is misleading, because any coefficient different from zero bespeaks a relationship between domestic and foreign prices dangerous to ignore in explaining domestic prices. It can be argued that to judge the hypothesis of parity a failure, and to abandon it, is to flee to evils that we know not of. The main point, however, is that the "failure" of purchasing-power parity in such an equation is not measured against a standard. How close does the slope have to be to the ideal of 1.00 to say that purchasing-power parity succeeds? The literature is silent. The standard used is the irrelevant one of statistical significance. A sample size of a million yielding a tight estimate that the slope was .9999, significantly different from 1.00000, could be produced as evidence that purchasing-power parity had "failed", at least if the logic of the usual method were to be followed consistently. Common sense, presumably, would rescue the scholar from asserting that an estimate of .9999 with a standard error of .0000001 was significantly different from unity in a significant meaning of significance. Such logic also could be applied to findings of slopes of .90 or 1.20, but usually it is not. lD The point is not that levels of significance are arbitrary. Of course they are. The point is that it is not known whether the range picked out by the level of significance affirms or denies the hypothesis. Nor is the point that econometric tests are to be disdained. Quite the contrary. The point is that the econometric tests have not followed their own rhetoric of hypothesis testing. For one thing, as we have said, the errors that tests of significance deal with are errors of sampling, but in many cases there is no sampling involved: we have the entire universe of observations of the general price level in the United States and the United Kingdom 18801940. For another, nowhere in the literature of tests of purchasing-power parity does there appear a loss function. We do not know how much it will cost in policy wrecked or analysis misapplied or reputation ruined if purchasing-power parity is said to be true when by the measure of the slope coefficient it is only 85-percent true. That is, the argument (due to Neyman and Pearson 1933) that undergirds modern econometrics has been set aside here as elsewhere in favor of a merely statistical standard, and an irrelevant one related to sampling error at that. We are told how improbable it is that a slope coefficient of .90 came from a distribution centered on 1.00 in view of the one kind of error we claim we know about (unbiased sampling error, with finite variance), but we are not told

135

The Success of Purchasing-Power Parity

whether it matters to the truth of purchasing-power parity where such limits of confidence are placed. Silence on the matter is not confined to the literature of purchasingpower parity. Most texts on econometrics do not mention that the goodness or badness of a hypothesis is not ascertainable on merely statistical grounds. Statisticians themselves are more self-conscious, although the transition from principle to practice is sometimes awkward. A practical difficulty in the way of using the Neyman and Pearson theory in pure form, say Mood and Graybill (1963, p. 278), is that the loss function is not known at all or else it is not known accurately enough to warrant its use. If the loss function is not known, it seems that a decision function that in some sense minimizes the error probabilities will be a reasonable procedure. Such a procedure might be reasonable for a general statistician who makes no claim to know what is a good or bad approximation to truth in fields outside statistics itself. The procedure is not reasonable for a specialist in international trade or macroeconomics. If the loss function is not known it should be discovered. Finding the loss function amounts to finding out how close the slope has to be to 1.00 in order for it to be reckoned close enough. Every student of the matter is more or less aware of the need for some standard against which to judge the closeness, a standard beyond the probability of being misled by sampling errors (if there really are any) into gullibility or skeptjcism, but no one has provided them. In a superb paper that has by its sheer weight and subtlety turned the tide of battle recently against purchasing-power parity, for example, Kravis and Lipsey (1978), in reporting some calculations of parity relative to base years and some correlation coefficients (both of which "ought" to be 1.00), remark: Each analyst will have to decide in the light of his own purposes whether the PPP relationships fall close enough to 1.00 to satisfy the theories. As a matter of general judgment, we express our opinion that the results do not support the notion of a tightly integrated international price structure. (Pp. 214ff.) The only guidance they provide to evaluating their "general judgment" is a footnote (p. 214) reporting that in the general judgment of Houthakker, Haberler, and Johnson deviations from parity of anything under 10 or 20 percent are acceptable to the hypothesis. ll It happens, incidentally, that the bulk of the Kravis and Lipsey evidence passes rather than fails such a test. But accepting or rejecting one unargued standard of truth by comparing it with another unargued standard of truth does not advance the art of argument in economics very much.

136

Donald N. McCloskey and J. Richard Zecher

To be fair, Kravis and Lipsey are in fact unusually sensitive to the case for having some standard, more sensitive than are economists working the field with more powerful statistical tools. So frequently does their paper make the point that it must be accounted one of the major ones made. Repeatedly, for instance, they draw a distinction between the statistical and the economic significance of their results: "Indeed, even high coefficients of correlation [between domestic and foreign prices] may conceal shifts in relationships that are economically important" (p. 204); "The difference [of slope coefficients from 1.00] may not be large enough to be picked up by a statistical test yet be economically significant" (p. 236); "even a high correlation does not preclude what may be economically significant variations between the two indexes" (p. 242); and so forth, passim. The intellectual sword is sharp. Remarkably, however, they use the sword only against, not in favor of, purchasing-power parity, and never turn it on themselves. After stating repeatedly that they do 'not have a standard by which to judge the hypothesis, they nonetheless conclude: We think it unlikely that the high degree of national and international commodity abitrage that many versions of the monetarist theory of the balance of payments contemplate is typical of the real world. This is not to deny that the price structures of the advanced industrial countries are linked together, but it is to suggest that the links are loose rather than rigid. (Italics added; p. 243.) Every italicized word involves a comparison against some standard of what constitutes likelihood or highness or typicality or linkage or rigidity, yet no standard is proposed. Indeed, the lone page (p. 204) in their fifty that addresses the issue, entitled "Criterion of Similarity of Price Movements," is devoted to dismissing the lone attempt in the literature to offer a criterion (namely, the Genberg-Zecher criterion: if the correlations of prices among different countries are as good or bad as they are inside one country, then the different countries act as one) and to remarking again that the merely statistical standard used elsewhere in the paper is in fact irrelevant. The irrelevance of the merely statistical standard of fit does not bedevil only that half of the empirical literature that finds purchasing-power parity to be wrong. Towards the end of a fine article favorable to purchasing-power parity, Krugman writes (1978 p. 405): There are several ways in which we might try to evaluate PPP as a theory. We can ask how much it explains [that is, R-square]; we can ask how large the deviations from PPP are in some absolute sense; and we can ask whether the deviations from PPP are in some sense systematic. The defensive usage "in some absolute sense" and "in some sense" betrays his unease, which is fully justified. There is no "absolute sense" in

137

The Success of Purchasing-Power Parity

which a description is good or bad. The sense must be comparative to a standard. Similarly, Jacob Frenkel (1978 p. 175) says "if the market is efficient and if the forward exchange rate is an unbiased forecast of the future spot exchange rate, the constant term [in a regression of the spot rate today on the future rate for today quoted yesterday] ... should not differ significantly from zero, the slope coefficient should not differ significantly from unity" (italics added). In a footnote on the next page, speaking of the standard errors of the estimates for such an equation in the 1920s, Frenkel argues that "while these results indicate that markets were efficient and that on average forward rates were unbiased forecasts of future spot rates, the 2-8 percent errors were significant" (italics added).12 What he appears to mean is that he judges a 2- to 8-percent error to be large in some economic sense. In any event, what his results imply about their subject is unclear-purchasing-power parity, because significance in statistics, however useful it is as an input into economic significance, is not the same thing as economic significance. 2.5

The Search for Standards

Results typical of the conventional tests of purchasing-power parity are easy to replicate. Regressions for 1880 to 1940 (running through the First World War, as a more extreme test of the argument) of the U.S. GNP deflator (to avoid the usual criticism of the domination of wholesale prices by traded goods) against Canadian or U.K. prices adjusted by official exchange rates are (all variables in logarithms and standard errors in parentheses): Price in U.S. = 0.83 (0.19) R 2 = .87,D.W. Price in U.S. = 0.26 (0.14) R 2 = .94,D.W.

+ 0.87 [Exchange-adjUsted] (0.04)

Canadian price

.

= 0.29 + 0.93 [Exchange-adjUsted] (0.03)

U.K. price

.

= 0.46

Additional regressions adjusting for autocorrelation and for the simultaneity of prices and exchange rates (during the floating-rate period) yielded ~'s for the United Kingdom that ranged from .91 to 1.02, and for Canada from .35 to 1.00. Most of these regressions have in common the result that at conventional levels of significance, the slope coefficients would be said to be different from 1.00, and about half of the trends would exhibit "drift." In other words, by the usual standards, these regression results would lead to the conclusion that purchasing-power parity "fails." But the regressions also have something else in common-a uniformly

138

Donald N. McCloskey and J. Richard Zecher

5.25 A. Natural Log of the U.S. Implicit Price Deflator 5.00 4.75 4.50

4.25

1889

1899

1909

1919

1929

1939

1919

1929

1939

5.20 B. Purchasing Power Parity Prediction of the U.S. Implicit Price Deflator Using Canadian Prices 5.00 4.80 4.60 4.40

1889

c.

1909

Standardized Prediction Errors, 1880 to 1940

1889

Fig. 2.1

1899

1899

1909

1920

1919

1929

1939

u. S. prices predicted by Canadian prices.

"high" correlation as measured by R 2. Consider the plots in figures 2.1 and 2.2 of the actual U.S. price, the price estimated from foreign prices, and their difference, for Canada and the United Kingdom. Although the period includes the Great War and the Great Depression, the foreign prices do predict the gross outlines of the U.S. price. Such eyeballing is another way of saying that the RZ's are high, as they are. A fuller treatment would make the sensitivity analysis explicit, introducing the cost of being wrong in adopting a model in which a was zero and f3 was 1.00. A second standard is suggested by the bottom graphs, which plot the forecasting error in units of standard deviation. The period 1880-1914 is a standard for relative tranquility. This is the standard, a relative one over

time. If one is willing to think of 1880-1914 as tranquil (one may not, of course), then it lends meaning to the "success" or "failure" of purchasing-power parity. Comparing 1880-1914 with 1921-40 reveals no difference in the average deviation from purchasing-power parity. The comparison involving the United Kingdom, indeed, can even include the Great War and its immediate aftermath with no change in result. The turbulence of the 1920s and 1930s, which is said to have loosened the economic ties among nations, appears not to have done so. The price relations are equally close, which suggests that the economic behavior causing prices to move in parallel was uniform. It is no surprise that turbulence would offer high rewards to arbitrageurs and that arbitrageurs would take them.

140

Donald N. McCloskey and J. Richard Zecher

A third standard (our favorite) is the Genberg-Zecher criterion (see section 2.4 above). Our paper, and papers by Genberg (1976, 1978), expanded on this theme in the early 1970s.13 It will suffice for the present to recall the conclusion of our work, namely, that the matrix of the United States, the United Kingdom, Germany, and Sweden (with special attention to the United Kingdom) had correlations of prices in the 1880-1914 period similar to those between different parts of the United States. Genberg performed similar tests with similar results for recent times. The few attempts since to undermine the conclusion leave us unmoved. Arthur Gandolfi and James Lothian (1982) have written an interesting paper on the subject, but demand that the hypothesis pass a test of a lower correlation between very close states of the United States than between far nations of the world. A common reaction to the standard is, "Well, suppose we do not even accept the premise that the United States is a unified market?" If one does not accept such a premise, then of course one will refrain from speaking of "the" U.S. price level. Nor will one talk of "the" U.S. GNP. If the United States itself is a mere collection of wholly local markets, there is no more use for the talk than for talk about macroeconomic variables for a federation of Fiji-Botswana-Iceland-Saskatchewan, that is, for a random assortment of places. Fourth and finally, purchasing-power parity can be tested against the standard of explaining major events or illuminating puzzles in history. We have mentioned the event of the United States returning to gold in 1879 at the wrong parity. Consider another episode, puzzling to all who have noticed it, the inflation of 1933-34. It was a spectacular outlier of the Phillips curve, an example of stagflation forty years before that hideous neologism was coined. Unemployment was very high, yet prices rose, whatever notion of "the" price level one uses: a 14 percent rise in wholesale prices would have caused alarm even in 1981; a 3 percent rise in consumer (retail) prices would have caused alarm at least in 1960 (table 2.1). Why did it happen? While recognizing the indirect force of the world price level, Friedman and Schwartz (1963, pp. 498-99), argued that: Another [factor] was almost surely the explicit measures to raise prices and wages undertaken with government encouragement and assistance, notably, NIRA [the National Industrial Recovery Act, leading to the National Recovery Administration, or NRA], the Guffey Coal Act, the agricultural price-support program, and the National Labor Relations Act. ... We have grave doubts that autonomous changes in wages and prices played an important role [after World War II. But] there seems to us a much stronger case for a wage-price or price-wage spiral interpretation of 1933-37.14

141

The Success of Purchasing-Power Parity

Table 2.1

U.8. Price Indexes 1932-36 (1933 = 100)

Definition of Prices 1. GNP deflator, 1929 weights 2. GNP deflator, 1954 weights 3. Personal consumption, 1954 weights 4. Consumer price index 5. Retail food 6. Wholesale prices, 1926 base 7. Average annual earnings per full-time employee, all industries

1931

1932

1933

1934

1935

1936

113

101

100

106

105

109

113

102

100

106

107

108

117

104

100

106

108

109

118 124

106 103

100 100

103 112

106 119

107 120

100

114

121

123

100

104

108

113

111

122

98.3

107

Source: U.S. Bureau of the Census 1960. Row 1, series F 1/2; row2, F67, 87; row 3, F68, 88; row 4, E 113; row 5, E 114; row 6, E 13; row 7, D 696.

This is an unusual line of reasoning for such crusaders against mixing up the determination of relative and absolute prices. It also seems to us to square poorly with the evidence. The chief factual difficulties with the notion that the official cartels sanctioned by the NRA codes caused a rise in the general price level is that most of the NRA codes were not enacted until after the price rise. Ante hoc ergo non propter hoc. Look at the plot of wholesale prices of 1933 in figure 2.3 (retail prices, including such nontradables as housing, show a similar pattern). Most of the rise occurs in May, June, and July of 1933, but the NIRA was not even passed until June. A law passed, furthermore, is not a law enforced. However eager most businessmen must have been to cooperate with a government intent on forming monopolies, the formation took time. As the Bureau of Labor Statistics described it: The monthly load of code approvals reached its peak in the period from October 1933 to March 1934; thereafter there was a rapid decrease. Many of the large employing industries were codified in the latter part of 1933 [mentioning cotton textiles, petroleum, bituminous coal, retail trade, fabricated metal products, retail food]. The National Recovery Administration estimated in a report issued in February 1934 that codification of American industry under the industrial self-government program contemplated by the act was 90 percent completed.I5

Fig. 2.3

Exchange Rate

Jan

I

Feb

£

g>

~

j

4-

:-...'

~/~

~

Scale)

April

~

f?

a:

c(

~

j

g

. -."... .....1 ~

~ March

c:

.:.t.

(5 J:

:2

~

~

.!Cl

~ ~c:

~

~

j

51o

(3

....

May

' '

June

J,...... ~ •J......

~g

::J -,..JU

~~~ OO

oc&i

~~ ~

a:~ o:!2

§15

~~

j~

c:-o

~~

o &.~

"

~/

July

,

~

a::

.~

"0

-64

Dec ~ - 6~ '0

~ -66

~ -67

&

8- 68

t-!e9 ~

-70

~ u..

~

-60

- 6:

~-

62

Nov

1934

Range _ 72

..

_----..{

,,--~

~ - 63

Oct

"

,.. ]

Wholesale Price Index, BLS Handbook of Labor Statistics 1936 Edition, pp, 684-85. Averages for Weeks Ending Friday (Same Friday as Exchange Rates).

sept

l

"

Friday Spot Exchange Rates in London from Paul Einzig, The Theory of Forward Exchange (1937).

Sources:

Aug

\ .. ~I••_ ......' "

f'....·........

~I ::J'

T"""

l

Close weekly relationship between the exchange rate and the wholesale index, U.S.A., 1933. Source: Friday spot exchange rates in London, from Einzig 1937; wholesale price index, BLS Handbook of Labor Statistics, 1936 Edition, pp. 684-85, averages for weeks ending Friday (same Friday as exchange rates).

$3.00 -

...............J......-.......

$3.50-

$4.00-

$4.50 -

$5.00-

($ pert)

143

The Success of Purchasing-Power Parity

By September 1933, apparently before the approval of most NRA codes-and, judging from the late coming of compulsion, before the effective approval of agricultural codes-three-quarters of the total rise in wholesale prices and more of the total rise in retail food prices from March 1933 to the average of 1934 was complete. On the face of it, at least, the NRA is a poor candidate for a cause of the price rise. It came too late. What came in time was the depreciation of the dollar, a conscious policy of the Roosevelt administration from the beginning, although not usually believed to have taken effect until the fall of 1933. There was certainly no contemporaneous price rise abroad to explain the 28-percent rise in American wholesale prices (and in retail food prices) between April 1933 and the high point in September 1934. In fact, in twenty-five countries the average rise was only 2.2 percent, with the American rise far and away the largest. The close link between exchange rates and U.S. wholesale prices is clear in the weekly series graphed in figure 2.3 above. Note especially the two sharp jumps around 20 April and 2 July in response to explicit announcements by Roosevelt of the intent of his administration to devalue the dollar. Wholesale prices move up simultaneously. It would appear, in short, that the economic history of 1933 cannot be understood with a model closed to direct arbitrage. The inflation was no gradual working out of price-specie flow; less was it an inflation of aggregate demand. It happened quickly, well before most other New Deal policies (and in particular the NRA) could take effect, and it happened about when and to the extent that the dollar was devalued. By the standard of success in explaining major events, parity here works. 2.6

Purchasing-Power Parity and Monetary Policy

In the style of the doubts expressed above about price-specie flow, the success of parity can be judged by the failure of the alternatives. A common view in much of purchasing-power-parity literature is that while international trade places limits on what exchange-adjusted domestic prices can be, there is nevertheless considerable flexibility for prices to move up and down. It is argued in particular that a country can raise its price level relative to the exchange-adjusted price abroad by expanding the domestic money supply. If the activities of traders and arbitrageurs fix prices within very narrow ranges, however, such policies would not work. Consider, then, the relationship between monetary policy and the "errors" in purchasing-power-parity forecasts. The model is monetarist, postulating that excessively rapid money growth will put upward pressure on prices, leading to domestic prices that are systematically over the purchasing-power-parity predictions. This is a

144

Donald N. McCloskey and J. Richard Zecher

test, then, to what degree a country, the United States in this case, can through its monetary policies affect its price level (Pus) relative to prices (adjusted for exchange rates) in the rest of the world (Pus). In general equilibrium terms, the hypothesis concerns a state of disequilibrium in the goods markets that is matched by an offsetting disequilibrium in the domestic money market. The regression model below represents the goods-markets disequilibrium by the difference between actual U.S. prices and the purchasing-power-parity prediction of U.S. prices using Canadian and U.K. prices and exchange rates. Disequilibrium in the U.S. money market is the difference between growth in money supply a (M S ) and money demand a (M D ). Two measures of money supply are used: (1) M2, which reflects the Martian view that the United States could control its total money supply over the 1880-1940 period, and (2) domestic credit, omitting the effects of specie flows on money supply, which reflects the view that for most of this period the United States could only affect domestic credit, not total money supply. Growth in money demand is represented by the sum of growth in real income and in prices, on the assumptions that money demand is unit elastic with respect to both of these variables: i.e., a M D = a y + a P, where y is real income and P is the price level. Thus the regression equation becomes:

a (Pus -

-"

Pus)

= 0:

+ ~a (M S - M D )t

+ ~a (M S

-

MD)t_l.

The regression results are reported in tables 2.2 and 2.3. Of the thirty-six estimated coefficients relating excess money growth to changes in the purchasing-power-parity forecast error, twenty-two are negative and sixteen are positive. Two of these coefficients are significant at the 5 percent level; both are negative and both are for regressions using M2 as the measure of money. Only two of the twenty-four regressions have R 2 s above .08. There is little support here for the notion that the errors in purchasing-power parity are related to domestic monetary conditions. The alternative to the international determination of prices appears to work poorly. 2.7

Conclusion

The argument and evidence presented here make a pronouncement of the failure of purchasing-power parity impossible, and without a pronouncement of failure much of modern macroeconomics is badly damaged. The failure of purchasing-power parity must be large indeed to leave the Martian models and thei'r empirical implementations unscathed, whether these are Keynesian, monetarist, or rationally expecta-

145

The Success of Purchasing-Power Parity

tionist. Under either the cleanly fixed exchange rates that typify the historical periods used to test the models or under the dirty float that typifies the years in which the conclusions thus tested have been used for policy, it is hard to believe that foreign prices or interest rates did not matter. Yet the silence of most American macroeconomists on the role of

Table 2.2

Calculated Forecast Errors Implicit Deflator Canadian

U.K.

CPI Canadian

U.K.

WPI Canadian

U.K.

Purchasing-Power-Parity Forecast Errors For the U.S. as a Function of Measures of Excess Money Growth, 1880-1940

Constant

Current Excess Money Growth

0.018 (0.038) 0.019 (0.037)

0.109 (0.166) 0.125 (0.166)

0.087 (0.098) -0.089 (0.098)

0.040 (0.429) -0.080 (0.431)

0.026 (0.060) 0.028 (0.060)

0.077 (0.263) 0.112 (0.262)

-0.117 (0.117) -0.117 (0.118)

0.154 (0.512) -0.167 (0.519)

0.489 (0.096) -0.488 (0.097)

-1.114 (0.421) -1.126 (0.427)

-0.404 (0.139) -0.402 (0.140)

-0.343 (0.610) -0.304 (0.615)

Lagged Excess Money Growth

0.163 (0.166)

0.417 (0.431)

0.360 (0.262)

0.132 (0.519)

-0.123 (0.427)

-0.406 (0.615)

R2

D.W.

0.007

1.905

0.023

1.893

0.000

2.160

0.016

2.108

0.001

2.208

0.033

2.327

0.001

2.481

0.003

2.484

0.106

1.609

0.107

1.621

0.005

2.100

0.013

2.124

Sources: Canada: Deflator, Firestone 1958; CPI, Urquhart and Buckley 1965, tables J165,

139; WPI, ibid., table J34. U.K.: Deflator, Feinstein 1971; CPI, McCloskey and Zecher 1976; Mitchell 1975 , series 12; WPI, Mitchell 1975 series 11; Board of Trade series spliced at 1919-20 and 1929-30. U.S.: Deflator, U.S. Bureau of the Census 1975, series F5; CPI, ibid., series E 135; WPI, ibid., series E 23; money supply and gold flows, Friedman and Schwartz 1963, pp. 704-7; real GNP, McCloskey and Zecher 1976; U.S. Bureau of the Census 1975, series F 3. Note: Values in parentheses are the standard errors of the coefficients.

146

Donald N. McCloskey and J. Richard Zecher

Table 2.3

Calculated Forecast Errors Implicit Deflator Canadian

U.K.

Purchasing-Power-Parity Forecast Errors For the U.S. as a Function of Measures of Excess Domestic-Credit Growth, 1880-1940

Constant

Current Excess DomesticCredit Growth

0.004 (1.175) -0.410 (1.223)

-0.002 (0.010) -0.004 (0.010)

-0.674 (1.069) -1.014 (1.116)

-0.008 (0.009) -0.010 (0.009)

0.146 (0.756) 0.012 (0.795)

-0.003 (0.006) -0.003 (0.006)

-0.094 (1.270) -0.233 (1.338)

-0.010 (0.010) -0.010 (0.011)

-0.405 (0.609) -0.088 (0.264)

-0.006 (0.005) -0.004 (0.005)

-0.012 (0.835) 0.098 (0.880)

-0.008 (0.007) -0.008 (0.007)

Lagged Excess DomesticCredit Growth

0.144 (0.124)

0.119 (0.113)

R2

D.W.

0.001

1.719

0.026

1.726

0.015

2.564

0.035

2.632

0.003

1.912

0.010

1.926

0.016

1.694

0.019

1.726

0.023

2.307

-0.076

2.246

0.026

1.937

0.029

1.898

CPI Canadian

U.K.

WPI Canadian

U.K.

0.047 (0.080)

0.048 (0.135)

-0.110 (0.063)

-0.038 (0.089)

Sources: See table 2.2. Note: Values in parentheses are the standard errors of the coefficients.

the rest of the world in their models implies such a belief. It is hard to believe that American prices and interest rates are not at all constrained directly by the forces of arbitrage. Yet the journals are filled with work embodying this belief. The failure of arbitrage necessary to validate a Martian model must be gross, not a matter of the fourth digit of accuracy but of the first. To the first digit of accuracy, and even to the second, the hypothesis of parity succeeds. One wonders what would happen to estimates of wage and price equations, or of the effects of domestic monetary policy on prices and interest rates, or of optimal economic policy under

147

The Success of Purchasing-Power Parity

rational expectations, if they were each asked to embody the international milieu of the U.S. economy to the second digit of accuracy. The hypothesis of parity survives the test for the reasons usual in economic arguments. Economists are embarrassed to assert in print that they possess the economic equivalent of a perpetual-motion machine, and the gross violations of rationality that opponents of purchasingpower parity believe they see entail such a machine. General equilibrium makes the hypothesis still more robust; even nontraded goods are substitutes in consumption and production with traded goods; a few characteristics are tradable in markets even when named goods are not; and the tendency for parity to hold in the long run gives opportunities for speculative profits if it does not hold in the short. The standards by which parity "fails" empirically are unclear, as many opponents of the hypothesis readily admit. The literature contains no articulation of standards. The introduction of standards casts into doubt all the recent attacks on parity. By standards that make intellectual sense the hypothesis succeeds. It succeeds in explaining the U.S. price level from 1880 to 1940 to a standard of accuracy demanded of such explanations. 16 It succeeds in explaining the price level in turbulent periods by the standard of tranquil periods. It succeeds in explaining the difference in prices among countries by the standard of the difference in prices among places in a single country. One of its competitors, closedeconomy monetarism, fails to explain the residual deviations from parity by the standard of statistical fit. Another of its competitors, the elasticity approach, fails to explain the balance of trade by the same standard, although again we express our doubt that much can be inferred from the uncontrolled experiments in curve fitting that characterize the literature and that we have dutifully followed here. And by the standard of good storytelling that underlies all economics, applied to episodes from their beginning to their end, the hypothesis of parity explains what happened under the gold standard. Purchasing-power parity is not a failure. On the contrary, by the standards we have examined, it is a great success. And at the least, speaking to the most skeptical reader, it is not so great a failure that macroeconomics can go on ignoring the rest of the world.

Notes 1. See R. Dornbusch and D. Jaffee's (1978) introduction to the special issue of the Journal of International Economics on purchasing-power parity. They remark that the Kravis and Lipsey paper in the issue leaves purchasing-power parity "rather in a shambles" (p. 159). 2. Samuelson, the chief theoretician of the real theory, however, is an implacable foe of purchasing-power parity.

148

Donald N. McCloskey and J. Richard Zecher

3. Gottfried Haberler once described Cassel's work as "one part Wicksell and nine parts water." One can also read the General Theory as taking prices to be given (not constant), because given by international factors such as purchasing-power parity. 4. Friedman and Schwartz (1982, p. 318) assume no quick operation of price-specie flow. Were it as quick as in the monetary approach, many of their other conclusions would be wrong, especially the effect of domestic money on prices. 5. Prices might have risen in anticipation of gold flows, though such rationality runs counter to the usual price-specie-flow argument. 6. See Darby 1982. Kravis and Lipsey (1978) make a similar point, arguing that the drift of purchasing power from parity over periods of a decade or so shows that parity is false. 7. Roll says further that "In an efficient market, something so easy to detect and so intoxicating to the arbitrager as a relative price difference in two locations would presumably display an immeasurably short half-life" (p. 136). 8. One of us (McCloskey 1981, chaps. 4, 6) has in fact done similar work with the notion that English businessmen failed to adopt profitable novelties in the late nineteenth century. The task is to see whether the second-guesser could have done better, recognizing the limitations of resources they faced (including a limitation on prescience). It is worth knowing for the present context that the second-guessing showed that the businessmen knew what they were doing. 9. The source for the trade statistics is U.S. Bureau of the Census 1960 based on declarations to American customs. The period covered is 1880-1940. 10. A good-or bad---example is the paper by J. D. Richardson (1978). Richardson regresses Canadian on U.S. prices (multiplied by the exchange rate) for a number of industries and concludes: "it is notable that the 'law of one price' fails uniformly. The hypothesis of perfect commodity arbitrage is rejected with 95 percent confidence for every commodity group" (p. 347). 11. To which may be added an authority overlooked by Kravis and Lipsey, Leland B. Yeager (1958). Yeager reckoned that if twenty-six out of thirty-five countries were within 25 percent of their 1937 parity by 1957, the hypothesis is confirmed. Richard Caves and Ronald Jones (1973) may be added to the affirmation. They remark on Yeager's results that "this performance seems rather good-just how good it is hard to say" (p. 338). The rueful remark, "just how good is hard to say," illustrates well how urgent it is to develop some standard. The state of play, largely favorable to purchasing-power parity, is well described in a comprehensive review by Lawrence H. Officer (1976). 12. In another article, Frenkel (1981), concludes that the "collapse" of PPP during the 1970s was due to a fundamental difference between exchange rates and prices. "Exchange rates reflect expectations about future circumstances while prices reflect more present and past circumstances" (p. 162). In this view, arbitrageurs in commodity markets look backwards in time while arbitrageurs in financial markets look forwards. For a different view, see Nattress and Zecher 1982, where a theory of the arbitrageur is developed. 13. Genberg writes (1976, p. 302): "it is evident ... that the differences between OECD countries are no greater on the average than those between cities within the United States. Thus, if we believe that the whole of the U.S. can be treated as a single market in a macroeconomic context, then the area composed of the above countries can be treated likewise." 14. Friedman and Schwartz speak of 1933-37 because of their commitment to analysis by cycles, 1933-37 being an upswing. Notice that by all measures except the consumer price index most of the price rise had occurred before 1935. The discussion of indirect influence of devaluation occurs at any rate on pp. 465ff, at the beginning of which a direct influence on "most farm products and raw materials exported by the United States" is mentioned. 15. BLS Handbook of Labor Statistics, 1937, pp. 512ff. The report referred to was "Report on the Operations of the National Recovery Act," p. 7. 16. Another related standard that we shall explore in later work is that of the relative

149

The Success of Purchasing-Power Parity

convergence of prices over time. Spooner's magnificent graphs (Braudel and Spooner 1967, pp. 470-71) show ranges of wheat prices expressed in silver in Europe and its offshoots of 6.66 to 1 around 1400, falling steadily to 1.88 to 1 around 1750. The divergences of the late nineteenth century, not to speak of the twentieth, look trivial beside these. Correspondingly, fixation on the "failure" of the unity of world markets for the period 1880-1939 looks odd indeed.

References Braudel, F. P., and Frank C. Spooner. 1967. Prices in Europe from 1450 to 1750. In The Cambridge Economic History of Europe, ed. E. E. Rich and C. H. Wilson, vol. 4. Cambridge: Cambridge University Press. Caves, Richard, and Ronald Jones. 1973. World trade and payments. Boston: Little Brown. Darby, Michael R. 1982. Does purchasing power parity work? In Proceedings of Fifth West Coast Academic/Federal Reserve Economic Research Seminar, December 1981. San Francisco: Federal Reserve Bank of San Francisco. Dornbusch, Rudiger, and Dwight Jaffee. 1978. Purchasing power parity and exchange rate problems: Introduction. Journal of International Economics 8 (May): 157-61. Einzig, Paul. 1937. The theory offorward exchange. London: Macmillan. Feinstein, Charles H. 1971. National income, expenditure, and output of the United Kingdom, 1855-1965. Cambridge: Cambridge University Press. Firestone, O. J. 1958. Canada's economic development, 1867-1913. Income and Wealth series VII. London: Bowes and Bowes. Frenkel, Jacob A. 1978. Purchasing power parity: Doctrinal perspective and evidence from the 1920s. Journal of International Economics 8 (May): 169-91. - - - . 1981. The collapse of purchasing power parities during the 1970s. European Economic Review 16 (May): 145-65. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. - - - . 1982. Monetary trends in the United States and the United Kingdom, 1867-1975. Chicago: University of Chicago Press. Gandolfi, Arthur E., and James R. Lothian. 1982. A comparison of price movements among countries and states. Citibank. Typescript. Genberg, A. Hans. 1976. Aspects of the monetary approach to balance-of-payments theory: An empirical study of Sweden. In The monetary approach to the balance of payments, ed. J. A. Frenkel and H. G. Johnson. Toronto: University of Toronto Press.

150

Donald N. McCloskey and J. Richard Zecher

- - - . 1978. Purchasing power parity under fixed and flexible exchange rates. Journal of International Economics 8 (May): 247-76. Kravis, Irving B., and Robert E. Lipsey. 1978. Price behavior in the light of balance of payments theories. Journal of International Economics 8 (May): 193-246. Krugman, Paul R. 1978. Purchasing power parity and exchange rates: Another look at the evidence. Journal of International Economics 8 (Aug.): 397-407. McCloskey, Donald N. 1981. Enterprise and trade in Victorian Britain: Essays in historical economics. London: Allen and Unwin. McCloskey, Donald N., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance of payments. See Genberg 1976. Mitchell. B. R. 1975. European historical statistics. New York: Columbia University Press. Mood, Alexander M., and Franklin A. Graybill. 1963. Introduction to the theory of statistics. 2d ed. New York: McGraw-Hill. Nattress, Dayle, and J. Richard Zecher. 1982. The theory of the arbitrageur and purchasing-power parity. Chase Manhattan Bank. Typescript. Neyman, Jerzy, and E. S. Pearson. 1933. On the problem of the most efficient tests of statistical hypothesis. Royal Society of London. Philosophical Transactions 231 (series A): 289-337. Officer, Lawrence H. 1976. The purchasing-power-parity theory of exchange rates: A review article. IMF Staff Papers 23 (Mar.): 1-60. Richardson, J. David. 1978. Some empirical evidence on commodity arbitrage and the law of one price. Journal of International Economics 8 (May): 341-51. Roll, Richard. 1979. Violations of purchasing power parity and their implications for efficient international commodity markets. In International finance and trade, ed. M. Sarnat and G. Szego. Cambridge: Ballinger. Taussig, Frank W. 1918. International freights and prices. Quarterly Journal of Economics 32 (Feb.): 410-14. U.S. Bureau of the Census. 1960. Historical statistics ofthe United States, colonial times to 1957. Washington, D.C.: Government Printing Office. - - - . 1975. Historical statistics of the United States, colonial times to 1970. Part 1. Washington, D.C.: Government Printing Office. Urquhart, M. C., and K. A. H. Buckley. 1965. Historical statistics of Canada. Toronto: Macmillan of Canada. Wicksell, Knut. 1918. International freights and prices. Quarterly Journal of Economics 32 (Feb.): 404-10. Yeager, Leland B. 1958. A rehabilitation of purchasing-power parity. Journal of Political Economy 66 (Dec.): 516-30.

151

The Success of Purchasing-Power Parity

Comment

Robert E. Lipsey

How to define purchasing-power parity and how to test its "success" is the central issue of McCloskey and Zecher's paper. The authors come to two conclusions about purchasing-power parity, both favorable. One conclusion is derived from their reasoning and one from their tests. The first conclusion is that purchasing-power parity works instantaneously and perfectly, immediately eliminating any genuine differences in prices or divergences in relative price movements. The second is that the theory works sufficiently well that foreign influences on prices cannot be ignored. I shall argue that the first conclusion, as stated, depends on a definition of purchasing-power parity that makes it a tautology, not susceptible to proof or disproof, and that the second conclusion is valid but would not be disputed by many "critics" of the theory. I shall argue in addition that the strongest test they propose has been performed, with results that are unfavorable to the theory by their own criterion. The Definition of Purchasing-Power Parity In discussing the meaning of the concept, the authors say that they view purchasing-power parity as resulting from arbitrage in commodity markets. To me that means they identify the concept with the operation of the law of one price rather than with vaguer notions of aggregate country price levels or price changes. For example, "purchasing-power parity is a consequence of rationality in arbitraging. If all the opportunities for riskless (or insured) arbitrage among countries that are profitable at existing interest rates and other costs of arbitrage have taken place, then the price level of the world may be said to have exhausted its ability to determine the price level of one country" (p. 127), and "If opportunities for arbitrage are exploited (allowing fully for the cost of transport and information), then the price level of one country is fixed by the rest of the world, even in the very short run" (p. 129). Two things can be said about this definition or "theorem" or "higherorder proposition" as the authors refer to the latter version. One is that it is not what most people mean by purchasing-power parity. A second is that, stated this way, it is only a definition, not a theory, and is not susceptible to testing. The authors make no effort to apply it empirically. If one includes and takes seriously all the qualifying phrases about costs of arbitrage over space and time, most of which are not measured and probably cannot be measured, it is impossible to show that purchasingpower parity has been violated. The price of product x in country a plus all costs of arbitrage (including information costs, advantages of continuity of supplier relationships, costs of adaptation of existing machinery and Robert E. Lipsey is professor of economics at Queens College of the City University of New York and research associate of the National Bureau of Economic Research.

152

Donald N. McCloskey and J. Richard Zecher

work habits, etc.) equals the price of product x in country b. That statement is an identity; it tells us nothing about how the world works. By this standard, prices on Mars and Venus satisfy purchasing-power parity even though there is no communication between them. The information and arbitrage costs are so high that any prices or price changes satisfy the authors' criterion: no one can make money by arbitrage between the two planets. Critics of purchasing-power parity do not deny this tautology. What they argue is that adjustment costs are large in some cases, that it may take a long time to overcome them, that prices can move relative to each other in the meantime, and that consumers and producers react to these price differences and relative price movements. The authors illustrate their point about arbitrage with examples of changing price differences for electrical equipment, cement, and vacuum cleaners. It is clear from these examples that they picture the mechanism often called the law of one price, referring to prices of individual commodities, as enfor.cing purchasing-power parity. They do not give vivid examples of arbitrageurs buying the U.S. CPI or WPI or GNP deflator and selling that of the U.K. Of course, if the law of one price operated exactly and instantaneously, that is, if prices of carefully defined individual products were identical everywhere or moved identically in different countries, the levels of prices in general and their movements would be similar or would move similarly. However, aggregate price levels and price changes would not be the same. Given the differences among countries in the composition of consumption and production and what are probably even greater differences in the way that aggregate price measures are constructed, one must be careful in moving from one kind of statement to the other. The authors take little note of this point and speak about the movements of vaguely defined and badly measured price aggregates in different countries and the responses to them as if they were useful for testing the opportunities for, or existence of, arbitrage. I do not think they are. Testing Purchasing-Power Parity The first test the authors propose is that a critic of purchasing-power parity show that he has gotten rich on arbitrage profits. "It surpasses belief that many opportunities to make easy money buying low and selling high persist long enough to be observed in economic data" (p. 128). "Often unrecognized by critics of purchasing-power parity is that their conclusion that it has failed usually implies an ability to make money. Anyone who knew that purchasing-power parity was true in some long run but not yet true in the present would have a rosy financial future. The divergences from purchasing-power parity detected in the literature are so gross and the statistics purporting to show the divergences so easy

153

The Success of Purchasing-Power Parity

to collect that the opportunities for profit are large. Go thee and prosper" (p. 131). "Economists are embarrassed to assert in print that they possess the economic equivalent of a perpetual-motion machine, and the gross violations of rationality that opponents of purchasing-power parity believe they see entail such a machine" (p; 147). The number of times the point appears shows that the authors take it seriously, but I have never seen a criticism of purchasing-power parity that implied irrationality on the part of purchasers. There is strong evidence that in the capital goods and other complex products that form a large part of the trade of developed countries, price differences and divergences in price movements exist. They are not arbitraged away immediately, but they do eventually bring about shifts in trade that tend to remove them (an indication that they are not simply illusory or due to differences in specifications). The reasons are implicit in some of the authors' own discussion: information is costly and the risks of purchasing unknown types of machinery and dealing with unfamiliar suppliers are high and uninsurable. A fall in the price of a Japanese machine might at first produce no shift in purchases because buyers were unaware of the change, uncertain about its permanence, or skeptical about the quality of the machine or the availability of spare parts. There would be no violation of the law of one price by the authors' definition, because the price difference was insufficient to offset information or risk or insurance costs. After the lower price had been in effect for a year, information would become more widely and cheaply available, risk and insurance costs would decrease, and some buyers would switch. After another year more information would be available and still more buyers would switch. At each point, the price in Japan plus information, risk, and insurance would have been equal to the price in the United States. Therefore this sequence of events, which would suggest a violation of the law of one price to most observers, would be in conformity with it by the authors' definition. In view of the impracticality of calculating in retrospect all the costs of arbitrage, the authors go on to suggest a "more practical" test. That test is "whether the shipments of goods implied by the supposed opportunities for arbitrage in fact occurred.... A country that exhibits divergences from purchasing-power parity convincing to the doubters should also exhibit lower exports and increased imports ... a place with 'high' prices would have a hard time selling and an easy time buying" (p. 131). One might expect that the authors would perform their "practical" test on the data for commodities defined as narrowly as possible to observe the action or inaction of arbitrageurs. Instead, they perform their tests of responses to deviations from purchasing-power parity between the United States and the United Kingdom and the United States and Canada using aggregate price indexes. If they seriously believe that profit

154

Donald N. McCloskey and J. Richard Zecher

seeking by arbitrageurs is the force at work, and they wish to observe that action, why measure purchasing-power parity from price aggregates? The arbitrageur cannot easily buy the V.S. wholesale price index or GNP deflator. A more appropriate test of the workings of arbitrage, and one the authors themselves suggest, would be to compare price and trade changes for identical or related goods such as the electrical equipment they refer to. As they say, "Further tests of the hypothesis that profitable opportunities for arbitrage arise when measured prices diverge might proceed commodity by commodity." In fact, Irving Kravis and I, using price indexes that were constructed so that the same goods with the same weights were represented in two countries' price measures did perform many versions of the "practical" test they suggest, although not with the intention of testing purchasingpower parity. That is, we investigated whether price changes and differences in price levels did lead to shifts in trade. We found strong evidence that they did and that these shifts took years before they were completed. We explained the reasons for these lags in our book on price competitiveness (Kravis and Lipsey 1971) and in a series of later articles. l The reasons we gave for the price and price-change divergences we found did not imply that there were overlooked opportunities for aboveaverage profit, given the costs of information, costs of adjustment, and uncertainties about the permanence of price changes. It was true that the first V.S. buyers of foreign electrical generating equipment and foreign steel paid less than their competitors who hesitated. We cannot say what the rational policy was for a buyer of generating equipment given the possible uncertainties at the time. The first buyers were public systems which may have faced less danger from mistakes than private utilities. The private utilities may have had little incentive to lower costs, given the way their prices were regulated and the lack of incentives for managers to break cozy relationships with domestic suppliers. The important point is that whatever the reason, gaps in prices persisted for a long time and produced not sudden but gradual shifts in trade; but they did produce the shifts that McCloskey and Zecher imply would refute purchasing-power parity. In looking back at these past episodes in which price differences gradually gave rise to shifts in trade, we do not know whether the first V. S. buyers of foreign electrical generators and transformers or the first V.S. buyers of foreign steel reaped exceptional profits for a time. Even if they did, the profits, as usually measured, may have done no more than compensate for the risks, given the uncertainties about the quality of foreign products, the reaction of the V.S. government and regulatory agencies, the commitment of foreign suppliers to technical assistance, service, and continuity of supply, and many other factors. Thus these events may not represent a violation of purchasing-power parity by their

155

The Success of Purchasing-Power Parity

definition (it would be difficult to do that), but fail their test of conformance with the theory by the "practical" test. If, as stated in the text (p. 127), the hypothesis has to do with the rationality of arbitrageurs, the tests they offer are irrelevant. I know of no writer skeptical of purchasing-power parity who has stated or implied or assumed that the reason for deviations from it is that traders ignore profit opportunities. They believe there have been deviations from purchasingpower parity because it was not profitable to exploit all opportunities or to exploit them immediately, given the costs and uncertainties involved. One does not add to understanding of the international economy by assuming away all these deviations from purchasing-power parity. The paper includes several of the authors' own tests, but these are tests of what I called at the beginning of my comments their second conclusion, or the second version of purchasing-power parity. That second, or weak, version is that "it is hard to believe that foreign prices . . . did not matter" or "that American prices are not at all constrained directly by the forces of arbitrage" (p. 146). They test this version by regressions of the U.S. GNP deflator against aggregate Canadian and British price indexes adjusted by exchange rates, and by plots of the U.S. price indexes predicted by Canadian and British prices and of the prediction errors from the equations. The correlations are "high," they conclude, and "the foreign prices do predict the gross outlines of the American price." Furthermore, they report, the equations predict as well in the turbulent period 1921-40 as in the "tranquil" period 1880-1914. Aside from the point that these equations represent the "uncontrolled experiments in curve fitting" the authors are so critical of in others, the equations are unconvincing in other ways. For example, for the first twenty years of the chart the Canadian price predicts nothing of U.S. price movements. We are then told that foreign prices did no worse as predictors of U.S. prices in turbulent times than in "tranquil" times. But for half of the "tranquil" period foreign price did not predict at all-not a very exacting standard for judging the predictive power of Canada's prices to estimate later U.S. prices. A more serious objection to these tests is that there is no consideration of the possibility that the high correlation stemmed from the effects of common factors, such as World War I and the Great Depression on U.S., Canadian, and British prices at the same time, rather than from the dependence of one country's prices on another's. Judging the Results of Tests of Purchasing-Power Parity A substantial part of the paper is devoted to standards by which to judge the results of tests of purchasing-power parity. The authors take many of their colleagues to task for using vague and ill-defined standards. I cannot see that they escape the same problem. Their paper is filled with the same undefined terms they deplore in others. One reason for their

156

Donald N. McCloskey and J. Richard Zecher

vagueness and that of everyone else is that no one standard can be used for all purposes. The precision required by a speculator might be greater than that needed by a company making a long-term investment or by a government or international agency trying to set an exchange rate, and all of these standards may be far above that required for satisfaction by devotees of purchasing-power parity. The authors' judgments of the results reported by others are clouded by the fact that they seem to conceive of only two possible conclusions. Either the theory is a "success" and is graded A + , or it is a failure and is graded F. They seem to be determined to come out with a grade of A + and to think that all analyses that find deviations from purchasing-power parity imply a grade of F. In fact, most of the studies they cite seem to imply judgments of B or C rather than total success or failure. In summary, my reaction to the two tests of purchasing-power parity proposed by the authors is that one test cannot be failed and that the other test, conducted with the right type of data, usually is failed. Fortunately, there is another theme to the paper, although it is obscured by the extravagant claims made for purchasing-power parity. It is the important and reasonable one that "it is hard to believe that foreign prices or interest rates did not matter. It is hard to believe that American prices and interest rates are not at all constrained directly by the forces of arbitrage," and "it [purchasing-power parity] is not so great a failure that macroeconomics can go on ignoring the rest of the world." If that is the point the authors really want to make, even many skeptics about purchasing-power parity could agree. Note 1. Some of these were referred to in the article (Kravis and Lipsey 1978) quoted by McCloskey and Zecher; we might add to the list Kravis and Lipsey 1982. Of course, similar studies by many others of price elasticities in trade have yielded similar results.

References Kravis, Irving B., and Robert E. Lipsey. 1971. Price competitiveness in world trade. New York: National Bureau of Economic Research. - - - . 1978. Price behavior in the light of balance of payments theories. Journal of International Economics 8 (May): 193-246. - - - . 1982. Prices and market shares in the international machinery trade. Review of Economics and Statistics 64 (Feb.): 110-16.

157

The Success of Purchasing-Power Parity

Comment

Milton Friedman

I believe that a fundamental confusion runs through the McCloskeyZecher treatment, both in this paper and in their earlier paper-a confusion between two very different propositions. One proposition is whether the quantity of money in a country is an endogenous or an exogenous variable-an important and very interesting question. There is no doubt that in a world of fixed exchange rates and stable barriers to trade, the quantity of money in each country separately is ultimately an endogenous variable. That proposition is perfectly clear and everybody has accepted it for a long time. It's the proposition that Keynes presented so well in the appendix to his Tract on Monetary Reform ([1923] 1971), that there are three things of which a country can choose any two: stable internal prices, stable exchange rates, and free trade. You can't have all three; you can have only two. That proposition is critical. It should be noted, however, that while the quantity of money is ultimately an endogenous variable, there can be and is much leeway in the short run, before the external forces overwhelm the independent internal effects. And we have repeatedly been surprised in our studies by how much leeway there is and for how long-frequently a number of years. There's a very different proposition that is easily confused with the endogeneity or exogeneity of the quantity of money, namely, if money is endogenous, there is no causal relation between money and prices. That is a whole different proposition. Whatever may determine the quantity of money within a country, that quantity of money may still largely determine-or at least, be the conduit through which other forces determineprices within a country. The confusion between these two wholly different propositions is apparent in the statement by McCloskey and Zecher that "if purchasing-power parity is found to be a useful characterization of the world"-and they should have added "and fixed exchange rates characterize the world"-"then closed-economy theorizing and empirical work in macroeconomics should be changed to allow for the direct effects of international price arbitrage. Whether monetarists or Keynesians or rational expectationists, economists should begin thinking and measuring in global terms" (p. 127). Economists have consistently thought and measured in global terms in examining the determinants of the quantity of money in a country, both for periods of fixed exchange rates and of dirty floating-and McCloskey and Zecher cite no examples to the contrary. In respect of the second proposition, the money supply Milton Friedman, Nobel laureate, is senior research fellow at the Hoover Institution, Stanford University, and Paul Snowden Russell Distinguished Service Emeritus professor of economics at the University of Chicago.

158

Donald N. McCloskey and J. Richard Zecher

may be endogenous after a sufficient interval, yet not in the short run. And whether endogenous or exogenous, the domestic money supply is absorbed primarily within the domestic economy; it is highly fruitful to examine the process whereby that occurs and the relation within a country between changes in the money supply and changes in other variables. There is no such sharp dichotomy between "closed economy" and "global" thinking and measuring as the straw man set up by McCloskey and Zecher. Some findings from Anna Schwartz's and my book Monetary Trends are relevant to this subject. We calculated correlations between the United States and the United Kingdom in the rates of change of various variables between cycle phases for almost a century. We pointed out, and this is strictly in accord with the McCloskey-Zecher view, that the correlation between prices in Britain and the United States is closer than the correlation between any other two magnitudes. It is closer than the correlation between money supplies in the two countries, closer than the correlation between income in the two countries, real or nominal. We stressed that that result is strictly consistent with what they call our Martian view of the economy, and indeed dema·nded by it. Because, we said, in a world of fixed exchange rates, the money supplies in the different countries have to accommodate themselves in such a way as to be consistent with equality of prices and goods among countries. And therefore, the relation between the quantity of money in the two countries would be expected to be less close than between prices because changes in the demand curve for money within an individual country must be accommodated by changes in money rather than in prices (Friedman and Schwartz 1982, pp. 310-15). Let me turn to a couple of other points. First, McCloskey and Zecher assert that "the turbulence of the 1920s and 1930s, which is said to have loosened the economic ties among nations, appears not to have done so" (p. 139). One comparison in our book supports a very different conclusion. We started with the Kravis, Heston, and Summers estimate for the purchasing-power-parity exchange rate between the United States and the United Kingdom in 1970. We used price indexes in the two countries to extrapolate the purchasing-power exchange rate annually from 1870 to 1970. We then calculated the ratio of the purchasing-power-parity exchange rate to the market exchange rate. The resulting chart is fascinating (Friedman and Schwartz 1982, chart 6.5, p. 291). Before about 1932, the ratio of the purchasing-power-parity exchange rate to the market exchange rate varied within a range of plus or minus 10 percent. All of us would say that is a fairly close relationship to purchasing-power parity. After 1931, the range is between minus onequarter and plus one-third-an enormously wider range. There is no

159

The Success of Purchasing-Power Parity

doubt that a comment by Jacob Frenkel is right: Hume traveling in the 1950s would have found more deviations from purchasing-power parity than he would have found in the 1780s. Exchange controls, tariffs, other impediments to trade were far more important in dividing the world than improvements in communication and other technologies were in uniting it. I don't see that notion is any contradiction to the purchasing-powerparity theory of exchange rates. It is simply a consequence of the fact that there has been an enormous increase in barriers to trade among countries since 1931. I want now to discuss two particular episodes for which McCloskey and Zecher take Anna Schwartz and me to task for our analysis in A Monetary History (1963). The first example is resumption in 1879. They quote our statement that "it would be hard to find a much neater example in history of the classical gold-standard mechanism in operation" (p. 99). Their look at that episode on the basis of monthly data is interesting and most welcome, but on closer examination it does not, contrary to their claims, contradict our interpretation of the episode. McCloskey and Zecher compare price rises to inflows of gold, concluding, "In the monthly statistics ... there is no tendency for price rises to follow inflows of gold . . . ; if anything, there is a slight tendency for price rises to precede inflows of gold, as they would if arbitrage were shortcutting the mechanism" (p. 126). Their comparison is the wrong one for determining whether prices were reacting to arbitrage rather than reflecting changes in the quantity of money. For that purpose the relevant comparison is with the quantity of money. Gold flows are relevant only as a proxy for the quantity of money. If we compare price rises with changes in the quantity of money directly, a very different picture emerges than McCloskey and Zecher draw (see table C2.1). Our basic estimates of the quantity of money for this period are for semiannual dates, February and August. Resumption took effect on 1 January 1879. From August 1878 to February 1879, the money supply declined a trifle, continuing a decline that had begun in 1875 in final preparation for resumption. From February 1879 to August 1879, the money supply rose sharply, according to our estimates, by 15 percent. The Warren-Pearson monthly wholesale price index fell in the first half of 1879, reflecting the earlier decline in the money stock. It started its sharp rise in September 1879, or at least seven months later than the money supply.1 As to gold, the total stock of gold, as well as gold held by the Treasury, had been rising since 1877 as part of the preparation for resumption. But it had been rising at the expense of other components of high-powered money, which actually fell slightly. However, the decline in the money stock before 1879 had been due primarily to a decline in the depositcurrency ratio and the deposit-reserve ratio. After successful resumption,

160

Donald N. McCloskey and J. Richard Zecher

both ratios rose, which enabled the stock of money to rise despite no initial increase in gold flows. The large step-up in gold inflows in the fall of 1879, to which McCloskey and Zecher call attention, was mostly absorbed in raising the fraction of high-powered money in the form of gold rather than in speeding up monetary growth. Table C2.1 shows these developments in more detail. On the basis of this reexamination, I am inclined to repeat the statement in our Monetary History, "It would be hard to find a much neater example in history of the classical gold standard mechanism in operation" (italics added), at least on a sophisticated interpretation of both the gold standard mechanism and the historical data. At any rate, the gold standard mechanism, as I understand it, has always incorporated a variety of channels of adjustment, subject to different lags. Any result one wants can be gotten, depending on the relative speed of adjustment of the various channels. It is an important scientific question to try to identify and isolate these relative speeds of adjustment. I believe that McCloskey and Zecher make a real contribution in examining aspects of that issue. The second episode is the behavior of prices after the u.s. went off gold in 1933. The figure 2.3 shows a close parallelism between the weekly movements of wholesale prices and the exchange rate. However, the different scales used for the price index and the exchange rate in the chart give a misleading impression? For example, from April 1933 to July 1933, wholesale prices rose less than a sixth, the exchange rate by nearly a half, yet the total impression from their chart is that prices rose more sharply. Logarithmic scales would give a more accurate picture and make clear Table C2.1

Aug. Feb. Aug. Feb. Aug.

1878 1879 1879 1880 1880

Relations between U.S. Prices, Money, High-Powered Money, and Gold, 1878-80 Wholesale Prices (P) (1)

Money Stock (M) (2)

Highpowered Money (HPM) (3)

90 88 86 105 97

1.57 1.55 1.78 1.94 2.05

0.767 0.752 0.815 0.897 0.972

Gold Stock (G)

MIHPM

(4)

(5)

GIHPM (6)

0.182 0.198 0.219 0.302 0.378

2.05 2.06 2.18 2.16 2.11

0.24 0.26 0.27 0.34 0.34

Sources: Col. (1) U.S. Bureau of the Census 1949, app. 24, p. 344; col. (2) Friedman and Schwartz 1970, p. 5; col. (3) Friedman and Schwartz 1963, p. 799; col. (4) ibid., notes to table A-I, p. 723 and table A-3, p. 765. Notes: Col. (1) Warren-Pearson index of wholesale prices (191~14 = 100); col. (2) currency held by the public plus adjusted deposits of commercial banks (billions of dollars); col. (3) in billions of dollars; col. (4) in billions of dollars.

161

The Success of Purchasing-Power Parity

how much narrower the relative movement in the wholesale price index was than in the exchange rate. In any event, as McCloskey and Zecher note, we pointed out in A Monetary History that there was a direct effect of devaluation on prices. However, the existence of a direct effect on wholesale prices is not incompatible with the existence of many other prices, as Moe Abramovitz has remarked, such as non-tradable-goods prices, that did not respond immediately or responded to different forces. An index of rents paid plotted against the exchange rate would not give the same result. An index of wages would not give the same result. It may be worth quoting what we actually said on the issue, especially in view of the McCloskey-Zecher comment on a quotation from the Monetary History that "this is an unusual line of reasoning for such crusaders against mixing up the determination of relative and absolute prices" (p. 141).3 Here is what we said: "The aim of the gold policy was to raise the price level of commodities, particularly farm products and raw materials.... Most farm products and raw materials exported by the United States had a world market in which this country ... was seldom dominant.... Hence, the decline in the foreign exchange value of the dollar meant a roughly proportional rise in the dollar price of such commodities, which is, of course, what did happen to the dollar prices of cotton, petroleum products, leaf tobacco, wheat and similar items" (Friedman and Schwartz 1963, pp. 465-66). Thirty-odd pages later, after noting that the rise in the implicit price index from 1933 to 1937 was of roughly the same order of magnitude as in 1879-82 and 1896- or 1897-99, but in wholesale prices decidedly larger: "What accounts for the greater rise in wholesale prices in 1933-37, despite a probably higher fraction of the labor force unemployed and of physical capacity unutilized than in the two earlier expansions? One factor, already mentioned, was devaluation with its differential effect on wholesale prices" (p. 498). This was followed by the passage McCloskey and Zecher quote in which we referred to "the implicit measures to raise prices and wages undertaken with government encouragement and assistance" (p. 498). Contrary to the impression McCloskey and Zecher give, we did not try to assess the relative importance of various factors in explaining the rise in prices from 1933 to 1934---the period to which they limit their chart and discussion. On the contrary, we explicitly cited these measures as helping to explain the "rise in wholesale prices in 1933-37." The wholesale price index continued to rise after its initial sharp rise in 1933 and did not reach its peak until mid-1937 when it was 47 percent above its low point in February 1933 and 28 percent above its level in July 1933. Hence there was ample time for the factors we referred to to play their part after the enactment of the legislation we listed.

162

Donald N. McCloskey and J. Richard Zecher

Note finally that McCloskey and Zecher have faced up in this paper to the problem of floating or flexible exchange rates to only a very minimal extent. This paper is written primarily for a world of fixed exchange rates, and indeed, fixed exchange rates with nonchangeable barriers to trade. To be applicable to the current world, those elements must be added. Notes 1. This paragraph and the next two were added after the conference in revising my comments for publication. 2. This sentence and the next two were added in revising these comments for publication. 3. This paragraph and the next three were added in revising these comments for publication.

References Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. - - - . 1970. Monetary statistics of the United States. New York: National Bureau of Economic Research. - - - . 1982. Monetary trends in the United States and the United Kingdom, 1867-1975. Chicago: University of Chicago Press. Keynes, J. M. [1923] 1971. The applied theory of money: A treatise on money. Vol. 4 of The collected writings of John Maynard Keynes. London: Macmillan and New York: Cambridge University Press for the Royal Economic. Society.

General Discussion ABRAMOVITZ suggested another way of posing the question that lies at the heart of the McCloskey-Zecher paper: How can tradable-goods prices remain equal to one another or move in similar ways in different countries without destabilizing the gold standard, in the face of differing national and sectoral rates of productivity growth? Abramovitz pointed out that McCloskey and Zecher offer one possible adjustment mechanism. In contrast, the traditional specie-flow mechanism, involving changes in prices and nominal wages in different countries, offers a rather different mechanism. MCCAULEY asked McCloskey and Zecher to justify the leap from purchasing-power parity, however defined, to the assertion that monetary policy cannot alter prices. This assertion appears in weak form (p. 146)-that American prices are not at all constrained by the forces of

163

The Success of Purchasing-Power Parity

arbitrage-and in rather stronger form (p. 128)-that American price levels are exogenous to American monetary policy. Surely as the share of world output produced by a single country approaches unity, an individual country's monetary policy becomes capable of raising prices. McCauley argued that these questions imply technical objections to McCloskey and Zecher's analysis. While McCloskey and Zecher analyze the residuals from their purchasing-power-parity equation, their procedure presumes that U. S. monetary policy is incapable of influencing price levels in the rest of the world. ZECHER responded that the Zecher-McCloskey paper effectively incorporates McCauley's point. In response to Friedman, Zecher disagreed with his statement that McCloskey and Zecher fail to take into account the difference between periods of flexible and fixed exchange rates. Under flexible exchange rates, McCloskey and Zecher's assertion is not that a country cannot affect its own price level or rate of inflation, but rather that relative prices, or the deviation from purchasing-power parity, is constrained by arbitrage in commodity and other markets. Thus, their analysis is capable of dealing with flexible-exchange-rate periods. FRIEDMAN restated the central point of his argument: suppose a country's money supply is endogenous, determined by the outside world. One can still examine the relationship between the quantity of money in that country (call it Illinois) and the price level and nominal income in Illinois. The change in the quantity of money, however produced, has effects internal to Illinois. MCCLOSKEY responded that all monetarists share a belief in a stable demand for money. But to go from the presumption that money demand is stable to the assertion that money supply in Illinois determines prices and interest rates in Illinois is a large jump. FRENKEL raised the question of what exactly McCloskey and Zecher mean when they speak of purchasing-power parity? He suggested that McCloskey and Zecher essentially mean the law of one price. Purchasingpower parity· is enforced by the mechanism of commodity and asset arbitrage, converting the whole discussion of purchasing-power parity into a discussion of financial flows and profit opportunities. Frenkel suggested posing a very different question, which was the original question underlying the development of purchasing-power parity: How can one determine an appropriate exchange rate for the period following a serious market dislocation? How much information can be obtained from aggregate price indexes? One issue is which "aggregate" to look at. Needless to say, this question is based on the presumption that aggregates provide useful information for determining equilibrium exchange rates; changes in relative prices call this view into question. The crucial question, therefore, is under what conditions it is likely that aggregate price indexes will provide useful information about equilib-

164

Donald N. McCloskey and J. Richard Zecher

rium exchange rates? It may be very important to know whether the shocks to the system originate from the real side or from the monetary side. In the wake of monetary shocks, Frenkel argued, it is advisable to focus on the aggregate that best represents monetary conditions. This view suggests looking at the price indexes that cover the broader domain of goods and services, which is what Cassel had in mind. In his view, purchasing-power parity was not a theory of individual prices but of price indexes because it was intended as a measure of the monetary conditions. His view is fundamentally different from the view that foreign exchange rates have nothing to do with aggregate price levels but only with individual commodity prices. Frenkel pointed out that the original view of purchasing-power parity refers to the ratio of rates of inflation of purchasing power, where inflation is to be understood as inflation of the quantity of money. These concepts are completely divorced from individual commodity prices per see Frenkel made a number of points concerning McCloskey and Zecher's econometric results. One interesting exercise in their paper asks whether deviations from purchasing-power parity have real effects? McCloskey and Zecher choose to concentrate on the trade balance; they ask if there is a visible, statistically significant relationship between apparent deviations from purchasing-power parity and the trade balance. They find no such evidence. Frenkel hesitated to infer from these results any particular conclusion about purchasing-power parity, since it is not clear that changes in relative prices should always have a particular trade-balance effect. If one thinks of the current account as the difference between income and spending, then there is a determinate theory that links changes in the relative prices of commodities to the aggregate saving ratio and hence to the current account. There is a determinate link between the terms of trade and the current account-the so-called Laursen-Metzler effect. However, under a variety of plausible circumstances, the LaursenMetzler effect might not hold. ZECHER responded to Frenkel by emphasizing the importance of questioning the extent to which national markets are integrated. The problem with many recent empirical studies, he suggested, is that after concluding that purchasing-power parity fails, subsequent theorizing simply neglects the rest of the world. It is important to attempt to define criteria that permit one to label markets as more or less integrated. PIPPENGER pointed out that several economists have analyzed the residuals from purchasing-power parity calculations. He himself had examined deviations from purchasing-power parity using annual data, going back in some cases to the 1870s. The evidence indicates that deviations follow a random walk. There appears to be no tendency for

165

The Success of Purchasing-Power Parity

relative prices to return to any normal long-run level. This random-walk property holds even for the regional price indexes within the United States, which is indeed curious. Of course, this result is inconsistent with the continuous maintenance of purchasing-power parity, because the result permits exchange rates to drift in any direction. It also raises some interesting questions about the proper way of interpreting time-series data. One possible interpretation is that purchasing-power parity simply fails to hold. Another, suggested by Richard Roll, is that the randomwalk property is evidence of efficient commodity markets. This latter interpretation would suggest that Zecher and McCloskey are right, but for a different reason. What we may be seeing is simply the fact that price indexes for different countries are made up of different commodity bundles. Many problems must be sorted out before we can distinguish between the traditional view and the Zecher-McCloskey view. BRUNNER argued that one may wish to distinguish a shorter run, perhaps up to one-and-a-half or two years, over which the money stock is exogenous, and a longer run over which it is endogenous. Support for this distinction can be found in the history of the Italian monetary affairs in the 1960s. At that time there were a number of one-and-a-half- or two-year periods when macroeconomic accelerations were fueled essentially by the domestic-credit component of the monetary base. Such credit creation was able to alter the money supply for one-and-a-half or two years with accompanying adjustments in prices. Only thereafter was the balance of payments affected. Such lagged responses are quite consistent with some of the adjustment mechanisms sketched here. DORNBUSCH suggested that McCloskey and Zecher had provided insufficient room in their framework for the considerations emphasized by Abramovitz. When an economy is growing and the composition of activity is changing over time, simple tests of purchasing-power parity will be biased. For example, anyone who tests the purchasing-power-parity hypothesis for the last twenty years will find that real exchange rates in manufacturing, of the United States as well as of any other industrialized country, are well explained by differentials in sectoral growth rates but not by national rates of price inflation. Therefore, serious tests of purchasing-power parity must incorporate time trends or other variables designed to account for differentials in sectoral growth rates and other real changes. Dornbusch also drew attention to the financial research of the last three years, which demonstrates that one cannot reject the hypothesis that the stock market devia~es for long periods from market fundamentals. On purely statistical grounds, even fifty years of Dow-Jones data are incapable of rejecting, at a .99, a .95 or even a .90 level of confidence, the hypothesis that the stock market is driven for long periods by fads and fashions.

166

Donald N. McCloskey and J. Richard Zecher

MCGOULDRICK asked whether empirical tests of purchasing-power parity should include in the price indexes not only the prices of goods but the prices of securities. MCCLOSKEY agreed that in principle the capital market should be considered in studies of purchasing-power parity. He pointed out that Lipsey takes the McCloskey-Zecher analysis of purchasing-power parity as an empirical test and argues that it is not in fact properly interpreted as a statistical test of a hypothesis. Rather, the analysis is better thought of as a way of looking at particular episodes which might persuade people of the plausibility of a particular view of how markets function. McCloskey suggested that Friedman was in substantial agreement with the authors' main point. Friedman concedes that under fixed exchange rates the money supply of Illinois does not determine prices and interest rates in Illinois. That was McCloskey and Zecher's main point. FRIEDMAN referred back to some of the work in his recent book written with Anna Schwartz. Friedman and Schwartz discovered an appreciable difference in the relation between interest rates in the United States and United Kingdom before 1896 and after 1896. Interest rates in the United States before 1896 are much higher relative to the British interest rates than after 1896. In other words, before-1896 interest rates act as if there was widespread anticipation of a depreciation of the U.S. dollar. After 1896, they act as if there was widespread expectation of an appreciation of the U. S. dollar. This behavior bears on the question of whether there can be significant deviations in prices and interest rates in various countries over substantial periods of time. FRATIANNI suggested parallels with the Italian experience. In Italy, he argued, systematic deviations from purchasing-power parity are matched by deviations from interest-rate parity. Also, the way in which governments finance budget deficits should be accounted for in regressions. GREGORY reported an experiment conducted by himself and colleagues (Baltagi and Sailors) at the University of Houston. They estimated a three-equation model, pooling thirty-four cross-section time-series observations for seven countries (France, Germany, Japan, Russia, Sweden, the United Kingdom, and the United States), to investigate the working of the gold standard. The model included a money-demand equation, a balance-of-payments equation, and a money-supply equation, with all variables expressed as first differences. The pooled regression results suggest that the classical gold standard was a fairly simple system, not the complex one that Bloomfield reported. According to the model, if supply shocks were to set off domestic inflation, domestic inflation would then cause a worsening of the external balance, and a worsening of the external balance would cause the domestic money supply to drop. While the pooled model yielded these statistically significant and plausible results, the individual-country time-series regressions

167

The Success of Purchasing-Power Parity

yielded generally insignificant coefficients. Gregory suggested that pooled data might serve as an avenue for expanding the empirical data set in exploring the working of the classical gold standard. THOMAS raised a major question with respect to the model that Gregory described. An outstanding characteristic of the period 1880-1913 was the unique role of Great Britain in the Atlantic economy, as major capital exporter and center of what was virtually a sterling standard; her status was different in kind from that of the borrowing countries of the periphery. In 1880 the United Kingdom was responsible for 41 percent of world exports of manufactured goods as against 3 percent for the United States; even as late as 1889 the U.K. share of world exports of capital goods was 44 percent as compared with the U.S. proportion of 23 percent. A unique feature of the growth process was the fact that the long swings in capital formation, productivity, and real income in the center country, Great Britain, were inverse to the corresponding long swings in the borrowing countries of the periphery. The voluminous evidence confirming the validity of these inverse long swings was summarized by Arthur Bloomfield in his well-known Patterns of Fluctuation in International Investment before 1914 (1968). No account of the working of the international gold standard can afford to neglect these special features of the pre-1913 period. Five of the seven countries on which the pooled model was testedGermany, France, Sweden, Russia, and Japan-were not Atlanticoriented either in trade or foreign lending, so that only two-Great Britain and the United States-reflected the special characteristics of the pre-1913 international economy, namely, the center-periphery interaction. However, the way the model was specified made it impossible to pick up this interaction. The pooled results have drowned the peculiarity of Britain's interaction with the United States in general averages for the seven countries. The working of the international gold standard between the "regions" of the nineteenth-century Atlantic economy has a close resemblance to the working of the internal gold standard between the regions of the United States. Under this internal standard the ease of adjustment was greatly facilitated by the existence of two fundamental conditions-free interregional migration of labor and the transfer of Treasury funds into weak regions. Among the most important reasons why the international gold standard worked fairly smoothly were, first, the high degree of international mobility of labor and, second, the fact that Britain, the dominant creditor, with a high propensity to import, was always putting money back into international circulation, either through a substantial upswing in imports or through a substantial upswing in foreign lending. Thomas also commented on the McCloskey and Zecher paper. Proceeding from the analogy between the international and the national

168

Donald N. McCloskey and J. Richard Zecher

economy, McCloskey and Zecher emphasize what they regard as the necessary corollaries-unified prices of products, assets, and labor across national boundaries. Hence the monetary theory, which is an equilibrium model; but the theory overlooks some awkward facts. One has to recognize the nature of the growth process in the pre-1913 Atlantic economy, particularly the inverse relation between investment upsurges in the center country and in the overseas country of new settlement. Export capacity in a given phase of the long swing was a function of the infrastructure investment in that country in the previous swing. There was a long-run symbiotic relationship, but it necessarily entailed opposite movements at the center and the periphery, and serious disequilibria when the peaks of the long swing were reached and the Bank of England had to protect its reserve. This process would occur particularly when under-effected transfer was experienced. The U.S. trade balance determined the gold flow, and the gold flow determined the rate of growth of the money supply. There is no basis for the notion that investment upswings, by generating excess demands, attracted net capital inflows that more than offset the unfavorable trade balance, thereby inducing gold inflows. Gold inflow, and as a consequence the money supply, rose most rapidly in the phases of the long swing when U.S. exports were surging upwards and infrastructure expenditure and imports were in a downswing. Simultaneously Britain was having an upswing in home investment, her exports as a proportion of imports were falling, and there was an external flow of gold from the Bank of England. When it was the turn of the United States to have its upswing in investment, her trade balance deteriorated, gold flowed out, and the rate of growth of the money stock fell. See what happened during the 1890s. Between June 1892 and June 1896 there was an absolute fall in the U.S. money stock, the first such decline since the 1870s, whereas the Bank of England's reserve increased spectacularly from £15 million to no less than £49 million. When the United States was struggling desperately to stay on the gold standard, Britain was enjoying such a surfeit of liquidity that the market rate of discount was below 1 percent. In the second half of the 1890s the reverse happened as a result of the massive upsurge in United States exports in relation to imports coinciding with the opposite in Britain. The Bank of England reserve as a proportion of liabilities fell almost as fast as it had risen, while the money stock of the United States went up by 52 percent. The Old Lady of Threadneedle Street was not managing the international gold standard: she was just minding her own business and doing it on an investment in gold stocks inexcusably small in relation to her responsibilities. Her status as central bank of the center country endowed her with clout. McCloskey and Zecher are scornful of Keynes's description of the Old Lady as "conductor of the international orchestra"; they

169

The Success of Purchasing-Power Parity

regard her as "no more than the second violinist, not to say the triangle player, in the world orchestra" ("How the Gold Standard Worked, 1880-1913," in The Monetary Approach to the Balance of Payments, ed. J. Frenkel and H. G. Johnson. [London: Allen and Unwin, 1976], pp. 358-59). As a superb understatement that must constitute something of a record. EICHENGREEN elaborated upon one of Brinley Thomas's points. The purpose of pooling national time series and of attempting international comparisons is to extract as much information as possible from historical data. This approach is predicated upon the assumption that the structural relationships under consideration are identical across countries. In the case of monetary relations under the classical gold standard, there is considerable historical evidence of the existence of important structural asymmetries that would call into question the validity of this assumption. Eichengreen drew attention to the work of Triffin and others that pointed to the unique degree of market power exercised by the Bank of England under the classical gold standard and to asymmetries in the impact that changes in the monetary conditions in different countries had on the balance of payments of the countries participating in the system. For example, changes in monetary conditions in Britain appear to have had a much more powerful impact on short-term capital flows than did comparable changes in monetary conditions abroad. As Triffin suggests, the Bank of England had an ability to influence international gold flows unrivaled by other central banks. Asymmetries of this sort are not taken into account in Gregory's analysis. ABRAMOVITZ pointed out that there is no obvious connection between long swings, such as fifteen-to-twenty-year Kuznets cycles, and international gold movements. This is not surprising, since many factors can substitute for the actual movements of gold: the growth of high-powered money from domestic sources, changes in high-powered money multipliers, and changes in the income velocity of money. What, then, produced the long swings in high-powered money that parelleled so closely the long swings in the growth rate of real output? Abramovitz suggested focusing on the growth rate of the sum of exports and capital imports. That sum traces long swings that parallel the long swings in the growth rate of nominal and real incomes in the United States. This line of inquiry suggests a further question: are exogenous changes in the growth rate of the sum of exports plus capital imports driving the growth rate of the money supply and of nominal income? Or is the growth rate of nominal income determined by independent changes in real output and world prices to which the money supply of a country must adjust? The latter is recognizably the view of the monetary approach to the balance of payments, which is arguably the right view when one has in

170

Donald N. McCloskey and J. Richard Zecher

mind very long periods of time. In shorter periods, however, and even in the long swings that run across ordinary business cycles, the answer is much less clear. In the long swings of income in the United States, there appears to have been. an interaction between the sum of commodity exports and capital imports which together constitute the positive elements underlying the balance of payments and the real-income changes with which they are associated. Independent movements of commodity exports or capital imports stimulated change in nominal income and in real income as well. The changes in real income, in turn, generated further increases in capital imports. Within limits, there was a selfsustaining cumulative process. Abramovitz noted that the central point, raised by Brinley Thomas, concerns the inverse pattern of long swings in Britain and the United States. Abramovitz maintained that there were no inverse swings in real output or in nominal income in the aggregate in the United States and United Kingdom. In point of fact, the long upswings in the United States were matched by surges of capital export from Britain and capital import into the United States. Similarly, surges of exports from Britain were matched by surges of imports into the United States, and declines in home investment in Britain were matched by rises in home investment on the other side of the Atlantic. In Britain fluctuations in commodity and capital export offset one another and left the British economy growing smoothly over the business cycle, in contrast to the United States where capital imports and exports were not matched so closely bf surges and declines in commodity imports and exports. These observations lead to a further question: Why did the long swings come to an end? Why did they result in the United States in serious depressions that culminated each of these episodes, while in Great Britain there were no comparable breakdowns that would have produced, had they occurred, the appearance of long swings in aggregate output as in the United States? One reason was the difference between the banking systems of the two countries. At intervals, the United States suffered banking and financial panics far more violent than those to which Britain was subject. A more severe impact on money and real income was felt in the United States. The two countries, however, differed in other respects as well, and the matter deserves a lot more study.

PART

II.

Technical Procedures: Rules of the Game

3

The Bank of England and the Rules of the Game under the International Gold Standard: New Evidence John Dutton

3.1

Introduction

This paper is an investigation of the Bank of England's actions under the gold standard in the decades immediately preceding World War I. In particular, I use monthly data on Bank reserves, domestic activity, price changes, and gold inflows from the late 1880s to mid-1914 to determine which variables the Bank reacted to. The years from 1870 or 1880 until 1914 are frequently regarded as a halcyon period in international monetary relations. Several financial crises occurred (in 189O-the Baring crisis-and in 1893 and 1907), but throughout the period the major central banks were able to maintain gold convertibility of their currencies. Had war not intervened in 1914, the system might have operated reasonably well for decades longer. At the center of the system was London (Lindert 1969 describes its importance), and at the center of the London financial community was the Bank of England. Actually, in 1914 the Bank was still a private profit-making institution; however, it had for a century and more increasingly taken on quasi-official functions. After Walter Bagehot's publication of Lombard Street in 1873, the Bank's position as central bank had become widely recognized and accepted. Because of London's key position in the international financial community, the Bank at that time was clearly the foremost central bank of the world-a role that has special interest for anyone studying the international financial system. John Dutton is assistant professor of economics at North Carolina State University, Raleigh, North Carolina. The author wishes to thank Michael David Bordo, Paul McGouldrick, John Pippenger, Anna Schwartz, Richard Sylla, and the numbers of a North Carolina State University International Trade Workshop for helpful comments and David Dickey for econometrics advice. They are absolved of all responsibility for the final result.

173

174

John Dutton

A salient feature of that international financial system is its apparent stability. Bloomfield (1959) points out that exchange rates among countries on gold were essentially unchanging; devaluations were a rarity. Implicitly, balance-of-payments problems were not the disrupting influence they came to be under the Bretton Woods system. Admittedly some economies may have experienced greater instability with respect to certain measures of prices and output than after World War II (see Bordo 1981 for statistics for the United States and the United Kingdom); nonetheless, the international aspects of the system were stable. Analysts of the pre-1914 era have stressed two explanations of international stability. In the short run, if trade imbalances developed (caused perhaps by exogenous shocks), capital would flow in the opposite direction. This capital counterflow would take place because of the effects of gold flows on money supplies and interest rates. In the United Kingdom, according to the descriptive literature (e.g., Clapham 1970; Sayers 1976), such interest-rate effects were produced by central-bank authorities. Of course, changed flows of capital in response to interest-rate changes were primarily stock-adjustment processes and therefore inherently temporary. For this reason capital flows are not a satisfactory explanation of long-term flow adjustment. Once any capital stock adjustment had taken place and capital stopped moving, the outward gold flows caused by the original trade imbalance would reappear. A second explanation of long-term adjustment was the Hume pricespecie-flow mechanism. According to that explanation, internationalpayments imbalances resulted in money-supply changes in the countries involved. These in turn caused prices to change (upward in the country receiving gold, downward in the country losing gold), and these price changes altered the international flow of goods in such a way as to eliminate the initial trade imbalance. Although it does not appear in analyses of international adjustment before World War I, after the war the phrase "rules of the gold standard game" was applied to a prescription for a specific central-bank role in the adjustment process. According to the "rules," central banks had the important role of facilitating international-payments adjustment, either by reinforcing the effects of payments imbalances on the domestic economy so as to speed the adjustment process, or at least by not hindering those effects. According to one definition of the rules articulated by Bloomfield (1959 pp. 47-48): A discount rate and credit policy ... was supposed . . . to have the effect of increasing central·bank holdings of domestic income-earning assets when holdings of external reserves rose, and of reducing domestic assets when reserves fell. In this way, the effect of changes in central bank reserve holdings on the domestic credit base would be magnified by central bank action.

175

The Bank of England and the Rules of the Game

Bloomfield mentions that Nurkse (1944) applied this definition to the period between the world wars and found that during that period central banks in general did not follow this form of the rules. Bloomfield himself applied the definition to annual data in the 1880-1914 period. He computed the number of times international and domestic interest-earning assets of each central bank moved together, indicating rules compliance, and moved in opposite directions, indicating rules violation. He found for all the observations covered a compliance rate of 34 percent and a violation rate of 60 percent, with the remainder being cases of no change. The comparable figures for the Bank of England in Bloomfield's study were 47 percent and 50 percent. Pippenger's (1974) study uses a definition of the rules somewhat like Bloomfield's. He regresses Bank of England monetary liabilities on the Bank's stock of gold during the 1890-1908 period and finds a positive and significant relationship when data for periods of a quarter and longer are used, though not when the data are monthly. Pippenger's conclusion is that the Bank responded to gold flows by changing the money supply in the medium to long run but not in the short run. Such long-run behavior, if the response were sufficient, would be in keeping with the rules. Pippenger's coefficient for annual data indicates a bit less than a one-forone transmission. Bloomfield notes in his description of the rules that the definition quoted above is by no means the only one. Another possible definition, which he does not explicitly mention, is that central banks refrain from countercyclical domestic policy. Under the rules, central banks were obligated to maintain the convertibility of their currencies. Given the limited nature of their policy tools, that obligation would have left them little capacity to pursue domestic stabilization. In addition, pursuit of the stabilization goal might well have interrupted the process of long-term adjustment to eliminate payments imbalances. Qualitative discussions provide mixed answers to the question of the incidence of domestic countercyclical policy. Bloomfield (1959, p. 24) himself says: "the view . . . of central banking policy as a means of facilitating the achievement and maintenance of reasonable stability in the level of economic activity and of prices was scarcely thought about before 1914, and certainly not accepted, as a formal objective of policy," but then he goes on to note increasing awareness and sensitivity on the part of central banks to the domestic effects of their actions. Sayers (1976) also notes some sensitivity of the Bank of England to the effects of its policy on business conditions. Apparently, critics of the effects of Bank policy on the domestic economy made their views known (see Dornbusch and Frenkel, this volume). Mints (1945, pp. 188-89) cites several writers of the pre-1914 period who decried those effects. Palgrave, writing in 1903, complains: "Great instability in the rate of dis-

176

John Dutton

count is a very prejudicial thing to the interests of commerce," and favors minimizing the transmission of disturbances from abroad to the domestic economy ([1903] 1968, p. viii). Whether the Bank actually pursued countercyclical policies or not, certainly the possibility of doing so must have been evident to its directors. One problem in testing for Bank countercyclical policy is a possible strong negative relationship between reserves and domestic activity. Ford (1962, p. 21) points out that at cyclical peaks high import demand might have coincided with high capital exports, with consequent demands on central-bank reserves. If low reserves led to high Bank discount rates, then the Bank might appear to be following a countercyclical policy without actually doing so. Ford apparently believes the Bank-ratedomestic-activity relationship to be such an indirect one, for he states, "the Bank did not pursue a conscious contra-cyclical policy in its use of Bank Rate" (1962, p. 34). Goodhart (1972, p. 199) presents evidence to support the negative reserve-activity relationship. A regression he reports of the balance-of-trade surplus on an activity variable for the 1890s and early 1900s yielded a significant negative coefficient. A possible test for separate effects of reserves, domestic activity, and inflation on Bank policy is to include all three in the equations estimated. Dutton (1978) reports such a regression on Bank rate using annual data for 1862-1913. Despite the presence of reserves as a control in the equation, domestic activity, as measured by the rate of unemployment, had a significant positive effect on Bank rate. Another problem in the interpretation of the effects of domestic activity involves the Bank's profits. If high activity led to high market interest rates, the Bank might have raised its own discount rate as a way of enhancing its profits. Such profit-motivated behavior might appear to be countercyclical policy. It would be hard to discriminate between profitmotivated and purely countercyclical Bank-rate policy. Most of the descriptive literature takes for granted that profit was secondary to other Bank policies. In this paper I assume that it was. Actually, if Bank policy was countercyclical, whether for profitability reasons or otherwise, the Bank would have broken the rules. The need to discriminate between profitability and stabilization as motives may not be important in testing for violations of the rules. In the present study, I use monthly data available for several series from the late 1880s to 1914 to find out how well Bank actions fit the rules of the game. In doing so, I test for both countercyclical policy on the part of the Bank (ignoring profit motives) and for relationships between Bank holdings of non-interest-bearing reserves and interest-bearing domestic assets. I expect to find, if the Bank followed the rules, that it did not react to domestic activity and inflation in its conduct of policy. I also expect to find, if the Bank followed the rules, that reserve changes and Bank

177

The Bank of England and the Rules of the Game

interest-bearing assets moved together, indicating that the Bank did not sterilize the effects of reserve changes on the money supply. 3.2 3.2.1

Specific Variables Considered Policy Tools

The Bank of England's chief instrument of monetary control was its discount rate, known as "Bank rate." According to Sayers (1976, p. 23), "the accepted doctrine both inside and outside the Bank was that its most important action was the fixing of Bank Rate." Of course, for Bank rate to have an effect, it had to influence market rates of interest. Accordingly, the Bank had to ensure that capital markets in London were dependent on it for at least part of their funds. In fact, many commercial concerns borrowed regularly from the Bank at Bank rate, or at rates closely related to Bank rate. Consequently, during much of the period, that rate constituted the opportunity cost of funds at the margin for market participants. A great deal of the energies of those governing the Bank was expended in making sure at certain key times that Bank rate did constitute that opportunity cost. "Making Bank rate effective" generally involved the use of other more limited tools of the Bank to back up Bank rate. Altering the Bank's holdings of interest-earning assets-the tool used frequently to "make Bank rate effective"-ean be thought of as an early form of open-market operations. To force market participants to borrow at Bank rate, the Bank would in one way or another gain command of additional funds in the market. It did so at times by selling securities, at other times by borrowing against its securities holdings, and at still other times by selling consols spot and buying them forward. All of these procedures removed funds from the market and eventually forced customers to borrow from the Bank. How common these operations were is difficult to say. Clapham notes their use as far back as the midnineteenth century and earlier; he goes on to say that "in the seventeen years from 1873 to 1890 there are only four in which no market borrowing is done, and in several the borrowings are repeated and complex" ([1944] 1970, pp. 295-98). Sayers (1976, pp. 37-43), on the other hand, stresses the relative infrequency of these operations before 1890, though he describes them as used with increasing frequency after that time (see also Morgan 1965 for a description of the practices involved). A third instrument, in addition to Bank rate and altering the Bank's portfolio, was use of the so-called gold devices. The Bank was required by law to purchase gold with notes and to redeem notes with gold sovereigns on certain set terms of exchange. However, the Bank was free to alter those terms somewhat in favor of persons bringing gold for

178

John Dutton

exchange. It was also free to force its customers to follow the law's conditions more or less exactly in presenting or seeking gold. Thus the Bank could draw more gold by making it more advantageous for customers to bring gold and more costly for them to remove it from the Bank. These practices-lumped together under the name "gold devices"-had the effect of increasing the spread between the gold points, the rates of exchange at which gold could profitably be imported or exported. Sayers (1953, 1976) describes these gold devices in some detail. Their use was in a sense a violation of the rules, since the devices interfered with the free convertibility of gold at clearly specified rates of exchange. They were usually used to retain or attract gold without resorting to extreme Bankrate changes that would otherwise be necessary. In this paper I ignore the gold devices largely because I have no data series on their use. The descriptive literature portrays them as a relatively minor part of Bank policy. The policy tool I focus on in the empirical work is Bank rate, since all descriptions feature it as the principal embodiment of policy. Bank-rate information is available in weekly form from several sources, including, for the period 1888 to 1909, a very useful U.S. National Monetary Commission statistical volume (1910). Following Goodhart (1972), I use for each month the Bank rate in effect at the beginning of the month. I also use a form of the second policy tool, the Bank's prototype form of open-market operations. The U.S. National Monetary Commission volume contains information on Bank holdings of government securities and other domestic interest-bearing assets. As a securities-holdings variable, I use the reported holdings of all such assets at the beginning of each month. 3.2.2 Target Variables By all accounts, the Bank of England devoted a large share of its attention to ensuring the adequacy of the liquid reserves in its Banking Department. Those reserves formed the base of the credit structure in the United Kingdom. As a result of its lending, the Banking Department's liabilities were generally two or three times its reserves. Some of those liabilities were deposits of the government or of customers of the sort a commercial bank might have. Others, however, were "bankers' balances," and as such constituted a large part of the reserves of the commercial-banking system. Thus any unusual demand for money anywhere in the U.K. economy or from abroad was likely to end up quickly as a demand on the Bank. Only the Banking Department's reserves were available in normal times to meet that demand. Generally the gold reserves in the other half of the Bank, the Issue Department, were considerably larger than Banking Department reserves; however, that

179

The Bank of England and the Rules of the Game

gold was by law (the Bank Charter Act of 1844) only to be given up in return for Bank of England notes. Gold held for backing notes not in the hands of the Bank was unavailable for meeting a money demand. Of course that gold was physically within the country and therefore available for emergency use in meeting a demand from abroad. In fact, the 1844 law governing Bank behavior with respect to gold was suspended by a special letter from the cabinet on rare crises occasions. (See Dornbusch and Frenkel, this volu~e, on the 1847 crises in which domestic liquidity demands supervened on foreign gold demands.) The Bank and government, however, were highly reluctant to suspend the law; for essentially all purposes, then, the Bank's available reserve was that of the Banking Department. For the purposes of this paper, therefore, I use Banking Department reserves as a variable indicating the Bank's ability to meet its liabilities, including its obligation to provide gold in return for its bank notes. I also use another widely watched indicator of reserve position, the "proportion. " The proportion was the ratio of reserves to liabilities of the Banking Department. It was of course closely related to the level of reserves. Either of these variables, if they act as expected, should demonstrate a significant inverse relationship to Bank rate in the empirical work. As reserves and/or the proportion dropped, Bank rate should have risen. In some equations a reserve-change variable will be used with Bank securities-holdings changes. In that case, if the Bank adapted its interestbearing asset-holdings to reinforce the effects of changes in reserve holdings, in accordance with the rules, then the measured relationship should be positive. In the description of the rules above, I mentioned the rules interpretation that would prohibit a countercyclical reaction of the Bank to domestic activity and prices. The empirical work described below tests for such reactions. On an annual basis, estimates of gross domestic product are available for years quite a bit earlier than those covered in this study. However, on a monthly basis they do not exist. I therefore employ proxies for domestic activity. Goodhart (1972) collected and used railway freight receipts, which were available by month back to the early 1890s. I have borrowed that series for this work. As an alternative I also use an unemployment-rate series originally collected from trade unions. Neither series is a perfect proxy for domestic activity. The railway-receipt series, of course, is unduly representative of certain sectors of the economy. Since it is a revenue series, it reflects not only quantity movements but price movements. The unemployment series applies only to a subset of workers, those in trade unions. In addition that subset altered as the number of trade unions reporting increased. Despite these limitations I employ the two series to test for Bank reactions. If the Bank followed the rules, then Bank rate should not respond to them. The rate of inflation, like domestic activity, should not have affected

180

John Dutton

Bank rate if the Bank followed the rules. I have computed a monthly rate of inflation based on a crude wholesale price index collected by Augustus Sauerbeck. Another variable I use, gold inflows, will serve as an indicator of internationally induced money movements. In some ways a gold inflow is like a positive movement in reserves. Therefore, I enter it in some versions of the securities-holdings equation to see if the Bank responded to gold flows rather than to changes in its own reserves. Unfortunately the data on gold flows are poor. Goodhart (1972) presents two monthly series, one collected by the Bank and the other by the English Board of Trade, both purportedly measuring the same thing. The two series are quite different, indicating that one or the other or both are defective. I have chosen to include gold flows, despite the limitations of the data. I use the Bank of England series; presumably the Bank would use that series, if it used any, to govern its behavior. 3.3 Modeling and Estimation The question posed is: What variables did the Bank of England react to? To answer it I treat the Bank's policy tools as dependent variables and regress them on several variables to which the Bank might have responded. Of course, any response is founded on a belief by the Bank that its tools in turn would affect these variables. The process of feedback of target variables to the actions of a policy authority is discussed by Sargent (1979, chap. 15). As he notes, proponents of the recently developed rational-expectations hypothesis tend to discount the effectiveness of systematically applied policy tools in influencing economic events. The argument is that individuals and firms, if they come fully to anticipate policy reactions, will take compensatory actions that will obviate the effects of policy. If these analysts are right, then feedback processes, being very regular and predictable, will be ineffectual in altering economic events. As Sargent points out, however, a sufficient condition for feedback policy responses to work is greater and more timely information availability to the policy authority. If such a condition holds, then individual agents will be unable fully to anticipate and compensate for policy actions; those policy actions as a result will have real effects on the economy. In this paper, I assume that the Bank of England at least believed its policy actions, including those determined by feedback rules, to be effective. The appearance of any significant coefficients in the feedback equation would bear out that assumption. For purposes of testing, the Bank is assumed to react each month to the values of target variables for that month, or more strictly speaking, to forecasts of target variables for that month. The feedback process is

181

The Bank of England and the Rules of the Game

modeled in linear form, i.e., the policy-tool variable is treated as a linear function of forecasts of the target variables. The forecasts, rather than the actual variables themselves, are used for two reasons. First, it is unlikely that the Bank would have full information about the most recent values of its targets at the time of undertaking policy action. Second, the forecasts serve to eliminate problems of simultaneous-equation bias. Two-way causality is implicit in the formulation of the model; policy actions influence targets, and targets, through the feedback mechanism, influence policy actions. The second avenue of causality is the one I test for. One way to reduce the chances of detecting the first direction of causality and mistaking it for the second is to use regressors that are predetermined at the time policy action is taken. In this case, I use forecasts of the target values based on information from previous months. The ideal forecast series, of course, would be those actually used by Bank officials. However, such series were likely never made completely explicit, much less written down. Proxies for the forecast series must be sought. One approach to forecasting used a great deal of late is that of Box and Jenkins (1976). To predict a variable, they use the past history of the variable, including both past values of the variable itself and past differences between the variable and its predicted value. The equations fitted for forecasting are of the form: Zt =

1

Zt-l

- a2 at -2 -

+ 2 Z t-2 + ... + p Zt-p + at ... -

a1 a t -l

aqa t - q ,

where Zt is the variable at time t (or in some cases changes in the variable), Zt _ i is the variable at time t - i, at is the prediction error at time t, at _ i is the prediction error at time t - i, and i and ai are parameters. The Zt-i terms form the autoregressive part of the process and the at-i terms are denominated the moving average part. Usually p and q are low numbers ranging from zero to two. If differencing is used, the process is said to be integrated. The acronym ARIMA, for "autoregressive integrated moving average" is commonly used with these models. In much of the empirical work that follows, I use ARIMA models to obtain forecasts for the target variables. Several objections to this method of generating forecasts may be raised. First, the Bank likely used information other than a variable's past history in generating forecasts for that variable. It probably used other variables, for example. Second, even if the Bank used only a variable's past history in generating its forecast, it might have used a different forecasting process than the equation estimated here. The forecasting equation estimated for a variable in this paper is based on data covering the whole period under consideration. At any given time, of course, the

182

John Dutton

Bank could have used only data preceding that time. If the pattern a variable followed were stable over the whole period covered, as well as over a suitably long preceding period, then the Bank in principle could have used the forecasting equation used here. If, on the other hand, that pattern were changing, then the equation I estimate here based on data up to 1914 could not, even in principle, have been available to the Bank before 1914. These considerations all imply that the forecast series are at best quite imperfect proxies for the Bank of England's forecasts. The problem can be viewed as one of errors in the variables. In such a case, the observed coefficients will be biased toward zero, making detection of any existing feedback less likely. Earlier I mentioned the use of predetermined forecasts as a way of eliminating simultaneous-equation bias. Unfortunately, use of those forecasts does not eliminate another "back-door" route by which causality running in the "wrong" direction could be picked up by the equations. If the policy-tool variable is autocorrelated, i.e., if its value at time t is correlated with its value at time t - i, then the possibility exists that any results from regressing the policy tool at time t directly on the forecasts could simply reflect relationships of both with the policy tool at time t - i. This problem, however, would not exist if the dependent variable at time t were not correlated with its own previous values. One way to meet this condition is to subtract from the dependent variable the part of it that can be predicted from its own past. For this purpose, I use an ARIMA model fitted to each dependent variable to generate predicted values, which are then subtracted from the actual values. The residuals can be thought of as the innovations that occurred each period in the policy variable. The innovations are uncorrelated with each other and therefore will not evidence any spurious back-door relationships with the predetermined target forecasts. This method should yield results reflecting causality going only in the direction from target variables to Bank policy. The equations fitted with this techinque are of the form at = X; ~ + Et , where at is the time t residual in the ARIMA equation for the policy tool, Xt is a vector of forecasts of the target variables, ~ is a vector of coefficients, and Et is the error at time t. Despite alterations of the Z and X series, the errors in this equation may still exhibit autocorrelation. If they do, the standard errors of the coefficients may be biased. To remove that autocorrelation and resulting bias, a generalized least-squares procedure may be used (see Theil 1971, p. 253, for a general description of the procedure) . The monthly data for the empirical work come from several sources, all detailed in a data appendix. Most of the series coyer the period from the late 1880s to mid-1914; the specific period for each series is listed in table 3.1. Most of the series exhibited marked seasonality. Some also exhibited distinctive time trends. To remove those influences the data were regres-

-

1

+ = in; - = out.

9.65 13.98 15.22 13.11 12.83 13.82 14.07 12.99 14.92 12.13 10.82 11.77 .039/mo.

JAN. 1888JUNE 1914

JAN. 1888JUNE 1914

4.15 3.54 3.31 3.22 3.07 3.06 2.81 2.92 3.13 3.58 4.12 4.12

Bank of England Reserves (million £)

Bank Rate (percent per annum)

Monthly Means and Trends

Sources: See Appendix.

January February March April May June July August September October November December Trend

Period

Table 3.1

-

37.3 50.0 48.3 44.3 47.0 48.6 44.5 48.0 51.9 45.6 45.9 48.1

4.96 4.54 4.48 4.01 3.99 3.94 4.07 4.40 4.51 4.57 4.39 4.90

442.5 509.3 520.3 471.1 492.9 442.1 445.7 442.4 523.8 565.8 558.4 468.9 1.73/mo. -

JAN. 1888JUNE 1914

JAN. 1887JUNE 1914

JAN. 1893JUNE 1914

Bank of England Proportion (percent)

Unemployment Rate (percent)

U.K. Railway Receipts (thousand £)

52,721 44,399 47,779 48,560 45,905 45,080 49,441 45,212 43,671 46,775 44,848 43,961 60.4/mo.

JAN. 1888DEC. 1909

Bank of England Securities (thousand £)

0.11 0.46 0.93 0.83 1.36 1.46 0.70 1.02 -1.28 -2.27 0.08 -0.03

JULY 1891JUNE 1914

Gold Flows l (million £)

184

John Dutton

sed on monthly dummy variables and trend terms. The residuals from those equations were used in subsequent steps. These residuals should be trendless and free of seasonal effects. Table 3.1 also presents the monthly means and estimated-trend coefficients. These statistics, though a secondary feature of this paper, are of interest in themselves. Note the unusually low average reserves and high average interest-earning assets of the Bank during January-the most usual month for issuing financial reports. Those averages apparently reflect the "window-dressing" in which English banks of the period engaged. The practice involved boosting gold and Bank of England note holdings just before public statements of financial condition were to be issued. The higher reserve holdings made the financial conditions of the banks appear more favorable to depositors and others. Table 3.2 presents the ARIMA equations used to generate the forecasts used as independent variables and the innovations used as dependent variables in the final equations. The Z's represent the detrended series with monthly means removed. L is a lag operator; i.e., .80L Zt = .80Zt - 1 and . 16L5 Zt = .16Zt - 5 . The a's are residuals that are assumed to be "white noise." A variable is a white-noise series if the value of the variable at time t is independent of its value in any other period. If an equation in table 3.2 has "captured" all the explanatory power of the variable's history in explaining its value in the present, then the residual series in that equation will be white noise. The Q-statistic probabilities on the right-hand side of the table are meant to answer the question: What is the probability of obtaining these residuals if they are from a white noise series? The Q-statistics are computed using autocorrelation coefficients of the residual series up through lags 6, 12, 18, and 24. The probabilities reported are all well above .1, indicating that we cannot reject the hypothesis that they are white noise. This fact demonstrates that the equations of the table are good fits and do capture most of the explanatory power present in each variable's history. One is led to ask why the particular lag structures of table 3.2 show up in the data. The low-order lags are what one might expect, and they show up fairly uniformly. The twelfth-order lags are also to be expected; they represent a bit of seasonality left in the data despite removal of monthly means. The other scattered lag terms of the Bank-rate and -reserves equations are harder to explain. I have been unable to divine any economic explanation of their presence but have simply accepted them. Before proceeding to report the results of the outlined procedures, I note an additional, similar mode of estimating the feedback equations described above. The use of the forecasting equations of table 3.2 effectively makes the policy variables into functions of lags of the target variables, with specific lag structures imposed from prior information. It seems reasonable also to estimate those equations without imposing lag

5

6

9

Zt(l- .80L - .16L + .16L - .09L ) = at (21.76) (2.74) (2.64) (2.28) Zt(l- .85L - .19L5 + .11L7 + .14Lll - . 16L12 ) = at (25.19) (4.06) (2.32) (2.39) (2.83) Zt(1-1.21L + .24L2 ) = at (1- .64L) (5.46) (8.78) (1.94) Zt(l- .91L)(1- .24L12 ) = at(l- .69L) (3.64) (18.52) (8.34) Zt(l- .72L - .17L2 - .07L3 ) = at (13.08) (2.51) (1.30) Zt(l- .24L) = at (4.70) Zt(l- L)(l- .20L12 ) = at(l- .57L) (11.01) (3.28) Zt = at(l + .21L - .30L2 ) (3.60) (5.18)

Fitted Equation

a

Forecasting Equations and Filters

.142 .343 .675 .243

.214 .671 .436 .154 .487 .416

.132 .232 .628 .503 .292 .642

.465

.233

.945

.991

.956

.727 .

.512

.424

6

Q = Statisticb Probabilities for Lags 12 18

.584

.394

.234

.849

.165

.330

.989

.732

24

.16Zt - 6 . Beginning year of data indicated; end was June 1914 except where indicated. Asymptotic t-statistics are in parentheses. bryhe probabilities of obtaining the estimated a's if they are from a white-noise series.

az indicates the variables before filtering; a indicates the variables after filtering and ideally is white noise. L is a lag operator; for example, .16L6 Zt =

Bank rate (1888) Reserves (1888) Proportion (1888) Railway receipts (1893) Unemployment (1887) Inflation rate (1885) Securities holdings (1888-1909) Gold inflow (1891)

Table 3.2

186

John Dutton

structures, i.e., to enter the lagged target variables themselves (with trend and monthly means removed) rather than the functions of those variables reported in table 3.2. Results of such an estimate procedure are reported below. It follows in some respects a procedure outlined in Nerlove, Grether, and Carvalho (1979, chap. 11). 3.4 Empirical Results

Table 3.3 reports coefficients for forecast variables regressed on the transformed Bank-rate variable. As expected from the descriptive literature on Bank policy, both the reserve level and the proportion showed up as strong influences on Bank rate. The relationship in each case was consistently negative, indicating that Bank rate was raised in response to low values of those variables and vice versa. Apparently the reserve level and the proportion are highly correlated; as a result, when both appear together in an equation, the effect of neither can be measured with precision. When they appear separately, however, each evidences statistically significant coefficients. Domestic economic activity was entered in some equations in the form of railway freight receipts and in others in the form of the unemployment rate. The former had consistently positive coefficients and the latter consistently negative ones. Both types indicate that Bank rate rose when activity was high and dropped when it was low. The level of statistical significance of the coefficients varied somewhat. The coefficient of unemployment was always high relative to its standard error, but the coefficient of railway receipts dropped when reserves or the proportion were included in the equation. Possibly railway receipts were more closely related to imports than was unemployment; if that were the case, it would be easier to estimate precisely a separate effect of unemployment than it would to estimate an effect for activity as represented by railway receipts. The rate of inflation also had significant coefficients in several of the equations. Its positive sign apparently indicates that Bank rate rose in response to high 'predicted rates of inflation and vice versa. Table 3.4 reports results when the change in Bank holdings of domestic interest-bearing assets is the dependent variable. The relationship between that variable and changes in reserves is of particular interest because of the emphasis Bloomfield (1959) and Nurkse (1944) placed on it. The negative sign on reserve changes (measured here as predicted reserves less actual reserves in the previous month, both adjusted for trend and seasonality) indicates that Bank securities holdings generally increased when reserves were predicted to fall, and vice versa. An increase in those security holdings, other things equal, would lead to an increase in Bank of England notes and gold held by commercial banks and the public, and to an increase in the domestic money supply. The

187

The Bank of England and the Rules of the Game

increase would tend to counteract any decrease caused by the withdrawal of reserves to the foreign sector. Thus the Bank at least in part seems to have sterilized the effects of international money flows on the domestic supply of money. As an alternative approach to the Bank's reactions, I estimate similar equations using gold inflows rather than reserve changes. The results, reported in equations 6 through 9, are similar to those described above, though the significance level of the coefficients is substantially lower. Gold inflows, which would normally suggest money-supply increases, led the Bank to respond by selling off securities, thus countering some or all of the change in the supply of money. In none of the equations of table 3.4 was either of the activity variables significant. The rate of price change, however, appears to be nearly significant. Its positive coefficient indicates that expected increases in prices led to higher Bank holdings of interest-earning assets. The response, if correctly measured, is a procyclical one; that is, higher inflation appears to cause the Bank to expand the British supply of money. This procyclical response contrasts with the apparently countercyclical response reported in table 3.3. The reserve level was also included in several equations. It had no significant effect in the reserve-change equations, though its coefficients were always positive. When gold inflows were substituted for reserve changes, the reserve level almost assumed statistical significance. Its positive coefficient indicates that the higher the reserves, the greater the inflow of interest-bearing securities into the Bank's portfolio, other things equal. Table 3.5 contains results of stepwise regressions paralleling the equations of table 3.3. These regressions are estimated by allowing independent variables to enter the equations in order of statistical significance. Only variables significant at the 0.15 level were included in the equations of table 3.5. The independent variables in these equations are lagged values of the detrended, deseasonalized series. Entering lagged values directly avoids imposing a lag structure of the sort imposed in using the ARIMA forecasting equations. The dependent variables are the same "innovations" in Bank rate used for table 3.3. Because of the loose specification of the equations, table 3.5 is harder to interpret than table 3.3. However, it appears that reserves and/or the proportion had the most significant overall effects on Bank rate. In each case, the sum of the effects is negative. The fact that much of the effect comes from lags of low order lends credibility to the results. The other variable that seems significant is inflation. Its overall effect is positive, as it was also in the table 3.3 equations. Neither of the activity variables appears to be important, though lags 2 and 12 of the railway-receipts variable enter equation 3. Possibly the specification of the equation is too loose for domestic activity to appear significant in the presence of re-

Nb

258

258

306

306

258

258

318

318

306

1. (1893)

2. (1893)

3. (1889)

4. (1889)

5. (1893)

6. (1893)

7. (1888)

8. (1888)

9. (1889)

-0.75 (0.64)

-1.47 (2.76)

-1.54 (2.82)

-1.72 (2.49)

-2.07 (2.74)

Reserves x 10- 2

-1.18 (1.38)

-0.99 (2.16)

-1.15 (2.29)

-1.57 (2.98)

-1.80 (3.20)

Proportion X 10- 2

Bank-rate Equations, 1888-1914

Equation a

Table 3.3

1.01 (0.62)

0.52 (0.30)

2.05 (1.28)

1.62 (0.95)

Rlwy Rcpts X 10- 3

-2.97 (2.12)

-3.07 (1.98)

-3.19 (2.70)

-3.35 (2.79)

Unemp X 10- 2

1.99 (2.34)

1.27 (1.31)

1.72 (1.92)

1.14 (1.16)

Inflation

Lag 3, - .10(1.76) Lag 12, .12(2.11)

Lag 2, .10(1.80), Lag 3, - .10(1.84) Lag 12, .10(1.80)

Lag 3, - .10(1.87) Lag 12, .10(1.74)1.74)

Lag 2, .11(1.72)

None

Lag 2, .12(2.16) Lag 12, .13(2.32)

Lag 2, .10(1.76) Lag 12, .13(2.29)

None

Lag 3, - .11(1.76)

Autoregressive Termsc

306

306

318

258

318

318

258

318

11. (1889)

12. (1889)

13. (1888)

14. (1893)

15. (1888)

16. (1888)

17. (1893)

18. (1888)

-1.60 (3.08)

-0.42 (0.51)

-1.02 (0.97)

-2.12 (3.28)

-1.55 (1.80)

-0.02 (0.02)

3.33 (2.31)

2.99 (1.92)

1.02 (0.61)

-3.90 (2.98)

-4.15 (2.84)

-2.87 (2.15)

Lag 2, .11(1.98), Lag3, - .09(1.69) Lag 12, .11(1.94)

Lag 2, .11(1.82)

1.71 (1.76) 2.14 (2.52)

Lag 3, - .11(2.07) Lag 12, .11(1.93)

Lag 3, - .10(1.72) Lag 12, .10(1.87)

None

Lag 3, - .11(2.07)

Lag 3, - .11(1.99) Lag 12, .13(2.25)

Lag 2, .12(2.16) Lag 12, .12(2.19)

Lag 2, .11(1.71)

1.92 (2.19)

1.70 (1.89)

1.27 (1.30)

Note: t-statistics are in parentheses. 8Date is beginning year of data; all data extend through June 1914. bN = number of observations. clndicated are lag number of autoregressive error term, estimated autocorrelation coefficient used in data transformation, and t-statistic of parameter.

258

10. (1893)

252

252

204

252

252

222

222

204

222

1. (1889)

2. (1889)

3. (1893)

4. (1889)

5. (1889)

6. (1891)

7. (1891)

8. (1893)

9. (1891)

-6.81 (2.08) -6.70 (2.06)

(l.31)

-6.61 (2.53) -5.53 (1.76) -8.96

Reserve

6.35 (1.95) 5.30 (1.33) 6.48 (1.80)

2.20 (0.61) 0.79 (0.17) 1.86 (0.48) 1.80 (0.47)

FCx 10

-4.93 (1.49) -3.83 (1.15) -4.94 (1.42) -4.52 (1.36)

Gold Inflow 2 X 10

FC

Rlwy Rcpts

-6.03 (0.61)

-3.21 (0.33)

FC

1.22 (0.14)

-3.11 (0.40)

Unemp x 10

Note: t-statistics are in parentheses. Fe

= forecast. aBeginning year in parentheses; final year is 1909. bN = number of observations. CFigures are lag number of autoregressive term, parameter estimate, and t-statistic (in parentheses).

Nb

Reserve Forecast - Reserves ( -1) X 102

Securities-Holdings Equations, 1888-1909

Equatibn a

Table 3.4

9.70 (1.62) 9.65 (1.73)

10.08 (1.70) 9.67 (1.91) 9.84 (1.96)

Inflation 3 X 10

None

None

Lag 7, .13(1.96)

Lag 7, .12(1.73)

None

None

None

Lag 7, .11(1.82)

Lag 7, .11(1.78)

Autoregressive Termsc

191

The Bank of England and the Rules of the Game

Table 3.5

Stepwise Regressions on Bank Rate

EQUATION Explanatory Variables Entered Variables a Significant

2

Reserves, Lags 1-5 Rlwy Rcpts, Lags 1-12 Inflation, Lags 1-12 Reserves Lag 1 2 4 5 -

.0865 .0578 .0411 .0210

3

Proportion Lags 1-12 Rlwy Rcpts, Lags 1-12 Inflation, Lags 1-12

Reserves Lag 1 2 4 7 8 -

(6.08) (3.56) (2.57) (1.50)

at 0.15 Proportion Lag 1 4 6 11 -

Level

R2

Reserves Lags 1-12 Proportion Lags 1-12 Rlwy Rcpts, Lags 1-12 Inflation, Lags 1-12

.0218 .0127 .0163 .0157

(3.75) (1.89) (2.35) (2.92)

.0989 .0609 .0289 .0391 .0482

(7.04) (3.83) (2.38) (2.49) (3.28)

Proportion Lag 6 - .0234 (3.90) 10 .0134 (1.87) 11 .0125 (1.82)

Railway Receipts None Significant

Railway Receipts None Significant

Railway Receipts Lag 2 .0012 (1.57) 12 - .0014 (1.93)

Inflation Lag 1 7 11

Inflation Lag 1 .372 (1.75) 5 - .471 (2.17) 7 .417 (1.91) 11 .468 (2.17) 0.19

Inflation Lag 1 .356 (1.76) 5 - .507 (2.40) 6 .431 (2.01) 11 .344 (1.69) 0.32

0.24

.448 (2.19) .591 (2.91) .543 (2.64)

at-statistics in parentheses following coefficients. Intercept included but not shown.

serves and/or the proportion. Of course, if either appeared significant despite the loose structure, then the hypothesis that the Bank engaged in countercyclical policy would have been strengthened. 3.5

Interpretation

The question addressed in the paper is: Did the Bank of England follow the rules of the gold standard game? The answer, insofar as it can be given, is a soft-spoken no. The rules, at least in spirit, required that the Bank not react countercyclically to domestic activity or price changes. The results of the equations of table 3.3 indicate that the Bank did react countercyclically. The reaction is present even controlling for the effects of reserves on Bank policy, suggesting that the domestic-activity and price-change variables are not simply acting as proxies for reserves in the equations. Those variables appear to have had direct effects on Bank policy independent of their indirect effects via reserves. This finding

192

John Dutton

conflicts with the beliefs of several previous analysts (e.g., Ford 1962; Bloomfield 1959) that the Bank did not engage in countercyclical policy of any sort. On the other hand, the finding is supported by much descriptive literature, which frequently mentions the Bank's sensitivity to the effects of its policies on the domestic economy. Table 3.3 suggests that the effects of reserves and the proportion were strongest. They showed up consistently and generally had the highest levels of statistical significance. This suggests that whatever domestic cyclical variables affected Bank behavior, they were outweighed by the need to maintain convertibility of the currency. The conclusion is also supported by the less constrained equations of table 3.5; there reserves and the proportion show up as the most significant variables and with low-order lag structures of the type that seem most reasonable. Thus it is necessary to speak softly of Bank rule-breaking; the results of the Bankrate equations indicate a strong preoccupation with convertibility. Table 3.4 also supports a negative answer to the question: Did the Bank follow the rules? Bank-reserve decreases seem to have led to increases in Bank holdings of interest-earning assets. Instead of amplifying the effects of reserve changes on the money supply, the Bank seems to have sterilized them. The sterilization mayor may not have been intentional. Bloomfield (1959) described Bank policy with respect to its securities holdings as somewhat passive; demands for additional liquidity were met by passively acceding to requests for additional discounting. Whether passive or active in the process, the Bank apparently acted as a buffer between reserve movements and money-supply changes. The rules would demand that it be an amplifier. The Bank at least to some extent seems to have violated the rules, yet the international monetary system was stable. Were the Bank's violations too minor to be important? Or were the rules themselves unimportant for the adjustment process? Must we rely on the unsatisfying attribution of stability to blind luck, the confluence of fortuitous circumstances?

Appendix

Data Sources

Sources for the monthly data used are Goodhart (1972, appendixes VA and VB), the U.S. National Monetary Commission statistics volume on the United Kingdom, Germany, and France (1910), and some National Bureau of Economic Research data sheets kindly supplied me by Anna Schwartz. The latter include price data based on a 1928 Journal of the Royal Statistical Society article ("Wholesale Prices of Commodities") attributed to "the Editor of the 'Statist' ," and unemployment data taken from the British Abstract of Labour Statistics. The unemployment data

193

The Bank of England and the Rules of the Game

are based on returns collected by the Board of Trade and the Ministry of Labour from trade unions paying unemployment benefits. The price data are based on price indexes for forty-five commodities computed by Augustus Sauerbeck; the overall index is a simple arithmetic average of the individual ones. As such, it suffers from obvious weaknesses. However, I have been unable to locate an alternative monthly index. The sources of the data used are listed below. Bank rate, Banking Department reserves, and the proportion: January 1888-June 1891, U.S. National Monetary Commission 1910; July 1891June 1914, Goodhart 1972. Railway receipts ("average weekly gross goods receipts of major British Rlys"): January 1893-June 1914, Goodhart 1972. The March and April 1912 figures were substantially lowered by a coal strike; because the effect on domestic activity was likely much less drastic, I have substituted higher numbers (1182 and 1161) for those two months. Unemployment rates: January 1887-June 1914, NBER data sheets. Price index: January 1885-June 1914, "Wholesale Prices" 1928. Securities holdings: January 1888-December 1909, U.S. National Monetary Commission 1910. This series consists of the sum of government securities in both Bank departments, plus "other securities" in the Banking Department. Gold inflows: July 1891-June 1914, Goodhart 1972. The series used is that attributed to the Bank of England.

References Bagehot, Walter. 1873. Lombard Street: A description of the money market. London: John Murray. Beach, W. Edwards. 1935. British international gold movements and banking policy, 1881-1913. Cambridge: Harvard University Press. Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard, 1880-1914. New York: Federal Reserve Bank of New York. Bordo, Michael David. 1981. The classical gold standard: Some lessons for today. Federal Reserve Bank of St. Louis Review 63: 2-17. Box, George E. P., and Gwilym M. Jenkins. 1976. Time series analysis: Forecasting and control. San Francisco: Holden-Day, Inc. Clapham, Sir John. [1944] 1970. The Bank of England: A history. Vol. 2, 1797-1914. Reprint. Cambridge: Cambridge University Press.

194

John Dutton

Dutton, John. 1978. Effective protection, taxes on foreign investment, and the operation of the gold standard. Ph.D. diss., Duke University. Fetter, F. W. 1965. Development of British monetary orthodoxy, 17971875. Cambridge: Harvard University Press. Ford, A. G. 1962. The gold standard, 1880-1914: Britain and Argentina. Oxford: Clarendon Press. Goodhart, C. A. E. 1972. The business of banking, 1891-1914. London: Weidenfeld. and Nicolson. Great Britain. Ministry of Labour. Statistics Branch. 1894-1926. Abstract of Labour Statistics, annual issues. Lindert, Peter H. 1969. Key currencies and gold, 1900-1913. Princeton Studies in International Finance, no. 24. Princeton: Princeton University Press. McCloskey, Donald N., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance of payments, ed. J. A. Frenkel and H. G. Johnson. London: George Allen & Unwin. Mints, Lloyd W. 1945. A history of banking theory in Great Britain and the United States. Chicago: University of Chicago Press. Morgan, E. Victor. 1965. The theory and practice of central banking, 1797-1913. London: Frank Cass & Co. Nerlove, Marc, David M. Grether, and Jose L. Carvalho. 1979. Analysis of economic time series: A synthesis. New York: Academic Press. Nurkse, Ragnar. 1944. International currency experience. Princeton: League of Nations. Palgrave, R. H. Inglis. [1903] 1968. Bank rate and the money market in England, France, Germany, Holland, and Belgium, 1844-1900. Reprint. New York: Greenwood Press. Pippenger, John. 1974. Bank of England operations, 1890-1908. Mimeo. Sargent, Thomas J. 1979. Macroeconomic theory. New York: Academic Press. Sayers, R. S. 1976. The Bank ofEngland, 1891-1944. Vol. 1. Cambridge: Cambridge University Press. - - . 1953. The bank in the gold market, 1890-1914. In Papers in English monetary history, ed. T. S. Ashton and R. S. Sayers. Oxford: Clarendon Press. Theil, Henri. 1971. Principles ofeconometrics. New York: John Wiley & Sons. Triffin, Robert. 1964. The,evolution ofthe international monetary system: Historical reappraisal and future perspectives. Princeton Studies in International Finance, no. 12. Princeton: Princeton University Press. U.S. National Monetary Commission. 1910. Statistics for Great Britain,

195

The Bank of England and the Rules of the Game

Germany, and France. Washington, D.C.: Government Printing Office. Wholesale prices of commodities in 1927. 1928. Journal of the Royal Statistical Society 91 (pt. 3): 403.

Comment

Donald E. Moggridge

The phrase "rules of the game" only came into the economist's vocabulary as the interwar gold standard neared its end. So far as I can ascertain, the term was first used by Sir Robert Kindersley, a director of the Bank of England, on February 1930 in the course of his evidence to the Committee on Finance and Industry (United Kingdom 1931b, question 1595). The phrase attracted Keynes's attention and found its way into his "private evidence" to the committee two weeks later (Keynes 1981, p. 42) and subsequently into the committee's report (United Kingdom, Committee on Finance and Industry 1931a, pars. 46-47). Between Kindersley's coining of the phrase and the present the exact meaning of the "rules" has varied. For Keynes, they meant that "you so conduct your affairs that you tend neither to gain nor to lose large quantities of gold" (Keynes 1981, p. 42). The Macmillan committee was even more general when it argued: It is difficult to define in precise terms what is implied by the "rules of the game". The management of an international standard is an art and not a science, and no one would suggest that it is possible to draw up a formal code of action, admitting of no exceptions and qualifications, adherence to which is obligatory on peril of wrecking the whole structure. Much must necessarily be left to time and circumstance. (United Kingdom, Committee on Finance and Industry 1931a, par. 47) Nevertheless, when economists have come to assess the possible reasons for the success or failure of fixed-exchange-rate regimes, and occasionally even more "flexible" ones (Chisholm 1979), they have often specified a set of rules conducive to the stability of the regime and tested for the relevant authorities' adherence or nonadherence thereto. The most famous of these exercises are those of Nurkse (1944) and Bloomfield (1959) for the interwar and pre-1914 gold standard periods. Both men tested a relatively activist rule that internationally equilibrating behavior would move central banks' foreign and domestic assets in the same direction as they reinforced the impact of reserve movements on financial Donald E. Moggridge is professor of economics at Scarborough College, University of Toronto, West Hill, Ontario.

196

John Dutton

markets. In both cases, using annual data, they found adherence to such a rule was the exception rather than the norm. Alternative suggestions for rules have been provided by Bloomfield (1959, 1968) and Michaely (1968). The former suggested that behavior would be equilibrating if central banks did not offset the effects of reserve changes or if their discount rate moved inversely with their reserve holdings or reserve ratios. The latter suggested that suitable behavior would see the money supply varying directly and the central-bank discount rate moving inversely with reserve changes. Bloomfield only chose to test his discountrate rule during the pre-1914 period while Michaely's rule was applied to the Bretton Woods period. Doubtless one could also present another rule involving relative rates of change of a suitable monetary aggregate and apply it to the pre-1914 period, but as yet I know of no such exercise. John Dutton's paper brings to the discussion of the observance or nonobservance of the rules of the game by the pre-war Bank of England a new twist to the rules and a new test of observance of the NurkseBloomfield reinforcement rule-and incidentally a check on the more passive no-offsetting and discount-rate rules. Dutton's new twist states that given equilibrating central-bank behavior under classical gold standard conditions, one would not observe central banks pursuing countercyclical policies. One can see the sense of such a rule in the abstract, for it would mean that the banks involved would avoid meeting potentially destabilizing dilemma cases and it would incidentally economize on the need for international reserves. However, I wonder whether it is an appropriate rule for the pre-1914 international economy given that Bloomfield (1959, p. 38), Morgenstern (1959, chap. 2) and Triffin (1964, chap. 1) have all noted the strong parallelism in movements of economic activity during the period. In such circumstances, nonadherence to the rule might still be consistent with the successful operation of the standard. As well as providing a possible new rule, Dutton's paper tests the Bank of England's observance of various rules in a new form. Rather than simply comparing the authorities' actual behavior to the rules, he proposes a more complex model in which the Bank reacts to forecasts of its possible target variables (the Banking Department reserve, the proportion, domestic activity, prices and gold movements) by altering the level of its discount rate and its domestic assets. The rationale for this more involved procedure is twofold. First the Bank might not have full information as to the most recent values of its target variables at the time of making policy adjustments, and second, the procedure eliminates some problems of simultaneous-equation bias. Leaving the second reason to one side, I cannot fully see the strength of Dutton's first procedural rationalization. It is true, given Dutton's-and Charles Goodhart's (1972)-problem of finding a good monthly index of

197

The Bank of England and the Rules of the Game

activity, that the Bank was unlikely to have had full information on that score. The same would almost certainly be the case as regards the price level. But why Dutton should think that the Bank did not have full information about its own reserve, its proportion, and gold movements strikes me as odd, given the information the governors received each working day at the daily "books" meeting (Sayers 1976, 1: p. 31). Thus it would seem to me that the justification for Dutton's technique must depend more on the usefulness of the new rule he wishes to test and on its statistical characteristics than on its being representative of the details of Bank behavior. This is particularly the case when the forecasting model assumes that the Bank had information on each variable's behavior over the entire period and used the same processes consistently. Both of these assumptions seem to me suspect, for the seasonal variability of certain matters such as internal drains changed over the period (Sayers 1976, 1: p. 32) and it is clear from the narrative material available that the Bank's procedures and techniques, as well as the balance among the latter, were changing markedly over the twenty-five years before 1914. Thus it would seem to me that one cannot really be certain exactly what Dutton's procedures are capturing at the end of the day. What appears to be going on is that the Bank reacted most markedly to the traditional stimuli, changes in the level of its reserves and the proportion, and in the expected direction. One would expect these reactions; they would be consistent with one Bloomfield rule and part of the Michaely rule of the game. There is as well the confirmation of the inverse relationship between changes in the Bank's reserves and its holding of securities that one might expect given Bloomfield's evidence on annual data that the Bank followed the reinforcing rule of the game just less than half the time between 1880 and 1914 (Bloomfield 1959, p. 50). What we do not know is whether this offsetting was partial as Bloomfield suggested (1959, p. 50) or complete-whether the Bank was merely inclined to lean against the wind or stand resolutely against it. The former might be within the spirit of a possible rule, given its Bank-rate reaction, while the latter would represent a violation of the rules in almost any common formulation. Nor do we know, although the Bank did, how much offsetting was an automatic reflection of the discount market being forced into the Bank and how much reflected deliberate policy. Perhaps some day the Bank or some private scholars will extract the necessary information from the "books" and thus help remove another puzzle. Finally we have the suggestion that the Bank responded countercyclically to the activity and price variables, something that Bloomfield (1959, p. 33) regarded as incidental and Ford (1962, p. 34) believed was not a conscious policy at all, although narrative accounts of the Bank's behavior have understood it as a subsidiary but growing preoccupation. Whether this behavior reflected a continuing but changing Bank concern or the long-standing

198

John Dutton

suggestion that given the power of Bank rate the money-supply process in Britain was to some extent endogenous (Ford 1962, p. 36; Goodhart 1972, chap. 15) is not pursued in this paper, concerned as it is with the rules of the game. Thus the paper leaves us in the position of confirming the Bank's adherence to some rules and suggesting violations of others. Perhaps further work by Dutton will clarify the extent of these violations, especially of the no-offsetting rule, and indicate whether the Bank became more inclined to violate some rules over time-perhaps because it did become more concerned about levels of economic activity. For the present we can thank Dutton for the questions he has raised and hope that discussion and more work will help us come to find answers. References Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard, 1880-1914. New York: Federal Reserve Bank of New York. - - - . 1968. Rules of the game of international adjustment In Essays in Money and Banking in Honour of R. S. Sayers, ed. c. R. Whittlesey and J. S. C. Wilson. Oxford: Clarendon Press. Chisholm, Derek. 1979. Canadian monetary policy, 1914-1934: The enduring glitter of the gold standard. Ph.D. diss., Cambridge University. Ford, A. G. 1962. The gold standard, 1880-1914: Britain and Argentina. Oxford: Clarendon Press. Goodhart, C. A. E. 1972. The business of banking, 1891-1914. London: Weidenfeld and Nicolson. Keynes, J. M. 1981. Activities, 1929-1931: Rethinking employment and unemployment policies. Vol. 20 of The Collected Writings of John Maynard Keynes, ed. Donald Moggridge. London: Macmillan for the Royal Economic Society. Michaely, Michael. 1968. Balance-of-payments adjustment policies: Japan, Germany, and the Netherlands. Occasional paper 106. New York: National Bureau of Economic Research. Morgenstern, Oskar. 1959. International financial transactions and business cycles. Princeton: Princeton University Press. Nurkse, Ragnar. 1944. International currency experience: Lessons of the inter-war period. Princeton: League of Nations. Sayers, R. S. 1976. The Bank of England, 1891-1944. Cambridge: Cambridge University Press. Triffin, Robert. 1964. The evolution ofthe international monetary system: Historical reappraisal and future perspectives. Princeton Studies in International Finance, no. 12. Princeton: Princeton University Press.

199

The Bank of England and the Rules of the Game

United Kingdom. Committee on Finance and Industry. 1931a. Report. London: His Majesty's Stationery Office. - - - . 1931b. Minutes of evidence. London: His Majesty's Stationery Office.

Reply

John Dutton

I would like to respond to two points in Professor Moggridge's "Comment. " First, he indicates concern that the behavior of the Bank of England varied substantially over the twenty-seven years studied. He is also bothered by the possibility of changes in seasonal and other patterns in the target variables over the period and by the effects those changes might have on the forecasting equations. I share those concerns and so have attempted to allay them. Tables C3.1 and C3.2 contain linear feedback equations, similar to those of tables 3.3 and 3.4, but for two shorter periods. Some changes in the coefficients appear. However, the earlier findings for the whole period remain intact in the results for the two subperiods. Bank rate in each subperiod reacted negatively to reserves, positively to economic activity, and positively to inflation. The reserve variable is clearly significant in each case (although its coefficient is much greater for the second period). The activity variables give much the same results as those for the whole period. Unemployment has a significant, or nearly significant, negative effect; railway receipts show a weak but consistently positive effect. Inflation has statistically significant positive coefficients for the first period but not the second. Likewise, in both periods, securities holdings of the Bank tend to change in the opposite direction from predicted reserve changes. Securities appear to have decreased about one-half to two-thirds as much as the change in reserves. This decrease is consistent with results for the whole period, indicating that the processes governing the Bank probably did not change dramatically between the two periods. Moggridge mentions concern that seasonal patterns of the variables changed over the twenty-seven years of the study. Figures C3.1 and C3.2 contain plots of monthly means of several variables for periods ending in 1900 and beginning in 1901. The patterns for the two periods seem remarkably similar. Gold inflows evidence the largest change (as well as large standard deviations for the means). Unemployment patterns during the first quarter of the two subperiods differ somewhat. The overall impression, however, is of substantial likeness between the overall seasonal patterns. Professor Moggridge rightly worries about stability of the forecasting

200

John Dutton Bank-rate Equations

Table C3.1

Equation a

Reserves x 10- 2

Nb

Rlwy Rcpts X 10- 3

Unemp X 10- 2

Inflation

Autoregressive Termsc

First Period (through 1900)

none

1. (1893)

96

-1.34 (1.82)

4.01 (1.01)

2. (1893)

96

-1.10 (1.47)

4.87 (1.23)

3. (1888)

156

-1.45 (1.95)

-3.49 (1.83)

4. (1888)

156

-1.28 (1.75)

-3.12 (1.65)

5. (1888)

156

-1.81 (2.23)

2.40 (1.59)

none lag 7, .14(1.73) lag 10, - .14(1.80)

2.18 (2.03)

lag 7, .15(1.87) lag 10, - .14(1.80) lag 10, - .16(2.08)

Second Period (through 1914)

lag 1, .13(1.68)

6. (1901)

162

-18.42 (6.19)

1.70 (1.05)

7. (1901)

162

-18.46 (6.18)

1.73 (1.07)

8. (1901)

162

-18.37 (6.37)

-3.79 (2.53)

9. (1901)

162

-18.43 (6.39)

-3.84 (2.56)

10. (1901)

162

-18.50 (5.92)

0.25 (0.39)

lag 1, .13(1.70) lag 1, .15(1.99)

0.34 (0.55)

lag 1, .18(2.02) none

Note: t-statistics are in parentheses. aDate in parentheses is beginning year. b N = number of observations. cIndicated are lag number of autoregressive error term, estimated autocorrelation coefficient used in data transformation, and t-statistic of parameter.

Table C3.2

Securities-Holdings Equations a

Period

N

1888-1900

155

1901-1909

108

Reserve Forecast - Reserves (-1) x 102 -7.39 (1.40) -5.25 (1.84)

Note: t-statistics are in parentheses. aThe errors in this equation showed no statistically significant autocorrelation. Securities changes measured in thousand-pound units, reserves in million-pound units. Coefficient x 10- 1 indicates portion of reserves changes offset.

201

The Bank of England and the Rules of the Game ~.j

11 ions of

ounds

Percent

15 10

Balk of

Eng I Old

Brnking

~oortrrent

reserves

55

/

//'\

50

\

\

'.-/

\ \,...- -"

45 40

.,r-\

Prnk rate \

/ /

4.0 3.5

/

~

.......

/

"""----- -_--....,

///

/

'-'"

3 .0

2.5

65

60

55 50

Fig. C3.1

Balk of

Eng IOld

securi ties holdings

Bank of England financial variables, means, by months, 18881900 and 1901-14. Earlier and later period values are indicated by broken and solid lines, respectively. Source: See section 3.6.

process over the whole period in question. The ARIMA forecasting equations are at best only rough proxies for the Bank's forecasts. One would expect the coefficients on these rough proxies to be biased toward zero. That significant coefficients are obtained for the period as a whole and for the two subperiods seems to indicate that the ARIMA forecasts are doing their proxying job reasonably well. The second of Moggridge's concerns to which I should like to respond is the validity of the no-countercyclical-activity rule of the game used in the paper. The rule seems reasonable to me because the Bank, with limited policy tools, would have needed to concentrate its efforts on the single major goal of maintaining convertibility. Efforts at countercyclical policy would likely have interfered with pursuit of that goal. They might

202

John Dutton ~'j

11 ions of

Percent

ounds

.6

,""-

-~

Railway receipts

,...----...

''----......... ' - - - - - - . . / /

//

,

//

,

",

5.0

"

Lharployrrent rate

4.5

4.0

1.5

0.5

--- -----

(?DId infICl'ls

-0.5 -1. 5

-2.5 .JAN

Fig. C3.2

F B

~~AR

APr

~~AT

JUN

JlTL

AUG

SEP

OCT

NOV

DEC

U.K. railway receipts, unemployment rate, and gold flows, means, by months, 1888-1900 and 1901-14. Source: See fig. C3.1.

also have had destabilizing effects. There is some reason to believe that gold inflows, Banking Department reserves, and domestic activity moved together (see Pippenger, this volume). If they did, then countercyclical Bank-rate policy would have tended to reinforce those cyclical movements of gold and reserves, rather than tempering them. Moggridge also suggests in his comment that the apparent countercyclical policy evidenced in my equations might have reflected a common response of Bank rate and economic activity to some third variable. Such a possibility cannot be completely dismissed. Possible candidate variables are reserves and the money supply. However, if reserves and the money supply were positively related to domestic activity, as seems likely, and Bank rate negatively related to those two, then any indirect effect of activity via reserves on Bank rate would be negative. The coefficients in the paper consistently indicate a positive relationship. Other scenarios for explaining the empirical results are of course possible. It is plausible, however, that the results could signify countercyclical actions on the part of the Bank. In any case, Professor Moggridge is certainly correct in calling for more work to reduce the extent of uncertainty about the Bank's policies.

Bank of England Operations, 1893-1913

4

John Pippenger

The Bank of England did not publish figures for bankers' deposits until 1967. The first economist to use that information was Goodhart (1972). This study builds on and reexamines the work of Goodhart, whose conclusions conflict with the conventional wisdom about the Bank and the gold standard. Section 4.1 reviews some of Goodhart's results, section 4.2 examines the long-run operations of the Bank, section 4.3 analyzes short-run behavior, and the final section 4.4 presents the conclusions. An appendix provides spectral estimates of key variables examined in this study. 4.1

Goodhart's Results

Goodhart (1972) analyzed the operations of the Bank of England and British commercial banks and their roles in the functioning of the gold standard from 1891 to 1914. His conclusions about the role of the Bank in the operation of the gold standard challenge the conventional wisdom. The strongest link in the causal chain of the classical analysis of the working of the gold standard mechanism is generally considered to be that connecting changes in the reserve base of the commercial banks with fluctuations in the (gold) reserve, or liquidity, position of the central bank. Yet in this study of the working of the system in the UK this is the link which shatters . . . . there is no simple direct relationship between the variations in the levels of bankers' balances at the Bank and in the level of the reserve in the Bank. (Goodhart 1972, p. 209) This conclusion rests primarily on two regressions. In the first, monthly data on bankers' balances at the head office of the Bank of England are John Pippenger is professor of economics at the University of California, Santa Barbara.

203

204

John Pippenger

regressed against time, reserves in the Banking Department, and seasonal factors. There is no link between bankers' balances and reserves.

= 7913.5 + 37.74 time

Bankers' balances

(967.4) (1.97) - 1.29 reserves + seasonals. (34.53) With seasonals R2

= 0.64, D.W. = 1.07. Without seasonals R2 = 0.56.

The numbers in parentheses are the standard errors. The second equation is in logs and adds railway freight receipts as a proxy for nominal income. Log bankers' balances = 4.843 + 0.0006 time (1.142) (0.0003) + 0.707 log freight receipts (0.167) + 0.092 log reserves. (0.40)

R2

= 0.679, D.W. = 0.96. Without seasonals R = 0.579.

With Seasonals

2

Now a positive relation between bankers' balances and reserves emerges, but the estimated response to income is several times larger than the response to reserves. Goodhart also estimates two other relationships that are relevant for the operations of the Bank of England. One attempts to explain the ratio of reserves in the Banking Department to total liabilities of that department----otherwise known as the proportion.

= 7.39 + 0.0006 time

Log proportion

(0.78) (0.0002) - 0.79 log freight receipts (0.11) + 0.53 log reserves + seasonals. (0.03)

R2

= 0.765, D.W. = 0.90. Without seasonals R2 = 0.584. With seasonals

The standard errors are in parentheses. Goodhart (1972, p. 206) interprets this result as follows: "It suggests that the Bank must have regularly accommodated, to some large extent, variations in the demand for cash

205

Bank of England Operations, 1893-1913

caused by changes in the level of domestic activity by varying its holdings of other assets, independently of the level of gold reserves." The final relationship attempts to explain Bank rate in terms of trend and the liquidity position of the Bank of England, first using the proportion and then the reserves as a measure of liquidity:

= 4.48 + 0.0012 time

Bank rate

(0.89) (0.0004) - 1.09 log proportion (0.22)

+ 0.746 Bank rate (t - 1) + seasonals. (0.037)

"R 2

= 0.795. Without seasonals "R = 0.736. Bank rate = 2.49 + 0.0021 time With seasonals

2

(0.55) (0.0005) - 0.714 log reserves (0.165) + 0.756 Bank rate (t -1) + seasonals. (0.038) With seasonals

"R 2

Without seasonals

= 0.791.

"R

2

=

0.749.

The numbers in parentheses are the standard errors. The results show the expected inverse relation between Bank rate and the liquidity position of the Banking Department. The next two sections reexamine the operations of the Bank of England, employing as much as possible the data used by Goodhart.2 The first section concentrates on the long-run and the second looks at the short-run behavior of the Bank.

4.2

Long-Run Operations

This section concentrates on long-run behavior by using annual averages of monthly data.3 The next section, in order to emphasize short-run operations, uses monthly changes. Sayers (1976, p. 8) points out that the governor of the Bank of England had three primary objectives. He had a statutory duty to maintain the convertibility of the note into gold coin; he had a political duty to look after the financial needs of government; and he had a commercial duty to maintain an income for

206

John Pippenger

the stockholders. Whenever possible, he was running all three horses at once, but if there was a conflict, he knew which he had to put first. He would think of his primary duty as the maintenance of the gold standard. Although a variety of special situations probably influenced the short-run operations of the Bank, the duties cited by Sayers, particularly the statutory and commercial duties, appear to dominate long-run behavior of the Bank. 4.2.1

Bankers' Deposits and Reserves

Goodhart's most challenging discovery is the weak relationship between reserves in the Banking Department and bankers' deposits at the Bank of England. His results threaten a crucial link in the conventional interpretation of the gold standard. Consider a very simple model of Bank-portfolio behavior in which desired reserves R depend on deposits and interest rates. (1)

R

=

<xii + <xzBD + <X3NBD + e,

where i is the market rate of interest (or vector of such rates), BD is bankers' deposits, and NBD is nonbankers' deposits at the Bank.4 Rewriting equation (1) yields an expression describing bankers' deposits.

(1') It is hardly surprising that Goodhart finds no link between reserves and bankers' deposits. Equation (1) implies that reserves are correlated with the error term in his regression. In addition, the influence of interest rates and other deposits is ignored. The inclusion of time compounds the problem because time tends to exclude any positive relation between bankers' deposits and reserves generated by growth. Goodhart's attempt to regress bankers' deposits against time, reserves, and a proxy for income has even less discriminatory power. Consider the following simple linear model for the determination of reserves. The demand for money depends on income and interest rates. (2) The supply ·of money by the banking system depends on notes N plus bankers' deposits at the Bank and interest rates. (3) Add a simplified balance sheet for the combined Banking and Issue departments of the Bank. (4)

BD+N=R+S,

where S is securities held by the Bank.

207

Bank of England Operations, 1893-1913

Equations (2)-(4) imply the following solution for reserves.

(5)

R = ko - 'Yo + !L y + k z + 'Y3 i 'Yl

'Yl

'Yl

- S + 'Yl - 'Y2 BD . 'Yl

In this equation bankers' deposits are unrelated to reserves if 'Yl equals 'Y2. This result however, ignores the fact that banks are free to choose how they hold liquid reserves. For simplicity, suppose they hold notes and deposits at the Bank in some fixed proportion.

(6)

N = 'Y4BD.

Equations (4) and (6) imply the following.

(7)

BD

=

[111 + 'Y4)](R + S) .

Substituting equation (7) into (5) yields a solution for reserves that is independent of bankers' deposits at the Bank of England. (8)

_K( 1 + 'Y4 + 'Y2 -

'Yl) S,

1 + 'Y4

where K equals 1/['Yl + ('Yl - 'Y2)/(1 + 'Y4)]. Bankers' deposits are unrelated to reserves in equation (8) because those deposits are determined primarily by the public's demand for money and the portfolio decisions of commercial banks. A more appropriate way to evaluate the link between reserves and bankers' deposits is to estimate an equation like (1). Equation (1) however ignores the Bank's statutory duties and treats the Bank as though it were only another commercial bank. In order to capture the influence of its role as a central bank, a proxy for income and a measure of foreign-relative-to-domestic interest rates are added to equation (1). If Goodhart's argument that the Bank essentially accommodated the demand for money is correct, then the coefficient for both variables should be negative. Higher rates abroad should lead to a loss of reserves and higher income should increase the money stock leading to a rise in bankers' deposits and an outflow of notes from the Banking Department. If the Bank operated only as a commercial bank, the coefficients for both these variables presumably would be zero. If however the Bank actively protected convertibility by responding to potential gold flows, then reserves in the Banking Department should increase as domestic income expands and foreign yields rise relative to domestic rates. The results from estimating such an equation using annual averages of monthly data are as follows, where t-statistics are in parentheses.

208

John Pippenger

R

= -8847.75 + 0.52NBD + O.57BD (1.90) (6.82)

(2.29)

- 2661.76i + 9040.07r + 9.35Y.

(5.17)

R2 = 0.87,

D.W.

(2.96) (1.54)

= 1.73.

The domestic interest rate i is the yield on fortnightly loans, r is the French market rate over i, and Y is railway freight receipts. Including both bankers' deposits and a proxy for income is likely to underestimate the influence of bankers' deposits because those deposits are related· to income through the demand for money. The proxy for income therefore is dropped and the equation reestimated. Since eliminating income substantially reduces the D.W. statistics, the equation is estimated using the Cochrane-Orcutt technique. R

= -8670.76 + 0.5NBD + 0.82BD (1.15) (6.21)

(3.25)

-2113.38i + 8864.09r.

(3.93)

(2.78)

R2 = 0.82, D.W. = 1.77, p = 0.352. The results do not support accommodation. The Bank held reserves against both bankers' deposits and other liabilities. Indeed the Bank seems to have been very conservative, holding up to eighty pounds in reserves for each one hundred pounds of deposits. The results also suggest that the Bank actively protected convertibility by increasing reserves as foreign rates increased relative to domestic rates. Although the coefficient on the proxy for income is not significant in the first regression, it is positive and therefore tends to refute accommodation. The Bank's concern for profit also emerges from these estimates. The coefficient on domestic rates is negative and significant, which is what we would expect from a bank concerned about paying dividends.

4.2.2 Proportion The proportion P of reserves in the Banking Department to total liabilities was the most common measure of the Bank's liquidity. Goodhart's estimates based on monthly data reveal an inverse relation between the proportion and his proxy for income, which he interprets as support for an accommodative Bank. The analysis of the Bank's demand for reserves however suggests that the proportion should reflect the Bank's concern for dividends and the desire to protect convertibility. The proportion therefore should depend directly on foreign relative to domestic yields and be related inversely to domestic interest rates. In

209

Bank of England Operations, 1893-1913

order to test Goodhart's hypothesis, his proxy for income also is included. Estimating such a relationship yields the following result. P

=

0.318 - 0.052i + 0.193r + 0.0002Y. (3.98) (5.29) (3.51) (3.38)

R2 = 0.73,

D.W. = 1.69.

The fit is good. All of the coefficients are significant at better than the 1 percent level and the Durbin-Watson statistic indicates that there is no serial correlation in the residuals. The evidence does not support accommodation. The Bank systematically reduced liquidity in order to earn income and protected convertibility by becoming increasingly conservative as domestic income increased or foreign rates rose relative to domestic rates.5 This result is particularly interesting because it provides insight into how the Bank handled the conflict between its statutory and commercial duties. For example, an upswing in business activity tended to make earning assets more attractive as interest rates rose, but the increased activity also posed a threat to convertibility. The Bank apparently responded to both influences, raising the proportion in response to increasing income and lowering it as interest rates rose. Such a policy of course would make it very difficult to identify a systematic pattern in Bank behavior over the business cycle and may help explain Bloomfield's inability to find evidence supporting the rules of the game in his seminal work (1959). 4.2.3

Bank Rate

The conventional story in which the Bank raises the discount rate in response to a loss of reserves is a disequilibrium process that is not relevant for this section because in the long run the actual and desired portfolio should be equal. Since the Bank of England was only one bank in the London money market and London was only part of the world capital market, in the long run Bank rate should follow rather then influence market rates. Bank rate therefore is assumed to depend primarily on market rates. A proxy for income and a measure of foreign relative to domestic rates were included to see if the Bank responded to potential threats to convertibility. Since the (-statistic for income is less than one, it has been dropped. The final result for Bank rate is as follows. BR

= -

0.443 + 0.882i + 1.000r. (1.225) (18.73) (3.09)

R2 = 0.95, D.W. = 2.26.

210

John Pippenger

The fit is very good. Both coefficients are significant at the 1 percent level and there is no evidence of serial correlation in the residuals. The results indicate that the Bank actively protected convertibility by raising Bank rate basis point for basis point with the rise in French market rates relative to domestic yields. The evidence also suggests that Bank rate did not fully respond to domestic yields. This result however may reflect the fact that i is the yield on fortnightly loans while Bank rate tended to apply to longer maturities. 4.2.4 Summary Although the results of this section indicate that the Bank was more concerned about profit than is often recognized in the conventional story about the gold standard, the evidence generally supports the conventional wisdom. In the long run the Bank was very conservative, maintaining up to eighty pounds sterling or more in liquid reserves for each hundred pounds in deposits. Such behavior implies that the Bank did play by the rules of the game and bought assets as reserves increased. Of course, if the monetary approach is correct, this behavior had no long-run effect except to alter the Bank's liquidity and earnings. The evidence also suggests that the Bank responded to foreign financial conditions and followed a mixed monetary policy over the course of the business cycle. Reserves, the proportion, and Bank rate all rose as French market rates rose relative to domestic rates. This movement suggests that the Bank was sensitive to the threat to convertibility from international capital flows. Although the commercial duties of the Bank promoted a procyclical monetary policy, concern for convertibility apparently led the Bank to increase the proportion as nominal income expanded. 4.3

Short-Run Operations

Analysis of long-run behavior of the Bank is relatively straightforward because simultaneity is not a serious problem. In the short run, however, the Bank's portfolio decisions can affect bankers' deposits, interest rates, and other variables, and equations like those estimated in the last section may be biased in the short run. Two-stage least squares can deal with simultaneity, but it does not appear to be applicable here. The technique requires the use of explanatory variables in the first stage that are independent from the error term in the second stage. No such variables appear to be available. One of course could use two-stage least squares anyway and pretend that the problem was solved. The choice here however is to use OLS and accept the bias due to simultaneity. The variance of economic time series, including the ones used here,

211

Bank of England Operations, 1893-1913

tends to be dominated by long-run or low-frequency components. Since we want to concentrate on short-run behavior, and differencing tends to filter out long-run components of the variance in time series, the analysis of this section uses monthly changes.6 4.3.1

Reserves

The results of the last section indicate that in the long run, desired reserves R D depend on bankers' balances BD, nonbankers' deposits NBD, market yields i, and foreign-relative-to-domestic interest rates r. (9)

R D = ao + alBD + a2 NBD - a3 i + a4 r + el,

where el is an appropriate error term. Equation (9) describes the Bank's equilibrium or steady-state demand for reserves. This equation can be converted to a short-run model by using a simple stock-adjustment model. (10)

dRt = 'AI(RP- R t -

l)

+ 'A 2LlG + ez,

where LlG is the gold flow into ( + ) and out of ( - ) the United Kingdom and e2 reflects other shocks such as internal drains due to holidays. The solution for reserves implied by equations (9) and (10) is given by equation (11). (11)

Rt

= 'Alao + 'AlalBD + 'Ala2NBD - 'A l a3i + 'AI a4 r + 'A z LlG + ('AI el e2) .

In estimating this equation in first differences, the seasonal components in ez are captured with seasonal dummies.7 There are two major potential sources for simultaneous-equations bias in this model. Consider a situation in which the Bank has excess reserves. In order to move toward portfolio equilibrium, the Bank buys securities. The purchase of securities raises bankers' deposits, reduces reserves, and tends to lower interest rates. As a result, unless the monetary approach holds even for monthly data, portfolio decisions by the Bank influence bankers' deposits and inter¥st rates. The influence on interest rates however is probably not as important as for bankers' deposits. Purchases of securities almost certainly had an initial pound-for-pound impact on bankers' balances, but the Bank of England was only one of many banks in the London money market and for periods as long as a month there almost certainly was a strong link between the London and world capital markets. When equation (11) is estimated in first differences, there is significant negative serial correlation in the residuals. The results reported in table 4.1 therefore are based on the Cochrane-Orcutt technique. The OLS results (table 4.2) however are almost identical, as they are for the other' two regressions reported in table 4.1.

212

John Pippenger

Table 4.1

Short-Run Determinants of Reserves, Proportion, and Bank Rate (with first-order autocorrelation correction) Dependent Variables

Independent Variables

~R

~p

~BR

Estimate (t-statistic)

Estimate (t-statistic)

Estimate (t-statistic)

-28.25 (0.24) -199.32 (0.94) -124.33 (0.21) 526.33 (10.32) -1366.81 (2.57) 3719.67 (6.49) -878.30 (1.44) -1981.79 (3.46) 1191.53 (2.07) 1119.74 (2.17) -183.78 (0.35) 240.20 (0.45)

-0.001 (0.59) -0.007 (1.68) -0.015 (1.12) 0.009 (7.82) -0.038 (3.10) 0.053 (4.05) -0.010 (0.81) -0.037 (2.96) 0.030 (2.35) 0.021 (1.82) -0.001 (0.08) 0.013 (1.08)

-0.008 (0.39) 0.609 (14.84) 0.317 (2.58) -0.026 (2.56) -0.081 (0.72) -0.036 (0.30) 0.083 (0.68) -0.058 (0.50) -0.081 (0.70) -0.008 (0.07) -0.027 (0.24) 0.202 (1.85)

Constant ~i ~r

~G

Mol M02 Mo3 Mo4 Mo5 Mo6 Mo7 Mo8

Although equation (11) does a good job of explaining changes in reserves with an R2 of 0.60, there is no evidence of a stock-adjustment mechanism. The coefficients for i, r, BD and R t - 1 are all insignificant. The explanatory power of the equation comes from the seasonal dummies together with gold flows and nonbankers' deposits. The insignificance of bankers' deposits probably results from a tendency for a purchase of securities to increase those deposits and reduce reserves. 4.3.2 Proportion The long-run model for the proportion is also converted to the short run by using a simple stock-adjustment model. The equilibrium-desired proportion is described by equation (12) and stock adjustment by (13). (12)

pD=bo -b 1 i+b 2 r+b 3 Y+El.

(13)

aPr =

'Yl[pf - Pr-l]

+ 'Y2 LlG + E2·

213

Bank of England Operations, 1893-1913

Table 4.1 (continued) Dependent Variables Independent Variables Mo9 Mo10 Moll

flR t -

1

flBD flNBD

flR

flP

flBR

Estimate (t-statistic)

Estimate (t-statistic)

Estimate (t-statistic)

1793.96 (3.45) -1931.22 (3.42) 109.68 (0.19) 0.03 (0.66) 0.07 (1.68) 0.25 (7.02)

0.027 (2.33) -0.016 (1.21) 0.011 (0.93)

0.210 (1.94) 0.056 (0.46) -0.048 (0.42)

-0.241 (4.09) -0.000 (1.11) -0.220 0.44 -0.70 0.037

-1.65) (3.10) 0.000 (1.25) -0.314 0.61 (D.W.) 2.06 0.331

flPr-l

flY -0.275 0.60 -0.61 1602.51

p

R2 h SE

Notes: fl first difference; R = Bank of England reserves; P = Bank of England proportion; BR = Bank rate; i = market yield on fortnightly loans; r = French market rate over market yield on fortnightly loans; G = gold flows into or out of the United Kingdom; Mol-Moll = seasonal factors; BD = bankers' deposits; NBD = nonbankers' deposits; Y = railway freight receipts (proxy for income).

Equation (14) is the solution for the observed proportion (14)

Pt

= "I1 hO - 'Yl bl i + 'Yl b2' + "I1b3Y + 'Y2 il G + ("11 E + E2)'

The error terms here have the same interpretation as in the model for reserves, and seasonal factors again are added. The results from estimating equation (14) in first differences are reported in table 4.1. They reveal a pattern similar to that for reserves. None of the factors explaining the equilibrium behavior of the proportion are significant. The coefficient for the lagged proportion is significant but negative. Gold flows have the expected sign and are significant. Several seasonal factors also are important. Although short-run models for reserves and the proportion explain a reasonable amount of the variance in those variables, the results do not show any evidence of a portfolio adjustment mechanism. Even after six

214

John Pippenger

Table 4.2

Short-Run Determinants of Reserves, Proportion, and Bank Rate (without first-order autocorrelation correction) Dependent Variables

Independent Variables

aR

ap

aBR

Estimate (t-statistic)

Estimate (t-statistic)

Estimate (t-statistic)

Constant

- 81.94 (0.56) -208.29 (0.97) -128.78 (0.22) 489.15 (9.77) -864.48 (1.59) 3537.34 (6.18) -146.97 (0.24) 1758.14 (3.09) 922.25 (1.59) 1200.92 (2.33) 59.66 (0.11 ) 214.27 (0.40) 1829.56 (3.53) 1867.23 (3.32) 278.10 (0.47) -0.09 (1. 71) 0.05 (1.45) 0.24 (7.00)

-0.002 (0.75) -0.008 (1.91) -0.015 (1.20) 0.008 (7.77) -0.032 (2.56) 0.047 (3.62) -0.001 (0.10) -0.039 (3.10) 0.029 (2.25) 0.026 (2.19) 0.002 (0.17) 0.012 (1.06) 0.030 (2.58) -0.014 (1.08) 0.011 (0.92)

-0.006 (0.21) 0.557 (13.14) 0.285 (2.33) -0.033 (3.15) -0.042 (0.36) -0.088 (0.72) 0.079 (0.64) -0.054 (0.46) -0.123 (1.04) -0.013 (0.12) -0.035 (0.32) 0.180 (1.62) 0.213 (1.94) 0.054 (0.43) 0.028 (0.23)

ai ar

aG Mol Mo2 Mo3 Mo4 Mo5 Mo6 Mo7 Mo8 Mo9 Mo10 Moll

aR t- 1 aBD aNBD apt-l

ay

R2 h

SE

Notes: See table 4.1.

0.59 -4.10 1634.42

-0.366 (6.53) -0.00005 (1.10) 0.43 -3.63 0.037

-1.63 (3.16) 0.0004 (1.08) 0.57 (D.W.) 2.56 0.347

215

Bank of England Operations, 1893-1913

lags are added for the relevant explanatory variables such as interest rates and income, there is no evidence of stock adjustment. The lack of any evidence of a stock-adjustment mechanism probably is due to three factors. The first is simultaneous-equations bais, which has already been discussed. The second is our inability to capture many of the short-run influences to which the Bank responded. For example, we can observe changes in reserves, but not the source of those changes. The Bank however had information about the source of reserve changes, whether they were internal or external and, if external, whether the gold came from France or South Africa. The third, and perhaps most important, factor is the evolution of the Bank's short-run operating procedures. The long-run objectives of the Bank and its equilibrium portfolio probably did not change very much from 1893 to 1913. The short-run operating procedures by which the Bank attempted to reach its equilibrium portfolio however, were continually evolving (see Sayers 1976, pp. 28--60). In the 1890s the Bank had a great deal of trouble making Bank rate effective. It resorted to manipulating gold points, borrowing from private depositors, and engaging in transactions that were very close to open-market operations. After 1907 it relied primarily on Bank rate and rarely used these other techniques. As a result of the evolving short-run operating procedures, the adjustment mechanism changed over time, making it difficult to identify a significant distributed lag for the long-run determinants of the portfolio such as deposits and interest rates. 4.3.3

Bank Rate

Even though they have not been successful, stock-adjustment models for reserves and the proportion can be justified on the grounds that portfolio adjustments are costly. This argument however seems much weaker for Bank rate and so a different approach is used here. As implied by the long-run results, Bank rate depends on domestic interest rates and foreign-relative-to-domestic yields. For the short run, two other factors are added. It is assumed that the Bank raises Bank rate when the desired proportion exceeds the actual proportion and when the domestic gold stock declines. (15)

BR t

= do + d1it + d2 r t + d3 [pf- Pt - d4 1i.G t + Ut ·

1]

Substituting equation (12) into (15) yields the solution for Bank rate. (15')

BR t = (do + d3 bo) + (d 1 - d3 b 1)i + (d2 + d3 b2 )r

+ d3 b3 Y - d4 1i.Gt + Ut )·

- d3 Pr-l

+

(d3 E l

216

John Pippenger

As in the earlier models, seasonal dummies are included and the equation is estimated in first differences. If one accepts an integrated world capital market, this model probably is less prone to simultaneity bias than the previous short-run models because of the linkage between the London and world capital markets. If however one accepts the conventional story in which high Bank rates induced capital flows and reduced nominal income, simultaneousequations bias could be very severe. The results from estimating equation (15') are reported in table 4.1. They strongly support the model. The 0.61 "R 2 is high for changes. All signs are correct and, except for income, all are significant. The high t-statistics for domestic yields may of course be due partly to a short-run influence from Bank rate to market rates. If the discount rate in France responded to Bank rate and the French discount rate influenced French market yields in the short run, then feedback also could run from Bank rate to r. It seems unlikely however that a rise in Bank rate would cause French market rates to rise by more than domestic rates, which is what is required for a positive coefficient for r. If Bank rate influenced the proportion and gold flow, then simultaneity presumably would work to reduce the coefficients for those variables, in which case their influence on Bank rate would be even stronger than indicated in table 4.1. The results support the conventional view that the Bank systematically used Bank rate to defend convertibility. A low proportion, gold outflow, or increased tightness in foreign financial markets caused the Bank to raise Bank rate. What remains unclear is what short-run influence Bank rate had on market rates, the proportion, gold flows, or income. Since weekly data exist for all of these variables but income, an analysis of their interdependence based on "causality" tests would be a fruitful project for further research. 4.4 Conclusions

Month-to-month changes in reserves and the proportion appear to be dominated by seasonal factors and external gold flows. There is no evidence that in the short run they responded to bankers' deposits, domestic yields, or foreign-relative-to-domestic interest rates. The absence of any evidence for stock adjustment probably is due to several factors, including simultaneity, important unobserved variables, and changing short-run responses to portfolio disequilibrium. The evidence for Bank rate however is consistent with the conventional wisdom. Both a gold outflow and a high desired-relative-to-actual proportion tended to result in a higher Bank rate. Analysis of long-run operations of the Bank strongly support the conventional wisdom about the gold standard. The Bank held fractional

217

Bank of England Operations, 1893-1913

reserves against deposits, which implies that it played by the rules of the game and bought securities as deposits and reserves increased. The Bank also actively protected convertibility by increasing both reserves and the proportion as French market rates rose relative to domestic yields. The evidence also indicates that the proportion rose as nominal income increased. In the long run Bank rate appeared to be dominated by domestic market yields. The evidence however, indicates that Bank rate rose as foreign rates increased relative to domestic yields. The conventional wisdom tends to forget that the Bank was private and had to pay dividends. The results show that, like a normal commercial bank, the Bank of England reduced reserves and the proportion as interest rates increased. The tendency to reduce the proportion as interest rates rose and raise it as nominal income increased meant that the Bank followed conflicting policies over the business cycle. This behavior and the inverse relation between securities and reserves implied by the monetary approach helps explain why Bloomfield (1959) and others have concluded that the Bank did not play by the rules of the game.

Appendix

Spectral Patterns

Figure 4.A.1 shows spectral estimates for monthly gold flows and changes in bankers' deposits. The variance in changes in bankers' deposits is dominated by cycles of one year or less with a dominant peak at three months and smaller peaks at four and six months. The dominant elements for gold flows are a strong annual and threemonth cycle. Unlike bankers' deposits, spectral estimates do not drop off sharply for cycles longer than one year. Instead estimates tend to drop off for cycles shorter than three months. The spectrum for differences in gold flows-which is used in the regressions-does fall off rapidly for cycles longer than one year. Figure 4.A.2 shows spectral patterns for monthly changes in reserves and total-earning assets. The spectrum for securities tends to decline as cycle length increases with peaks at about 2.3 and 3.0 months and another at the one-year cycle. Estimates for reserves decline somewhat beyond one year, but reveal much stronger seasonal elements. There are significant peaks at about 2.3, 3.0, 6.0, 9.0, and 12.0 months. Figure 4.A.3 shows the spectrum for monthly changes in sixty-day U.K.-Treasury-bill rates from 1897 to 1908 and, for comparison, the spectrum for monthly changes in the ninety-day Treasury-bill rate from 1959 to 1970. The pre-World War I period is restricted to the period 1897-1908 so that both periods have approximately the same number of observations. The spectrum for the post-World War II data is essentially

218

John Pippenger SPECTRUM 10

5.0

,\I' , ,

1.0

0.5

V

I

1 \ 1 \ 1 \ \ I \ \ I \ I \ I I 1

,\

1\ I \ I , I \

,

'/ ' 1 \.J

GOLD FLOWS

\ \

90 PERCENT CONFIDENCE INTERVAL

\ \

I

,, ,,

\ \

\

\

\

v

I I I I

,

0.1

0.05

0.01...........~~--r-----r------r"--r----,-----­ LENGTH OF CYCLE 2.5 4 .. 25 12 6 IN MONTHS

Fig.4.A.l

Spectral estimates for monthly gold flows and changes in bankers' deposits.

flat. Although estimates seem to rise up to about the one-year cycle and then decline, the Treasury-bill rate behaved essentially like a random walk. Spectral estimates for the gold standard era also increase as cycle length increases, but then drop rapidly for cycles longer than one year. The pattern implies structure in changes in short-run rates under the gold standard. In addition, spectral estimates are much higher at every frequency for the gold standard era, indicating greater variability in interest rates in both the long and short run. In order to combine the two series effectively, figure 4.A.4 shows the spectral-density estimates for changes in the proportion and changes in the log of the Bank rate. Once again the short run tends to dominate and estimates drop off rapidly beyond the one-year cycle. The proportion has a very strong peak at three months and Bank rate has a definite peak at one year. The decline in spectral estimates for the Bank rate for cycles shorter than about two-and-one-half months probably reflects the inertia

219

Bank of England Operations, 1893-1913 SPECTRUM

20

5.0 95 PERCENT CONFIDENCE INTERVAL

\ \ \

"

1\

I

I~ I \ \

I

\

\

1 \

\

\ \

1.0

\

\ I \ I \1 V RESERVES

I

\

I

\ \

I I

\ I \ I \ I \ I \ I \I

0.5

\

\

O.l-t--.----,-----r------r---~-__r_---__r_-----

25

Fig.4.A.2

12

4

2.5

LENGTH OF CYCLE IN MONTHS

Spectral estimates for monthly changes in total securities and reserves.

in the rate generated by the Bank's concern for domestic trade and its reluctance to make small changes (see, for example, Sayers [1936] 1970, pp. 50-54). The spectra for monthly changes in most of the series tend to be dominated by high-frequency components. The spectral estimates also show that, except for short-term interest rates, most series have strong seasonal components.

Notes 1. No Durbin-Watson statistics are reported for these regressions. 2. Goodhart's series for bankers' balances is not for the same week in the month as his

220

John Pippenger SPECTRUM

400

1897·1908

100

95 PERCENT CONFIDENCE INTERVAL

50

10

5.0

1.0-+------------------...-------------, 12

6

3

LENGTH OF CYCLE IN MONTHS

Fig.4.A.3

Spectral estimates for changes in Treasury-bill rates, 18971908 and 1959-70.

other series for the Bank. I therefore used data from the U.S. National Monetary Commission (1910) and information supplied by Goodhart to construct series for the Bank that correspond to the same day of the month as his series on bankers' balances. Bank rate, but not the proportion, is taken from Goodhart's book. 3. The use of nonoverlapping annual averages of monthly observations smooths the data and eliminates most of the short-run noise. When monthly data at twelve-month intervals are used, the signal-to-noise ratio drops and most of the significance disappears. 4. Nonbankers' deposits also include the small item under liabilities called "seven day and other bills." 5. If the Bank held more reserves for bankers' deposits than other liabilities, the proportion would be positively related to those deposits. In that case, Ycould be acting as a proxy for BD. That, however, does not appear to be the case because, when bankers' deposits replace the proxy for income, both the R,2 and D.W. decline.

221

Bank of England Operations, 1893-1913 SPECTRUM

3.0

PROPORTION

1.0 BANK

f\ RATE

I I I I

0.5

0.1

I

I

I

I

/

/

/

95 PERCENT CONFIDENCE INTERVAL FOR WHITE NOISE

, ,,, , ,, , ,,,

0.05

O.O-+----,r----r----.,..--~--_,------~----

25

12

Fig.4.A.4

4

LENGTH OF CYCLE IN MONTHS

Spectral-density estimates for monthly changes in bank rate and proportion, 1893-1913.

6. For those who are interested in the behavior of the variables, their spectral patterns are shown in the Appendix. 7. The Bank's balance sheet gives data for the second week of the month while gold flows are reported for the calendar month. As a result changes in reserves from the second week of June to the second week of July are regressed against the gold flow in June minus the gold flow in May. The same applies to regressions for the proportion and a similar situation also exists for Bank rate.

References Bloomfield, Arthur. 1959. Monetary policy under the international gold standard. New York: Federal Reserve Bank of New York.

222

John Pippenger

Goodhart, C. A. E. 1972. The business of banking. London: Weidenfeld and Nicolson. Sayers, R. S. [1936] 1970. Bank of England operations, 1890-1919. Reprint. New York: Greenwood. - - . 1976. The Bank of England, 1891-1944. Vol. 1. Cambridge: Cambridge University Press. U.S. National Monetary Commission. 1910. Statistics for Great Britain, Germany, and France, 1867-1909. Washington, D.C.: Government Printing Office.

Comment

Charles A. E. Goodhart

John Pippenger quotes Richard Sayers on the three main duties of the governor of the Bank of England in the pre-1914 period: to maintain convertibility into gold, a political duty to look after the financial needs of government, and to maintain an income for the stockholders. Such stockholdings ceased in 1946 when the Bank was nationalized, and the fixed-exchange-rate system was abandoned in 1972. But the importance, and at times the difficulties, of the Bank's relationships with the politicians in Whitehall have increased over time. So it is something of a pleasure for me to turn back again to examine the history of a period when it could be said that the operations and objectives of the Bank were largely independently decided within the context of the Bank Act of 1844. My reading of the papers by John Pippenger and John Dutton leads me to the view that there is a large measure of common agreement among us on the question of the actual way in which the Bank operated-its positive actions. There may be less certainty, or agreement, on what were the reasons, the normative motives, that led to such behavior. I would characterize the Bank during this period as having a hierarchy of objectives, a lexicographical utility function in our jargon. Although it was only on rare occasions a matter of concern, the fundamental objective of the Bank was, I believe, the maintenance of the basic fabric and structure of the banking and financial system. It is worth remembering that the Bank Act was suspended at times of extreme crises and that the ultimate responsibility of the Bank was to the stability of the financial system itself, not to the gold standard. Under normal circumstances, however-and, with the exception of the Baring crisis, the whole period up till the outbreak of war in August 1914 could be described as normal-the most important function of the Bank was to protect the convertibility of its notes into gold, i.e., to maintain the Charles A. E. Goodhart is a senior policy adviser for the Bank of England.

223

Bank of England Operations, 1893-1913

gold standard. The instrument(s) which it predominantly used for this purpose were short-term interest rates, administered changes in Bank rate and market operations to make Bank rate effective, i.e., to keep market rates in line with Bank rate. The gold devices were a rather lesser adjunct. Most earlier writers have seen the Bank as adjusting its short-term interest rates solely in response to variations in its liquidity position (i.e., the interest-rate instrument was assigned to the protection of convertibility). Dutton mentions Ford and Bloomfield as taking this view, and in my own work (1972, p. 207) I also related changes in Bank rate only to the Bank's liquidity position. Dutton has now challenged this view since he finds that domestic cyclical variables, especially unemployment, had a perhaps minor but nonetheless significant effect on the decision to make Bank-rate changes. While I accept Dutton's carefully worked econometric findings, it is nevertheless possible to interpret them in more than one way. One interpretation is that there was a trade-off in objectives. For a given loss of reserves, the Bank would raise Bank rate by more (less) if domestic activity was higher (lower). But another possible interpretation was that the state of activity in the country was treated by the Bank as an indicator of the risk of future gold drains, either internal or external, from its reserves. Subject to the above qualification, it is, I hope, generally accepted that the Bank played the gold standard game faithfully in respect of varying short-term interest rates in response to changes in its liquidity position. In my earlier work I noted that the elasticity of external gold flows in response to relative interest-rate changes was sufficient to provide some accommodation of monetary changes to domestic activity. More interesting, I felt, were my findings that bankers' balances at the Bank were not related to the reserves or the proportion in the Banking Department, but were related to the level of activity. This finding suggested to me that the Bank must have accommodated the demand for cash by buying more securities at times when the reserves and proportion were low and interest rates were high. These results are now supported by Dutton and Pippenger. "Bank-reserve decreases seem to have led to increases in Bank holdings of interest-earning assets. Instead of amplifying the effects of reserve changes on the money supply, the Bank seems to have sterilized them" (Dutton, p. 192). Pippenger also reports that the proportion (of gold reserves to liabilities in the Banking Department) fell when U.K. interest rates were higher. "The Bank systematically reduced liquidity in order to earn income" (Pippenger, p. 209). Having assigned to interest rates the task of protecting convertibility, the Bank's market operations provided a procyclical impulse to the monetary base in the United Kingdom. This conclusion about facts, reached from differing starting points by the three of us, is, I believe, an important finding.

224

John Pippenger

Why did the Bank behave this way? Here there are a range of differing interpretations. The first, espoused by Dutton, is that the setting of interest rates by a central bank leaves it, during the period such rates are fixed, passively responding to demands for accommodation at that rate. Pippenger notes the inertia in Bank rate and the Bank's reluctance to make small changes. Milton Friedman has regularly complained that the preference of central banks for setting interest rates rather than the monetary base imparts a procyclical bias to monetary changes. It is interesting to find that this may have been the same in the United Kingdom under the gold standard, as subsequently. The second hypothesis, strongly advanced by Pippenger, is that the Bank was concerned to maintain its profits for its stockholders, thus "the commercial duties of the Bank promoted a procyclical monetary policy" (Pippenger, p. 210). There is certainly evidence that the Bank was concerned to maintain (though not to seek to maximize) its profits during this period, and I do not doubt that Pippenger's preferred explanation has some weight (as also does Dutton's inertia hypothesis). My own reading of the literature, however, makes me doubt whether profitability concerns played quite such a major role as Pippenger suggests. I would, however, tentatively offer a slightly different, but broadly similar, reason for the Bank's behavior. In the earlier part of the nineteenth century, the Bank had achieved a position of market dominance as much (or more) because of its size as from its strategic central position as holder of the main gold reserves and lender-of-Iast-resort. As the nineteenth century progressed, the relative size of the Bank fell progressively compared with the joint-stock banks around it, and, with the growth of the main London clearing banks, in part by a series of mergers, the Bank came to feel dwarfed and even threatened in power by them. So, in addition to profits, the Bank may have retained some concern with market share and size. As the size of the bill market increased, other things being equal (i.e., with interest rates raised high enough to maintain convertibility), the Bank, retaining a more than purely residual banker's instinct, would have wanted to keep its share. If a short digression may be permitted, it was quite largely this fear of being left face-to-face in market operations with the overmighty subjects, in the form of the London Clearing Banks (LCB), that led the Bank in the nineteenth century to encourage and sustain the discount houses, to provide institutional buffers between itself and the LCB. If Americans try to imagine a world in which the Fed had to contend with, say, eight colossal banks throughout the United States dominating the banking system, they might also appreciate the need for buffers in such circumstances. It has been fascinating to me, as a monetary historian, to find that the provisions for revising monetary operations issued by the Bank in August 1981 restated the Bank's preference for dealing through the

225

Bank of England Operations, 1893-1913

discount houses and the bill markets, and for exactly the same reasons-a disinclination to deal face-to-face with the huge clearing banks-as had held in the nineteenth century. Plus ~a change, plus c'est la meme chose; perhaps a comforting maxim for central bankers! There could be, I suppose, a third hypothesis to account for the Bank's behavior in this respect-that it was consciously following a procyclical policy, e.g., in order to support "the needs of trade." Although there was, as Dutton notes, some sensitivity in the Bank to the effects of its policy on business conditions, my own assessment is that this hypothesis is the least likely of the three to account for the procyclical variations in the Bank's security holdings. But I have not the evidence, either from contemporary accounts or econometric test (how could this be set up?) to discriminate between these hypotheses. Since I interpret Pippenger's results, as well as Dutton's, as closely in accordance with my own, why then does Pippenger think he is challenging my findings? You must ask him. Perhaps he thought my claim that "the Bank must have regularly accommodated, to some large extent, variations in the demand for cash" (1972, p. 206), implied that the Bank's motive was consciously to do so. If so, he read too much into those words, for it remains entirely consistent with my position for such accommodation to have been the unconscious result of other sources of action, including Pippenger's profit motive. Nonetheless there are some differences in approach between us. Subject to the gold devices, the maintenance of convertibility required the Bank to buy or sell gold at fixed, known prices. So in the short run, changes in the Bank's reserves were determined by others, not by the Bank. It was not, within this time frame, a choice variable that the Bank could determine. The Bank's choice variables were Bank rate and its market operations. Thus I would follow Dutton in setting up equations in which Bank rate or securities holdings are functions, inter alia, of the level of reserves or the proportion. It may be that this concentrates attention on the shorter run: indeed, my reason for including a time variable in my own equations was to try to eliminate the influence of long-run common trends. I hope Pippenger would accept the above. I think he criticizes me for ignoring the possible effect of longer-term changes in the Bank's demand for reserves. But since any individual short-run observation will tend to be off the Bank's underlying demand function, he has to transform his data into annual averages of monthly data to try and capture his longerterm relationships, so both the time period and data base of our studies were rather diferent. In any case, I find no difficulty in accepting his findings of the relationships between U.K. and French interest rates and the Bank's reserves and proportion, though I would interpret the direction of causality somewhat differently. A problem with his longer-term

226

John Pippenger

approach (see the equation on p. 208) is that both the dependent and (some of) the independent variables will have been growing over time, so that deflation by a scale variable (perhaps preferable to my use of a time dummy) would help to remove common trends. Even in the case of his equation for the proportion (see p. 209), there was a rise between 1893 and 1913 (from 45.7 to 48.5), so that when time is entered as an additional variable, the significant positive coefficient on Y (which appears to contrast so sharply with my own result), disappears.! What Pippenger's reformulated equations have done, for me at least, is to raise the question of the determinants of deposits in the Banking Department that were not bankers' balances, NBD in his terminology, and also whether the Bank responded differently to changes in BD and NBD. For example, in the second part of his paper where he sought to examine short-run relationships, Pippenger found a significant positive relationship between monthly changes in the Bank's reserves and in NBD. One possible explanation may be that most of the main gold dealers in London had accounts at the Bank (though there could be some question in one or two cases whether the accounts would be classified as a banker's balance or not), so payments for gold purchases (sales) might in the first instance be met by crediting (debiting) an NBD account? In private correspondence with me, John Pippenger also raised the possibility that the Bank may have felt a stronger obligation to hold reserves against NBD than against BD, perhaps on the grounds that the Bank could rely on the banks' need to hold some minimum level of operational balances. Although the Bank did have an estimate of the latter,3 I doubt that it is a likely possibility. First, it would suggest that the level of reserves was more under the short-term control of the Bank than, I believe, is justified. Second, it hardly squares with the relative sizes of the coefficients for BD and NBD in Pippenger's earlier equations (p. 208). Be that as it may, Pippenger's work suggests that the behavior of nonbankers' deposits (NBD) may also be worth studying. Whether we will also obtain more enlightenment from his spectral patterns, I find more difficult to tell; I did not get much from them. Both Pippenger's and Dutton's papers applied econometric methods to examine the historical behavior of the Bank. Amidst the differences of econometric techniques and some academic disputations, I haveperhaps hopefully-perceived an emerging consensus of views about the positive facts of-if not the normative motives for-such behavior. That is an advance. Notes 1. My thanks are due to John Pippenger for running these extra regressions for me.

227

Bank of England Operations, 1893-1913

2. lowe this information to A. F. A. Carlisle who found a record of these accounts in the Bank's archives. 3. See Goodhart 1972, pp. 109-17 and Presnell 1968, pp. 167-228.

References Goodhart, Charles A. E. 1972., The business of banking. London: Weidenfeld and Nicolson. Presnell, L. S. 1968. Gold reserves, banking reserves, and the Baring crisis of 1890. In Essays in money and banking in honour of R. s. Sayers, ed. c. R. Whittlesey and J. S. G. Wilson. Oxford: Clarendon Press.

General Discussion of Dutton and Pippenger Papers MUNDELL made several points drawn from the session as a whole. He noted that it was -an historical occasion to hear a discussion by so many economists of an issue that had been close to his heart for some time. From his casual knowledge of economic history he believed that the gold standard has never received so much attention by so many economists in such a concentrated period of time. This testifies to the importance of the subject to the economics profession. He was struck by two key issues. One was the way in which the international adjustment process worked under the gold standard; the other was the overall global approach to the nineteenth century gold standard. He thought that the transfer problem was swept under the rug at first, but it quickly won attention when empirical issues were discussed. It is important to realize that the transfer problem itself is crucial for an understanding of the gold standard. Thus periods in which international transfers were being made should be used to illustrate the economic events of those years. Mundell stressed that the transfer problem exists even outside the context of capital movements, reparations payments, or other unilateral transfers. It exists by the very idea that money is transferred from one group to another, which is a shift of purchasing power from one country to another, and a gold flow accompanies the shift either as cause or effect. There is an accompanying transfer of real resources because the country that receives gold has to have income in excess of its expenditures, and the deficit country-the gold exporting country-to effect the transfer has to have expenditure in excess of its income. That is the heart of the earlier mechanism of Gervaise up through the absorption approach to the transfer problems in the balance of payments in the 1950s.

228

John Pippenger

In the literature over the past fifty years there is a great controversy as to whether, in order to effect a transfer, relative prices must change, and whether that change comes about in the terms of trade or in the ratio of international-to-domestic prices. Even Viner at some point understood that there are cases in which relative changes are not crucial to the movement of transfers. Take as an example a small country in the world economy that has to lend or pay money to another small country. With perfect capital and goods mobility, there is no need for any changes in prices. There is simply a shift of expenditures, without any further effects. This shift is a crucial part of the mechanism quite apart from pure capital flows. Liquidity flows, then, are the starting point; secondly, there is the transfer problem; and thirdly, the issue of long-term lending flows or transfers that was involved in the discussion with Brinley Thomas and Moe Abramovitz. In that context of the long swings of economic activity, Britain lends to America, the lending is a result of an expansionary boom in the United States like the railway boom, capital flows to the United States, but the money or the trade transfer is not as large as the capitalmovement transfer. The explanation comes in the correct solution of the transfer problem, which was discussed most exactly and very precisely in the literature of the 1930s-not in the literature of the Keynes-Ohlin controversy but rather in the literature following the review by Sir Dennis Robertson of Viner's Studies in the Theory of International Trade (1937). According to Viner, in the discussion of where the gold goes, the key issue is what happens to the demand for money. If the demand for money increases in the gold-receiving country, then the gold has to go to the receiving country. The crucial distinction is whether the demand for money is a function of domestic expenditure or a function of national income. Robertson took the position that the demand for money is a function of national income while Viner took the opposing position that the demand for money is a function of domestic expenditure, which he cans "final purchases" and which Keynes and Meade later on call domestic expenditure. FRATIANNI pointed out that both the Humean price-specie flow and the monetary approach to the balance of payments suggest that following an expansion of the domestic component of the monetary base, there will be an outflow of gold. In this view, causation runs from domestic credit expansion to gold flows. The reaction-function framework used by Dutton and Pippenger suggests the opposite; according to that framework, the monetary authorities adjust through purchases and sales on the open market to changes in gold flows. The data clearly cannot discriminate between the two views. Thus, the crucial unanswered question is how to go about differentiating between these mechanisms. Fratianni also raised the question of whether, in Dutton's paper, right-hand-side variables such as gold flows are truly exogenous in the

229

Bank of England Operations, 1893-1913

sense that changes in the independent variables occur before changes in domestic credit expansion. This question is crucial for differentiating between the two possible interpretations of the results. MCGOULDRICK also expressed the opinion that the correlation between Bank rate and gold flows results from the influence of the first variable over the second. DUTTON responded that econometrically the right-hand-side variables are completely predetermined. Whatever the Bank of England chooses to do at that point is a result rather than a cause. MUNDELL stated that whether a positive correlation between gold flows and changes in the Bank's domestic assets was evidence of passive behavior of the Bank or evidence of deliberate policy was important and similar to the debate for the 1945-71 period. PIPPENGER responded to Goodhart by arguing that the idea that the Bank of England controlled interest rates in London over any substantial period of time is almost inconceivable. Throughout the nineteenth century, London was the center of an international capital market of vast proportions; the Bank of England was only one relatively small bank in London, a city that was only one part of a large international capital market. Changes in Bank rate merely reflect the fact that the Bank of England sometimes found itself in portfolio disequilibrium. One of the ways the Bank restored portfolio equilibrium in periods when its actual proportion fell below desired levels was to raise Bank rate relative to discount rates in order to discourage discounting at the Bank of England. This maneuver would reduce the Bank's holdings of assets and raise its proportion. The idea that the Bank of England could use Bank rate to control interest rates in Great Britain would imply that over the longer run, if its choice of Bank rate was inappropriate, then its proportion would explode to plus or minus infinity. GOODHART responded that there was not that much difference between Pippenger's and his own views. Pippenger concentrates on the long run, whereas Goodhart in his earlier work had concentrated on the short run. In Goodhart's view, Pippenger's analysis of the long run is correct. Under a fixed-exchange-rate system, there is no way that in the long run the United Kingdom could have maintained interest rates out of line with those in the rest of the world without the Bank of England's proportion moving up or down endlessly. In the short run, on the other hand, the Bank of England actually did exert some control over interest rates. PIPPENGER disagreed with Goodhart even for the short run. In his view, the Bank of England could no more control market interest rates in London in the 1800s than the Federal Reserve Board can control interest rates in New York today. But this statement is different from saying that the Bank of England cannot influence interest rates--eontrol and influence are different matters. Pippenger agreed that the Bank of England

230

John Pippenger

was capable of influencing interest rates. However, if the Bank of England had attempted to set an inappropriately low rate, it would have been inundated with borrowing and would have found its proportion going to zero. If it had set Bank rate at 20 percent, it would have found that no one borrowed from it, and furthermore that it was earning no income and might go bankrupt. Therefore, it could not act independently of market rates, even in the short run. GOODHART responded by pointing out that the Bank of England actually operated in markets to try and make Bank rate effective. On many occasions in the nineteenth century, there are indications of concern within the Bank of England that it was not able to make Bank rate effective, at least until the Bank developed mechanisms by which it could affect the amount of cash available in the market. Of course, the range of freedom of any central bank-if not the Fed then certainly the central bank of a relatively small open economy-is severely limited. The Bank of England cannot go out tomorrow and set rates at 30 percent any more than it can go out and set them at 2 percent. But if Bank rate today is 13, then the Bank of England can surely change it up to 15 or change it down to 11. It could do that in the nineteenth century as well. FREEDMAN returned to the distinction between the long run and the short run. He pointed out that Goodhart and Pippenger seemed to nave agreed that in the long run the Bank of England was incapable of maintaining a Bank rate out-of-line with interest rates elsewhere. But that conclusion depends on assumptions about the response of other central banks. If the Bank of England lowered Bank rate and was inundated with discounts, it is important to know how the Bank of France and other central banks responded to the Bank of England's initiative. If they adjusted their discount rates in the same direction, then it was at least conceivable that the Bank of England could impose its rate on the rest of the world. DORNBUSCH pointed out that the Bank of England never actually discounted at Bank rate. It posted Bank rate but intervened at rates very close to market rates. The actual rates at which transactions took place in any given week took the form of a distribution centered around the market rate. Although Bank rate was infrequently moved, the effective rate of interest charged by the Bank moved week-to-week. MUNDELL suggested that it maybe useful to divide the nineteenth century into two parts when addressing the question of speeds of adjustment. Conditions certainly changed with the invention of the telegraph. The ability of the telegraph to increase the speed with which information was diffused permitted a global market to become established. Mundell commented also on another difference between the pre-1870 and post-1870 periods, namely, the introduction of flexible silver prices and the bimetallic system, which broke down after 1870. There were

231

Bank of England Operations, 1893-1913

really two worlds: one of silver standard countries and another of gold standard countries. That made a very large difference in the interpretation of the two periods. EICHENGREEN addressed the seemingly paradoxical results concerning changes over time in the sensitivity of the Bank of England to internal economic conditions. According to Dutton's paper, the Bank of England became less sensitive to changes in internal conditions after 1890, which seems counterintuitive. Eichengreen reported some results by Richard Grossman of Harvard University, who attempted to estimate a similar function for the period 1925-31, finding that the Bank of England again became less sensitive to changes in internal conditions after the First World War. This result is clearly inconsistent with the vast majority of historical studies that indicate the Bank often hesitated to raise Bank rate in the interwar period for fear that its action might injure British industry or arouse Treasury officials. It is certainly conceivable that the vast majority of historical studies are simply wrong, but it is also possible that the reaction-function methodology, the specification, or the empirical techniques adopted by the authors is inappropriate. One way in which the specification may be deficient is that it fails to recognize the existence of nonlinearities in the relationship of Bank rate to internal economic conditions. Such nonlinearities were recognized at the time by the Bank of England: for example, Bank officials apparently believed that Bank rate had to exceed a certain crucial threshold-usually taken to be 4.5 percent-before it began to affect short-term interest rates, and that only when Bank rate remained above that threshold for extended periods were long-term market rates of interest affected. Although commercialbank overdrafts were extended at rates 0.5 to 1 percent above Bank rate, and although exceptions were sometimes made for favored customers, these rates were normally subject to a floor of about 5 percent. Thus, so long as Bank rate remained at or below 4.5 percent, it could be argued that domestic-credit conditions were unaffected by measures designed to attract gold and short-term capital inflows. Such nonlinearities could be readily incorporated into the reaction-function framework, but the authors' failure to model such institutional detail may bias their results toward insignificance. MCCLOSKEY argued that the Bank of England's actions only matter if one believes in the price-specie-flow mechanism. That theoretical construct is the intellectual origin of the question of whether the Bank played by the rules of the gold standard game. If one does not believe in the price-specie-flow mechanism, then the question of whether the Bank played by the rules of the game or sterilized gold flows and whether it raised or lowered Bank rate in response to changes in its financial position is irrelevant to the question of what determined the level of interest rates in England and the rest of the world.

232

John Pippenger

GOODHART responded to other discussants by pointing out that Bank rate, at least from 1900 onwards, was varied with frequency. While the Bank of England was concerned to make Bank rate effective, it is also true that movements in Bank rate were regarded as signals and that most of the Bank's business, as Dornbusch noted, was done at rates related to existing market rates. However, when the Bank wished to signal a change of gear, to use Richard Sayers's phrase, it would raise or lower Bank rate and then try to enforce the change in market rates by operating to do so. PIPPENGER concluded by summarizing his view that the way the Bank of England behaved in the long run is straightforward and fits the conventional view with only minor modifications. What is interesting, therefore, is how the gold standard worked and the Bank's role in its operation, particularly in the short run. DUTION concluded with a dissent from the view that the Bank of England's operations had no impact. Even if the Bank had no control over the money supply, Dutton argued, it still retained an influence over the proportion of the money suply that was backed by gold. In any case, it had some control over its gold stocks and had a desire to defend their level. From reading descriptions of the period, Dutton suggested, it certainly appears that in the short run at least, the Bank of England believed Bank rate could be used to change London money-market rates relative to rates in the rest of world, and, by so doing, that it could induce capital to flow among financial centers. Bank rate did have some effect, despite being simultaneously determined with other interest rates. It could be and was used as a policy tool.

5

The Gold Standard and the Bank of England in the Crisis of 1847 Rudiger Dornbusch and Jacob A. Frenkel When there occurs a state of panic-a state which cannot be foreseen or provided against by law-which cannot be reasoned with, the government must assume a power to prevent the consequence which may occur. -Sir Robert Peel (1847)1

5.1

Introduction

The acceleration of world inflation during the 1970s along with the rise in the rate of unemployment and the general instability of money and prices have renewed interest in the operation of the gold standard. Recent proposals for a return to some variant of the gold standard stem from the belief that a return to such a standard will restore macroeconomic stability. The belief is based on a casual look at history with the consequent inference that the gold standard contributed to the stability of the system. That view of price stability was supported by Keynes who argued in Essays in Persuasion: The course of events during the nineteenth century favoured such ideas.... The remarkable feature of this long period was the relative stability of the price level. Approximately the same level of price ruled in or about the years 1826, 1841, 1855, 1862, 1867, 1871, and 1915. Prices were also level in the years 1844, 1881, and 1914.... No Rudiger Dornbusch is professor of economics at the Massachusetts Institute of Technology and a research associate of the National Bureau of Economic Research. Jacob A. Frenkel is professor of economics at the University of Chicago and a research associate of the National Bureau of Economic Research. This paper is a revision of one presented at the conference "A Retrospective on the Classical Gold Standard, 1821-1931," at Hilton Head Island, South Carolina in March 1982. An earlier version was presented at the Workshop in Economic History at the University of Chicago in October 1971. The authors are indebted to Eliana Cardoso, Karl Brunner, Charles Goodhart, and the conference participants for helpful comments and to Lauren Feinstone and Alberto Giovannini for research assistance. A special debt is owed to Anna Schwartz for drawing the authors' attention to the important role of public deposits. Financial support was provided by grants from the National Science Foundation. This research is part of the NBER Program in International Studies and Business Fluctuations. The views expressed are those of the authors and not necessarily those of the NBER, Inc.

233

234

Rudiger Dornbusch and Jacob A. Frenkel

wonder that we came to believe in the stability of money contracts over a long period. The metal gold might not possess all the theoretical advantages of an artificially regulated standard, but it could not be tampered with and had proved reliable in practice. (Keynes 1932, pp. 88-89) But another view of the history of the gold standard during the nineteenth century reveals a succession of crises of varying length and depth. As documented by Hyndman [1932] 1967, the nineteenth century in the United Kingdom witnessed at least eight serious crises: in 1825,1836-39, 1847, 1857, 1866, 1873, 1882, and 1890. Given this perspective of the gold standard era the relevant question should be not only how the gold standard worked but also why did it fail. The origins of the various crises during the gold standard era vary. Some were "real" and some were "financial," some autonomous (like a massive harvest failure), and some induced by mistaken policies. Of course, no proponent of the gold standard has suggested that it would eliminate harvest failures. The question, therefore, is whether and to what extent the policies that are induced by the rules of the game mitigate or exacerbate the severity of crises. Our paper examines the 1847 crisis in Great Britain. That year, well documented by parliamentary inquiries, is of special interest because the origin of the crisis was "real." A harvest failure gave rise to commercial distress and financial panic, the extremity of which was remarkable. Our analysis examines the operation of the gold standard, the policies of the Bank of England, as well as the speed and extent of international adjustment in the form of gold and capital flows. The paper proceeds as follows: Section 5.2 provides a brief account of the main events in the United Kingdom during 1847. Section 5.3 studies the institutional setting of the gold standard and spells out a formal model of the financial markets. The two crises of April and October 1847 are studied in section 5.4. In section 5.5 we discuss whether suspension of Peel's Act was necessary. The paper concludes with some observations on the gold standard as a monetary system. 5.2

Outline of Events

The events of 1847 were initiated by a major harvest failure in Ireland and England in 1846. The shortage of domestic food supplies led to large price increases and trade deficits which in turn brought about an external drain of bullion from the Bank of England. These developments occurred against the background of the "railway mania" which commenced in 1845. The railway mania along with the food shortage resulted in a massive financial crisis, the analysis of which is the subject of this paper. The characteristics of the 1847 crisis were stated by John Stuart Mill:

235

The Bank of England in the Crisis of 1847

It is not, however, universally true that the contraction of credit, characteristic of a commercial crisis, must have been preceded by an extraordinary and irrational extension of it. There are other causes; and one of the more recent crises, that of 1847, is an instance, having been preceded by no particular extension of credit, and by no speculations; except those in railway shares.... The crisis of 1847 belonged to another class of mercantile phenomena. There occasionally happens a concurrence of circumstances tending to withdraw from the loan market a considerable portion of the capital which usually supplies it. These circumstances, in the present case, were great foreign payments, (occasioned by a high price of cotton and an unprecedented importation of food,) together with the continual demands on the circulating capital of the country by railway calls and the loan transactions of railways companies. . . . This combination of a fresh demand for loans, with a curtailment of the capital disposable for them, raised the rate of interest, and made it impossible to borrow except on the very best security. Some firms ... stopped payment: their failure involved more or less deeply many other firms which had trusted them; and, as usual in such cases, the general distrust, commonly called a panic, began to set in, and might have produced a destruction of credit equal to that of 1825, had not circumstances which may almost be called accidental, given to a very simple measure of the government (the suspension of the Bank Charter Act of 1844) a fortunate power of allaying panic, to which, when considered in itself, it had no sort ·of claim. (Mill 1871, bk. 3, chap. 12, section 3) Table 5.1 reports selected data for the period 1845-48. 1847 was characterized by a deterioration in the balance of trade and the terms of trade as well as by a significant rise in the price of wheat and the other price indexes. The trade-balance deficit caused gold outflows and an accompanying reduction in the supply of Bank of England liabilities and credit. While Table 5.1

Selected Data for Great Britain, 1845-48

Exports Imports Trade balance Terms of trade Price of wheat Price of agricultural products Price of industrial products

1845

1846

1847

1848

69.4 88.4 -19.0 119.6 50.8 120.0 99.0

67.0 87.3 -20.3 115.1 54.7 118.0 99.0

70.5 112.1 -41.6 112.5 70.0 125.0 104.0

61.2 88.2 -27.0 121.7 50.5 107.0 92.0

Source: All data are from Mitchell 1962. Notes: The balance-of-payments data are measured in £ million; the terms of trade are an index of the net barter terms (1880 = 100); the price of wheat is in shillings per imperial

quarter; and the prices of agricultural and industrial products are the Rousseaux price indexes (1885 = 100).

236

Rudiger Dornbusch and Jacob A. Frenkel

the external bullion drain was one direct consequence of the harvest failure, a second one was the extensive commercial failures arising from speculative forward purchases of foodstuffs for delivery in mid-1847. By the time these contracts came to maturity, a good harvest for 1847 was expected, and that change in expectations led to a drastic decline in spot prices and default of many trading establishments. In addition, the precarious financial position of many enterprises that had taken part in the railway speculations reduced confidence in the financial integrity of the system and resulted in an internal drain of bank reserves. Figures 5.1-5.3 show the weekly data for the stock of bullion in the Bank of England, the Bank's note reserve, the stock of notes held outside the Bank, and the price of consols (the sources of the data are listed in the Appendix). T,wo major crises occurred during the year. The April crisis arose from the reversal of Bank of England credit policy. Having followed to that date a policy of sterilization of the credit effect of the deficit by lowering its reserve-deposit ratio, the Bank reversed its policy in April by raising the discount rate and only sparingly accommodating the discount market. The suddenness and severity of the action led to a panic as the best houses in the trade found it impossible to obtain domestic credit. The panic in October by contrast was due to an internal drain that resulted from a loss of confidence in the convertibility of bank deposits into Bank of England notes and was essentially due to the operation of Peel's Act, which is described in the next section. That crisis was overcome by the joint effect of a suspension of the prohibition of fiduciary issue and of a discount rate at an unprecedented level of 8 percent. The

RESERVES

BULLION

10

-------~---------- 17.5

7.5

15.0

,

\

\

5.0

12.5 \ \

,

,

\

\

2.5

\/

J

,

10

\,,,

Bullion ''\

~

o '--40 1845

Fig. 5.1

4 1846

20 1846

36 1846

52 1846

16 1847

32 1847

"

I

48 1847

-----1

7 .5

12 1848

Bank of England holdings of bullion and note reserves (in million f).

237

The Bank of England in the Crisis of 1847 21.4

~--------------------,

20.2

19.0

17 .2

L....-

I 1847

Fig. 5.2

--I

9

25

33

1847

1847

1847

.&--_....I

5

41 1847

1848

Stock of currency, 1847 weekly data (in million £).

95.0 ~------------------_ .~

en

U

en en

90.0

85.0

80.01...I

Fig. 5.3

.... II

21

31

41

51

61

Price of 3 percent consols, 1847 weekly data.

policy led domestically to a full return of confidence and a reduction in defensive liquidity positions while at the same time attracting gold and capital from the rest of the world. By the end of the year the Bank of England had restored its note reserve as well as its stock of bullion to the levels prevailing prior to the year of crisis. Both in April and in October the timing of the public-sector-deposit withdrawals aggravated significantly Bank of England liquidity problems and contributed to precipitating the crises. The policies pursued to alleviate the October panic were an early application of by now well-accepted central-banking principles. In the subsequent section we develop an analytical framework to study the interaction of external payments and the financial system.

238

5.3

Rudiger Dornbusch and Jacob A. Frenkel

Institutions and a Model of Financial Markets

In what follows we discuss the events of 1847 in the context of a simple model. The model establishes how the rate of interest is determined by the existing stock of bullion, currency preferences of the public, and reserve behavior of the Bank of England. The analysis starts with the institutional setting, namely the provisions of Peel's Act of 1844. 5.3.1

The Provisions of Peel's Act and the Money Supply

Peel's Act, passed by Parliament in 1844, essentially enacted the doctrine of the "currency school."2 The main provisions were the following: 1. The Bank was separated into Banking and Issue departments. 2. The fiduciary note issue was limited to fourteen million pounds sterling, and the supplementary note issue required a lOa-percent marginal bullion reserve. 3. Notes were issued for bullion at 13.17s. 9d pound sterling per ounce of gold. The consolidated balance sheet of the Issue and Banking departments is shown in table 5.2. Several points are worth noting. First, public deposits, including in particular the Exchequer and the account for debt service (entitled "For Payment of Dividends"), are distinguished from private deposits that include bankers' accounts. Second, on the liability side, the item "circulation" refers to Bank of England notes including seven-day and other bills. The latter remain less than one million pounds throughout. On the asset side, the item "bullion" refers to silver and gold bullion as well as coin in the Bank. To study the Bank of England as it operated under Peel's Act, we show in table 5.3 the corresponding disaggregated balance sheets of the Banking and Issue departments, respectively. In analyzing the balance sheet two comments are relevant: First, the Issue Department has security holdings in the amount of £14 million that back the fiduciary component, (F), of note issue, (N), but at the margin there is 100 percent backing of note issue. Furthermore, note issue is confined to the Issue Department. Second, the Banking Department holds part of the Bank of England note issue as reserve, (R), against its deposit liabilities, (D + G). The money supply-by which is meant here the supply of Bank of England liabilities-equals the sum of currency and private deposits, and the monetary base equals the sum of currency and reserves. With these definitions we express the proximate determinants of the money supply in equation (1): (1)

l+c M == - - [B + F] c + ro.

=

m (c, r) [B + F] ,

239

The Bank of England in the Crisis of 1847

240

Rudiger Dornbusch and Jacob A. Frenkel

Table 5.3

Bank of England Disaggregated Balance Sheet Banking Department

Issue Department

Assets

Liabilities

Assets 3

Liabilities

Note reserves (R) Loans (L)

Private deposits (D) Public deposits (G)

Gold (B) Securities (F)

Notes (N)

3

As usual, we suppress the equity component in the balance sheet.

=

where m(c, r) (1 + c)/(c + rex) is the money multiplier. The ratio of total to private deposits is denoted by ex (D + G)/D, c denotes the currency-private-deposit ratio of the nonbank public, and r is the actual reserve-total-deposit ratio of the Banking Department.3 From equation (1) it can be seen that an increase in bullion B, given the reserve- and currency-deposit ratios, will increase the money stock as will a reduction of the ratio of total to private deposits and of the reserve- and currencydeposit ratios. Throughout this paper the discussion is confined to the supply of Bank of England note and deposit liabilities. 5.3.2

=

The Financial Model

The currency-deposit ratio is determined by institutional factors as well as by the reserve-deposit ratio. Specifically, a rise in the actual reservedeposit ratio, r, is assumed to enhance confidence in the convertibility of deposits into notes (internal convertibility), and therefore it reduces the desired currency-deposit ratio, c. This relation between c and r is expressed by equation (2): (2)

c = c(r), c'

~

o.

With this assumption the money-supply function (1) becomes:

(3) where mer)

M=m(r)[B+F];m'~O.

= m(c(r), r).

Our specification implies that a rise in the reserve-deposit ratio exerts two effects on the money multiplier. First, the ratio reduces the multiplier directly through the increased use of high-powered money by the Banking Department; and second, the rise in the reserve-deposit ratio raises confidence and thereby reduces the currency-deposit ratio, which in turn increases the money multiplier. In what follows we assume that the net effect of a higher reserve-deposit ratio is to lower the money multiplier, that is, iii' < O. This may appear plausible at first sight but is in fact a strong assumption since it eliminates the possibility of a dominating impact of the internal convertibility problem. The demand for real balances is assumed to depend on real income, y,

241

The Bank of England in the Crisis of 1847

as well as on the rate of interest, i, in the conventional way. Monetary equilibrium requires that the real money stock, MIP, equals the demand for real balances, L( ) as in equation (4):

(4)

m(r)[B+F]IP=L(i,y);

LiO.

Focusing on the short run of weeks or months rather than a year or more, we take both prices and output as exogenous to the financial sector.4 With this assumption and with a given fixed stock of fiduciary issue, equation (4) can be solved for the equilibrium interest rate as a function of the reserve-deposit ratio and the stock of bullion: (4')

i = i(r, B; ... );

The adjustment of the Bank's lending policy, motivated by prudence and profit, is described in equation (5). The Banking Department adjusts gradually, raising the reserve-deposit ratio through credit contraction, in proportion to the discrepancy between the desired reserve-deposit ratio ( ) and the actual ratio, r. Thus,

(5)

;. = v [(i) - r], ' < O.

In equation (5) the desired reserve-deposit ratio, (i) , declines as the rate of interest increases. This decline reflects the behavior of the Banking Department: in response to more profitable loan opportunities the desired liquidity of the balance sheet is reduced. A specification of the rate of inflow of bullion completes the mode}. The rate of inflow of bullion or the balance of payments, denoted by B, depends on the exogenous trade balance as well as on the rate of capital inflow. Capital flows respond positively to the international interest differential, i - i*, and the foreign interest rate, i*, is taken as given: (6)

B=B(i-i*; ... );Bi>O.

Again, we concentrate on the short term and therefore leave relative prices and output as exogenous to the model. 5.3.3

Formal Dynamics

Equations (4'), (5), and (6) represent a dynamic model of the interaction between the Banking Department's credit policy and the balance of payments. Substituting equation (4') in equations (5) and (6) yields the following pair of ·equations: (7) (8)

;= G(r, B); GrO. B = H(r, B); H r > 0, H B < O.

where the signs of the partial derivatives follow from the previous assumptions. It is readily verified that the system shown in figure 5.4 must be stable.

242

Rudiger Dornbusch and Jacob A. Frenkel

r

In figure 5.4 the = 0 schedule shows the locus of reserve-deposit ratios and levels of bullion at which the Banking Department is in equilibrium with respect to its liquidity position. Therefore, along that schedule the reserve-deposit ratio is neither rising nor falling. At points above the schedule, the high reserve-deposit ratio implies a low realmoney supply and thus high interest rates. The preferred reserve-deposit ratio is low, and therefore above the r = 0 schedule the reserve-deposit ratio is being lowered. Conversely, below the = 0 schedule, the Banking Department seeks to become more liquid because interest rates are low, and ther~fore the reserve-deposit ratio is raised. Along the B = 0 schedule the balance of payments is in equilibrium. Points below and to the right of the schedule correspond to high money supplies, low interest rates, capital outflows, and therefore deficits and falling bullion. By contrast, points to the left of the sch~dule involve high interest rates and growing levels of bullion. Along the B = 0 schedule the interest rate is compatible with external balance. The interest rate is higher above and to the left of the schedule and lower below and to the right of the schedule. The relative slopes of the two schedules are implied by the previously assumed restrictions. As the arrows indicate, the dynamic model of the financial sector must be stable and the approach to equilibrium cannot be oscillatory. From any initial reserve-deposit ratio and stock of bullion, the adjustment process leads to the steady state at point A where the Bank's liquidity position is in equilibrium and external payments are balanced. The response of the Bank's reserve-deposit ratio to the rate of interest is reflected in the slope of the r = 0 schedule. The less responsive the

r

Reserve - Deposi t Ratio

8=0

r= 0 r

o

~----_-----&_----- Bullion

o Fig. 5.4

Financial model.

B

243

The Bank of England in the Crisis of 1847

Bank, the flatter the schedule; in the extreme, when the Bank is entirely unresponsive, the desired reserve-deposit ratio is constant and the = 0 schedule is horizontal. As the reserve-deposit ratio becomes more responsive, th~ schedule steepens until, in the limit, its slope coincides with that of the B = 0 schedule. The responsiveness of the reserve-deposit ratio, of course, determines the extent to which interest rates move in the adjustment process. If the reserve-deposit ratio declines in response to high interest rates, then a shortage of bullion will in part be offset by increased lending on the part of the Bank and interest rates therefore will tend to be lower, the balance of payments will be smaller, and the rate of adjustment will be slower. Conversely, if the reserve-deposit ratio is unresponsive, a shortage of bullion implies a sharper reduction in the money stock, high equilibrium interest rates, larger capital flows, and faster adjustment.

r

5.3.4 The Adjustment Process The traditional model of the price-specie-flow mechanism, originating with David Hume, emphasizes the impact of relative prices on the trade balance and hence on the balance of payments and the international flow of bullion. A deterioration of the external balance due to increased aggregate spending or an adverse development of net exports will lead to bullion export, monetary deflation, declining spending, and price deflation. Both the decline in spending and deflation work to restore external balance. The model we have sketched here, on the contrary, places emphasis on capital flows and banking policy as the main factors in the adjustment process. The two views of the adjustment mechanism are of course complementary, although they may well correspond to different adjustment periods. In the short run, banking policy and capital flows are likely to be the main factors determining bullion flows, since, in the short run, prices and trade flows do not adjust to the full, possible extent. The role of capital flows in the adjustment process was recognized by contemporaries. John Stuart Mill, in particular, noted: It is a fact now beginning to be recognised, that the passage of the precious metals from country to country is determined much more than was formerly supposed, by the state of the loan market in different countries, and much less by the state of prices. (Mill I8?1 , bk. 3, chap. 8, section 4) In addition to the difference between the balance of trade and the capital account in facilitating adjustment, there is another aspect of the adjustment process that deserves emphasis. The traditional representation of the gold standard takes it to be an automatic, nondiscretionary

244

Rudiger Dornbusch and Jacob A. Frenkel

adjustment. Bullion flows are matched one-for-one by changes in the amount of currency outstanding. This is, of course, not the case once the reactions of the Banking Department are taken into account. Changes in the reserve-deposit ratio of the Banking Department affect the money stock independently of the existing stock of bullion. The question then arises whether during the 1847 episode the Banking Department's credit policy might in fact have amounted to partial or even complete sterilization of bullion flows. The possibility of credit expansion by the Bank and of loss of note reserves to finance the export of bullion is suggested by the data which reveal a high correlation between weekly changes in bullion and in note reserves. Consider an autonomous, transitory improvement in the trade balance which leads to an inflow of bullion and therefore to a monetary expansion. The monetary expansion lowers the interest rate, and, with a constant reserve-deposit ratio (or a flat; = 0 schedule in figure 5.4), the lower interest rate leads to capital outflows and thereby to restoration of the initial equilibrium. Now if, on the contrary, the reserve-deposit ratio rises due to the Banking Department response to the reduced profitability of loans, then the rise in the reserve-deposit ratio dampens the decline in interest rates and therefore slo.ws down the adjustment process. The Banking Department's reaction to the interest rate will only slow the speed of adjustment but will not eliminate the adjustment process. Thus our model is also capable of incorporating a partial sterilization policy with an effect of dampening interest-rate movements and reducing the speed of adjustment. Changes in the reserve-deposit ratio enter consideration in another respect. If the Bank, perhaps in response to a loss of confidence on the part of the public, decides to raise the reserve-deposit ratio, then this raise, of course, leads to a reduction in the supply of money and credit. Interest rates rise and that state persists until bullion inflows accommodate the desired increase in reserves. The model suggests therefore that changes in the Bank's reserve preferences may be an important source of macroeconomic disturbance. The possibility of internal inconvertibility turns out to be an important issue in the 1847 crisis. Internal inconvertibility would arise if the Banking Department should become sufficiently illiquid not to be able to redeem its deposit liabilities in notes. Thus there is a clear distinction between external or gold convertibility and internal or note convertibility. Note convertibility involves the Banking Department's reservedeposit ratio. If the reserve-deposit ratio falls too low, the public loses confidence and reacts by raising the currency-deposit ratio. While our model embodies this reaction of the public, the reaction is for the moment not allowed to exercise a dominating influence.5

245

The Bank of England in the Crisis of 1847

5.3.5

Some Evidence

Before discussing in detail the various crises that occurred during 1847, we look at some evidence that is consistent with the general analytical framework outlined in this section. The dynamic model was summarized by equations (7) and (8). Changes in the reserve-deposit ratio depend negatively on the level of that ratio and positively on the stock of bullion, while changes in the stock of bullion depend positively on the reserve-deposit ratio and negatively on the stock of bullion. In table 5.4 we report regressions of the changes in the reserve-deposit ratio and bullion on the previous-week levels of these variables. The coefficients have the predicted sign and are statistically significant. We view these estimates as providing support for the analytical framework that was developed in this section, and we turn next to a more detailed analysis of the crises of 1847. 5.4 Financial Markets and the Balance of Payments in 1847

5.4.1

The April Crisis

The harvest failure of 1846-47 depleted the bullion in the Bank in the fall of 1846 and more so in early 1847. Table 5.5 shows the development of bullion, note reserves, the reserve-deposit ratio, the discount rate, and the stock of notes in the hands of the public during the first half of 1847. The table brings out forcefully the magnitude of this depletion. Indeed, over the period 2 January to 17 April 1847, bullion fell by about 40 percent and note reserves in the Banking Department declined by about 70 percent. The extraordinary decline in the reserve-deposit ratio from 46 percent to 19.6 percent implies that the Bank sterilized substantially the effect of gold outflows. The decline in the reserve-deposit ratio occurred The Dynamic Model, 1847 Weekly Data (standard errors in parentheses)

Table 5.4 Dependent Variable Constant -0.119 (0.099)

~'t

0.337(107 ) (0.099)107

~Bt

't

't-1

-0.625 (0.151) 0.262(107 ) (0.118)107

Bt -

1

0.301(10- 7 ) (0.122)10- 7 -0.413 (0.116)

R2

D.W.

P

.36

1.85

.75

.66

2.00

.85

Notes: and B t denote, respectively, reserve-deposit ratio and bullion in the Bank of England; ~'t and ~Bt denote the weekly change in these variables. R 2 denotes the coefficient of determination and p the first-order autocorrelation coefficient.

246

Rudiger Dornbusch and Jacob A. Frenkel

along with an increasing discount rate. The table reports the weighted average discount rate applied by the Bank. From a level of 3 percent at the beginning of the year, the rate was gradually raised toward 5 percent in early April 1847. These developments suggest that part of the effects of the external drain on the money supply were sterilized. Whether sterilization was a conscious policy, or whether it was a banking response to increasing interest rates and credit tightness, is open to question. But it is certainly interesting to note that F. T. Baring, the ex-chancellor of the exchequer, argued that the possibility of sterilization was a major defect of the Bank Act of 1844: I believe, if we look back, we shall find that the operation of the deposits and the question of reserve was not sufficiently considered, either by those who were favourable or those who were opposed to the bill. I cannot find in the evidence before the committee of 1840 more than a few sentences leading me to suppose that danger arising from such a cause was contemplated or referred to; yet this was a most important consideration; for it was by the reserve, the bank was enabled to do what was contrary to the spirit of the bill when gold was running out, not to reduce their circulation by a single pound. I do not think that the system works satisfactorily in this respect; and in fact, the point did not receive anything like a sufficient consideration. Perhaps it was impossible before the bill was in practical operation to see how the reserve of notes would operate; but it certainly never entered into the contemplation of anyone then considering the subject that £7,000,000 in gold should run off, yet that the notes in the hands of the public would rather increase than diminish. (MacLeod 1896, pp. 141-42) The relative constancy of notes in the hands of the public, to which Baring refers, is shown in table 5.5. Through late April 1847, notes were practically unchanging while bullion declined by nearly one-third. In this period the Bank of England expansion "financed" the export of bullion.

Table 5.5

2 January 6 March 3 April 17 April 1 May 5 June

The April 1847 Crisis (million £)

Note Reserves

Bullion in Issue Department

ReserveDeposit Ratio (%)

Discount Rate (%)

Notes in Public Hands

8.23 5.71 3.70 2.56 2.74 5.09

14.26 10.99 9.55 8.80 8.51 9.43

46.0 36.0 23.9 19.6 23.6 32.0

3.10 4.14 4.25 5.25 5.25 5.20

20.0 19.3 19.9 20.2 19.8 18.3

Sources: See table 5.2 and Appendix.

247

The Bank of England in the Crisis of 1847

Bullion losses did not exert their full contractionary effect on money and credit because the reserve-deposit ratio was declining.6 In March and April things became troublesome. The ongoing decline of bullion tightened credit-market conditions, and the failure of the Bank to change its accommodating stance in the face of a deteriorating balance sheet evoked concern about a sudden reversal of policies that would leave the public without notes and without loans. In April, therefore, the ongoing drain of bullion was reinforced dramatically by the seasonal payment of the dividend, which meant a significant run-down of public deposits. During the week of 17 April reserves fell to a level of only £2.56 million; the reserve-deposit ratio fell to less than 20 percent. Consol prices fell in March-April by 4.5 percent and short-term interest rates skyrocketed as the Bank moved vigorously to restore its liquidity position by reduced discounts, consol sales, and high discount rates. Figure 5.5 shows the series for private and public deposits. The figure makes it clear that whatever influence the bullion drain had on the liquidity position of the Bank, the sharp public-deposit withdrawal could not but accentuate the problem. Table 5.6 shows that the public-sectordeposit withdrawal led only partially to a loss of note reserves and that the money stock (currency held by the public plus private deposits at the Bank of England) did not change substantially. The table confirms that the Bank of England managed to face the runoff by selling securities. Figure 5.6 shows the weekly series of the reserve-deposit ratio during the year. The effects of the extraordinary loss of reserves (by 17 April) and the reaction of the financial markets and the Bank of England have been described by MacLeod (1896, p. 142):

12.0,..........-------------------.....,

9.5

, ~~

,

, 4.5

,J

"II ~

\

u , = ,

\ \ Q. ,

, L--

I 1847

Fig. 5.5

9 1847

I I I ,

I \.~,'

\ <%

2.0

,

I I

7.0 ,

\

I

I

, I

\.-1

Public Deposits

I

~'C;;

J

"

17 1847

~

Q> .Qcn u'~

au 25 1847

33 1847

41 1847

---I

49 1847

5 1848

Public and private deposits at the Bank of England (in million f).

248

Rudiger Dornbusch and Jacob A. Frenkel

Table 5.6

Money, Public Deposits, and Note Reserves in the April 1847 Crisis (million £)

3 April 10 17 24

Money Stock a

Public Deposits

Note Reserves

Bullion b

30.3 32.7 32.2 29.8

6.0 5.0 3.0 2.6

3.7 2.8 2.6 2.7

10.2 9.8 9.3 9.2

Sources: See table 5.2 and Appendix. aNotes in the hands of the public plus private deposits at the Bank of England. bTotal bullion in the Issue and Banking departments including coin.

0.475

CIl CIl

0.375

U

0.275

0.175

0.075

Fig. 5.6

1

9

1847

1847

17 1847

25 1847

33 1847

41 1847

49

5

1847

1848

Reserve-total-deposit ratio, 1847 weekly data.

When, therefore, the public saw that the whole banking resources of the bank were reduced to £2,558,000, a complete panic seized both the public and the directors. The latter adopted severe measures to check the demand for notes. The rate was not only raised to five percent, but this was only applicable to bills having only a few days to run, and a limit was placed upon the amount of bills discounted, however good they might be. Merchants who had received loans were called upon to repay them without being permitted to renew them. During some days it was impossible to get bills discounted at all. These measures were effectual in stopping the efflux of bullion; and a sum of £100,000 in sovereigns, which had been actually shipped for America, was relanded. During this period the rate of discount for the best bills rose to nine, ten and twelve percent. The tightening of credit led to internal dishoarding and to some reflow of bullion. Accordingly, the Bank's reserve-deposit ratio quickly rose to about 30 percent. The crisis was overcome and the adjustment process of

249

The Bank of England in the Crisis of 1847

tight money was underway. The external part of that adjustment process involved importation of bullion and foreign investment in London as well as English borrowing abroad. In response to high yields on consols, the emperor of Russia decided to substitute foreign securities for gold as backing for the Russian currency and made substantial purchases of consols. These developments and their effect on the restoration of confidence were described by a contemporary as follows: Between the 25th and 28th of April confidence in a slight degree revived. The Bank was then discounting more freely; and the important news was announced that the Emperor of Russia had issued a Ukase "ordering an investment of about four and a half million sterling in home and foreign securities." Under the impression that a large amount of the money would find employment in Consols, as ultimately was the case, this circumstance, coupled with greater disposition of the Bank to grant facilities for accommodation, tended to abate the pressure. (Evans 1849, p. 62) It seems fairly clear from the events that the Bank's policy in the first half of the year was certainly poor. It had all the characteristics of a policy of "too late, and (therefore) too vigorously." The continued expansion of credit in the face of falling reserves and the prospect of further decline in bullion meant that a crunch was quite inevitable. To wait too long before tightening, and ultimately to administer it with severity, led to an unnecessary panic in the money market. 5.4.2 The October Crisis The severe credit tightening following the April crisis restored the Bank's ability to maintain convertibility. Table 5.7 shows that by June the reserve-deposit ratio was again substantial. Currency outstanding was significantly lower than during the April crisis. But during the late summer, and especially in early fall, conditions deteriorated, setting the basis for the October crisis.

Table 5.7

5 4 23 24

June September October December

The October 1847 Crisis (million £)

Note Reserves

Bullion in Issue Department

ReserveDeposit Ratio (%)

Discount Rate (%)

Notes in Public Hands

5.09 4.19 1.55 7.79

9.43 8.40 7.87 11.61

32.0 28.9 11.6 44.5

5.2 5.9 8.1 5.7

18.3 18.2 20.3 17.8

Sources: See table 5.2 and Appendix.

250

Rudiger Dornbusch and Jacob A. Frenkel

The credit tightening starting with the April crisis raised the cost of credit substantially above what it had been in past years, and indeed above anything the public could remember. This raise is the burden of "The Petition of Merchants, Bankers, and Traders of London against the Bank Charter Act" issued in July 1847, in which the opening statement reads: "That there has lately been apparent throughout the commercial and manufacturing community of this country an extent of monetary pressure, such as is without precedent in the memory of the oldest living merchant" (Gregory 1929, 2: p. 3). Figure 5.7 confirms that 1847 interest rates were at a peak relative to the preceding twenty-five years. In May, first-class bills had been discounted at an all-time high of 7 percent, but by July the rate had in fact returned to 5.5 percent which, as the figure shows, was still very high. By October the increasing tightness of credit, the extraordinary height of the rate of interest, commercial failures, and the threat of default by financial institutions had increased even further. Sir Charles Wood, the chancellor of the exchequer, is quoted as stating: When he came to town in October he found the City in a state of panic. He saw persons of all classes and descriptions from the time he was up until he went to bed, and he never passed so painful a week. The interest of money rose to an exorbitant rate, and 60 percent per annum was charged for what were called "continuations" for one day. It was thought impossible that the loans could be repaid. (Gregory 1929, 2: p. 11) Developments between June and September resulted in some decline in both bullion and note reserves. The large payments for grain shipments received in June and July from Russia became due and caused a further drain of bullion and reserves. By now, of course, the cumulative external drain had reduced both bullion and note reserves. Bullion in the Issue 1 0 . 0 _ - - - - - - - - - - - - - -__-~

7.5

5.0

2.5

0.0 L - -

Fig. 5.7

-----'

I

1

I

I

I

I

I

I

1

I

1824

1827

1830

1833

1836

1839

1842

1845

1848

1851

Rate of discount on first-class bills, 1824-48 (percent per year). Source: United Kingdom, Parliament 1848a, app. C, p. 467.

251

The Bank of England in the Crisis of 1847

and Banking departments still remained at a comfortable £8.3 million, but note reserves were down to only £1.55 million, well within the possibility of depletion by a scramble for currency. The same point is evident from the reserve-deposit ratio that had declined from 32 percent in June to only 11.6 percent by late October. In the October crisis, once again, public-sector-deposit movements played a role, but this time in combination with a scramble for currency by the public. Table 5.8 shows a large decline in public-sector deposits of nearly £4 million between 9 and 16 October. Three-quarters of that deposit reduction is matched by a reduction in Exchequer bills on the asset side. In the week of 23 October, dividend payments further reduced public-sector deposits and showed up in increased currency in circulation. The increase in currency during the first three weeks was only £2 million and as such did not appear large, but that gain must be compared to note reserves in the Issue Department which by the 23rd had fallen to only £1.5 million. The crisis thus involved potential minor increases in currency holdings reducing the Bank to insolvency. While the external drain set the preconditions for the crisis, it was the internal run and scramble for currency shown in figure 5.8 that caused the panic of October. The poor liquidity position of the Bank deteriorated further as a result of commercial failures. These caused the public to question the soundness of private banks and thus required, on the banks' part, increased liquidity to demonstrate convertibility of their liabilities into Bank of England notes or specie. Again this factor exacerbated the shortage of Bank of England notes and the inadequacy of note reserves.? It is not certain who broke the Old Lady's back. There is some indication in the inquiry that private bankers threatened the withdrawal of deposits from the Bank in excess of the amount of notes on hand (Gregory 1929, 2: p. 113).8 The Bank, however, put up an admirable stone face and would claim in the inquiry that convertibility was never in question. Thus, in 1848, the governor of the Bank, James Morris, told the House of Lords: Table 5.8

9 October 16 23 30

The Bank in October 1847 (million £) Public Deposits

Private Deposits

Circulation a

Exchequerb BulBills lion c

9.4 5.5 4.8 4.7

7.7 8.7 8.6 8.9

19.5 20.3 21.3 21.8

3.9 0.8 0.7 1.2

Sources: See table 5.2 and Appendix. aIncluding seven-day bills. bHeld by the Banking Department. CTotal bullion, including coin in the Issue and Banking departments.

8.4 8.4 8.3 8.4

252

Rudiger Dornbusch and Jacob A. Frenkel 12.0

...-----------------~ 22.0

10.0

20.5

8.0

19.0

6.0 "-------------------~ 17.5 1 9 17 25 33 41 49 5 1847 1847 1847 1847 1847 1847 1847 1848

Fig. 5.8

Private deposits and currency.

The Question was put to me over and over again whether we were able to take care of the Bank. I always stated that, so far as the Bank itself was concerned, we had no Difficulty; but that, whether Her Majesty's Government might have any political Reasons, such as Fear of Mills being stopped, or Riots in the Country, was a Question for them to decide, and one which we could not answer. (Gregory 1929, 2: p. 11) Indeed, the way the Bank proposed to restore its financial position suggests that it might well have been able to maintain convertibility by an extraordinary contraction of credit. Here it is worth quoting what the governor told the House of Lords: We should have had no Difficulty whatever in meeting all our Liabilities. We should not have been able to give the same Extent of Accommodation that Parties were requiring from us. Parties came and thought they had only to ask for Money and they would have it at on,ce. We might have put into the Account a considerable Amount by selling Consols. We had going off weekly Bills to the Extent of £1,500,000, so that by discounting even at the Rate of £100,000 a Day to give the Public some Accommodation our Reserve would still have increased at the Rate of £900,000 a Week. It is certain that in a very short Period we should have had as large a Reserve as would be necessary for our Purposes, and therefore I maintain that the Bank was never at any Period in Jeopardy. (Evans 1849, p. 89) Whether the Bank might have been successful or not, the contraction of money and credit was so severe, commercial failure was so widespread and reaching increasingly the banking system, the government felt it was wise to suspend Peel's Act, authorizing the Bank to issue notes without gold backing. The letter of instruction, dated 25 October 1847, is given below (the letter is reproduced in Turner 1897, pp. 159-60 and in Evans 1849, p. 87).

253

The Bank of England in the Crisis of 1847

Downing Street, Oct. 25th, 1847. GENTLEMEN,-Her Majesty's Government have seen with the deepest regret the pressure which has existed for some weeks upon the commercial interests of the country, and that this pressure has been aggravated by a want of that confidence which is necessary for carrying on the ordinary dealings of trade. They have been in hopes that the check given to dealings of a speculative character, the transfer of capital from other countries, the influx of bullion, and a feeling which the knowledge of these circumstances might have been expected to produce, would have removed the prevailing distrust. They were encouraged in this expectation by the speedy cessation of a similar state of feeling in the month of April last. These hopes have, however, been disappointed, and Her Majesty's Government have come to the conclusion that the time has arrived when they ought to attempt, by some extraordinary and temporary measure, to restore confidence to the mercantile and manufacturing community. For this purpose, they recommend to the directors of the Bank of England in the present emergency to enlarge the amount of their discounts and advances upon approved security; but that in order to retain this operation within reasonable limits a high rate of interest should be charged. In present circumstances they would suggest that the rate of interest should not. be less than 8 per cent. If this course should lead to any infringement of the existing law, Her Majesty's Government will be prepared to propose to Parliament on its meeting a Bill of Indemnity. They· will rely upon the discretion of the directors to reduce as soon as possible the amount of their notes if any extraordinary issue should take place within the limits prescribed by law. Her Majesty's Government are of opinion that any extra profit derived from this measure should be carried to the account of the public, but the precise mode of doing so must be left to future arrangement. Her Majesty's Government are not insensible of the evil of any departure from the law which has placed the currency of this country upon a sound basis; but they feel confident that, in the present circumstances, the measure which they have proposed may be safely adopted, and at the same time the main provisions of that law, and the vital principle of preserving the convertibility of the bank-note may be firmly maintained. We have the honour to be, Gentlemen, Your obedient, humble Servants, J. RUSSELL. (Signed) CHARLES WOOD. The Governor and Deputy Governor of the Bank of England.

254

Rudiger Dornbusch and Jacob A. Frenkel

The authorization for fiduciary issue, coupled with the high discount that the Bank was charging, rapidly restored financial stability. The removal of the restriction of fiat-money issue dissipated the concern for the internal convertibility of deposits into notes. High interest rates, at the same time, brought about very substantial inflows of capital. The capital inflows, in turn, expanded the bullion in the Bank and thus led to internal monetary expansion that over time alleviated the extreme tightness in credit markets. By 24 December 1847, the reserve-deposit ratio had risen substantially while interest rates had declined from their panic peaks of late October. It is important to recognize the role of international capital flows in the adjustment process. A key aspect, in the eyes of the government, was that suspension of Peel's Act be implemented in a manner in no way prejudicial to external convertibility, as it would be if fiat-money issue financed export of bullion. To prevent such a course of events, a high interest rate was an essential part of the suspension of Peel's Act since the very size of the international interest differential would ensure that gold imports were advantageous. They would in turn provide an external basis for domestic monetary expansion. The 1847 episode was probably responsible for the popular maxim that "7 percent will draw gold from the moon."9 That high interest rates do attract gold flows is immediately obvious from the bullion gain. Between October and December the Bank's holdings of bullion rose by more than 50 percent. The interest responsiveness of bullion flows is substantiated by a regression of the flow of bullion on lagged interest rates. To examine the interest responsiveness of bullion flows we regressed the change in the stock of bullion on current and lagged values of the consol yield (up to a lag of eight weeks). We experimented with various lag structures and, consistently, the only significant coefficients were on the fourth and the seventh lag. Equation (9) reports the regression of gold flows on the fourand seven-week lagged interest rate; standard errors are reported below the coefficients: (9)

Ji.B t = - 0.892(10 7 ) (0.131)10 7

+ 0.179(10 9 )it _ 4 + 0.795(108 )it _ 7 • (0.043)10 9

(0.396)108

R 2 =0.75, D.W. =2.16, p=0.43.

The results are consistent with the expectation that the rate of capital inflow is related positively to the domestic rate of interest. Since in the pretelegraph period information on interest differentials could not be transmitted instantaneously, it is reasonable that gold flows responded to lagged values of interest rates. The length of the lag, in turn, should correspond to the length of time needed for a round trip between the home country and its trading partners. The round trip was necessary since the information on rates of interest had to be transmitted and then the

255

The Bank of England in the Crisis of 1847

gold had to be shipped. The four- and seven-week lags that are reported in equation (9) correspond, respectively, to the length of time of the round trip between London and New York and London and St. Petersburg. tO If this interpretation of the lag structure is correct, one might expect that in subsequent periods, following the introduction of the telegraph, the lag structure would be shortened by a factor of 50 percent. What the regression bears out is also quite clear from contemporary accounts-eapital did move in response to interest rates. The season was advanced and the navigation on the Baltic near its close; but even at the disadvantage of a double rate of insurance, orders had been sent to St. Petersburg, under the impulse of an 8 percent rate of interest, which sufficed to bring back all, and more than all the gold which had been exported in the beginning of the year. (Hubbard 1848, p. 23) That same principle was understood by the Bank of England. It was desirable that capital and bullion should be attracted to this country, and it was only by the attraction of a high rate of interest that this desideratum could be accomplished. He [the chancellor of the exchequer] was convinced, therefore, that the mode in which the Government had acted was the one best calculated to attain the end they had in view-namely, the influx of capital and the importation of bullion, and thereby the removal of the panic. (Evans 1849, p. 98) It is relevant to note in the context of the discussion of capital flows that in the case of a transitory real disturbance, such as a harvest failure, the correct response is indeed financing through the capital account of the balance of payments as opposed to the price-specie-flow mechanism of adjustment which operates through the trade account. The 1847 episode illustrates that principle. 5.5

Was the Suspension of Peel's Act Necessary?

The authorization for uncovered note issue, as we already mentioned, immediately removed the panic, so much so that there was almost no need to actually issue uncovered notes. In fact during the period of suspension the Bank issued only £400,000 in notes in excess of the limits set by the Act of 1844. The rapid restoration of confidence and the normalization of affairs allowed the government to revoke the suspension on 23 November (reprinted from Evans 1849, p. 102): Downing Street, Nov. 23, 1847. GENTLEMEN ,-Her Majesty's Government have watched with the deepest interest the gradual revival of confidence in the commercial classes of the country. They have the satisfaction of believing that the course adopted by

256

Rudiger Dornbusch and Jacob A. Frenkel

the Bank of England on their recommendation has contributed to produce this result, whilst it has led to no infringement of the law. It appears from the accounts which you have transmitted to us, that the reserve of the Bank of England has been for some time steadily increasing, and now amounts to £5,000,000. This increase has in great measure arisen from the return of notes and coin from the country. The bullion exceeds £10,000,000, and the state of the exchanges promise a further influx of the precious metals. The knowledge of these facts by the public is calculated to inspire still further confidence. In these circumstances it appears to her Majesty's Governmentthat the purposes which they had in view in the letter which we addressed to you on the 25th of October has been fully answered, and that it is unnecessary to continue that letter any longer in force. We have the honour to be, Gentlemen, Your obedient humble servants, (Signed) J. RUSSELL CHARLES WOOD,

The Governor and Deputy-Governor of the Bank of England. The rapid normalization led in some quarters (in the Bank and elsewhere) to the belief that there had been no real reason for the panic and no need to suspend Peel's Act. Thus Hubbard (1848, p. 25) commented: How utterly baseless were the apprehensions of the panic-mongers is now proved by the fact that the Bank not only never availed itself of the power of additional issue, but met from its own resources all the demands made upon it, including the extraordinary applications which would naturally be encouraged by the prospect of their being favourably received. Of course, the statement reflects, in an exemplary way, the lack of understanding of an internal convertibility crisis. A convertibility crisis or run occurs only if in fact not everybody can be paid off. Suspension of the act removed any conceivable basis for panic and therefore immediately restored a measure of financial stability. The special characteristics of an internal drain and the remedies that are called for were fully perceived by Bagehot ([1873] 1962): A domestic drain is very different. Such a drain arises from a disturbance of credit within the country, and the difficulty of dealing with it is the greater, because it is often caused, or at least often enhanced, by a foreign drain.... What then ought to be done? In opposition to what be at first sight supposed, the best way for the bank ... to deal with a drain arising from internal discredit, is to lend freely. (P. 23) Since the key issue underlying an internal drain is lack of confidence, it is clear that

257

The Bank of England in the Crisis of 1847

what is wanted and what is necessary to stop a panic is to diffuse the impression, that though money may be dear, still money is to be had. If people could be really convinced that they could have money . . . they would cease to run in such a mad way for money. Either shut the Bank at once . . . or lend freely, boldly, and so that the public may feel you mean to go on lending. (P. 31) Chancellor of the Exchequer Sir Charles Wood told the committee of inquiry that the basis of the panic was indeed a lack of confidence in internal convertibility. He quotes commercial traders and bankers as stating: We do not want notes-what we desire is that you should give us confidence, it is only for you to say that you will stand by us, and nothing in the world else will give us confidence. We do not want notes, but only to know where we can get them.... Charge 10 or 12 percent interest if you like-we do not mean to take notes, but let us know that at some rate of interest we can get them and that will amply suffice. (Evans 1849, p. 96) Sir Robert Peel for his part expressed the view that the Bank Charter Act had failed in one important respect, namely, it failed to secure a gradual and early, as opposed to severe and sudden, adjustment: If the Bank had possessed the resolution to meet the coming danger by a contraction of its issue, by raising the rate of discount . . . if they had been firm and determined in the adopting of those precautions, the necessity of extrinsic interference might have been prevented, it might not have been necessary for the Government to authorise the violation of the Act of 1844. (Andreades 1924, p. 339) Many felt that the act, as opposed to Bank of England policy, had no effect in aggravating the crisis. S. J. Loyd, for example, stated to the committee that the Act had no effect whatever in aggravating the Pressure. It protected the Public from the additional evil, which would otherwise have occurred, of a Failure in maintaining Convertibility of the Notes, and the consequent complete Destruction of our Monetary System. (Gregory 1929, 2: p. 44) But this view was not how the House of Lords came to see it. While also agreeing on the fundamental importance of external convertibility, the committee concluded that the recent panic was materially aggravated by the operation of that statute, and by the proceedings of the Bank itself. This effect may be traced, directly, to the Act of 1844, in the legislative restriction imposed on the means of accommodation, whilst a large amount of bullion was held in the coffers of the Bank, and during a time of

258

Rudiger Dornbusch and Jacob A. Frenkel

favourable exchanges; and it may be traced to the same cause, indirectly, as a consequence of great fluctuations in the rate of discount, and of capital previously advanced at an unusually low rate of interest. This course the Bank would hardly have felt itself justified in taking, had not an impression existed that, by the separation of the issue and the banking departments, one inflexible rule for regulating the Bank issues had been substituted by law in place of the discretion formerly vested in the Bank. (Turner 1897, pp. 162-63) Likewise, the Committee are fully aware that Alternations of Periods of Commercial Excitement and of Discredit, of Speculation and of Collapse, are likely to arise under all Systems of Currency; it would be visionary to imagine that they could be averted altogether, even if the Circulation were exclusively Metallic. But it is on this Account that greater Care should be taken to avoid increasing an Evil, perhaps inevitable, by any arbitrary and artificial Enactments. The Committee are of opinion, that the Principle on which the Act of 1844 should be amended is the Introduction of a discretionary relaxing Power; such Power, in whomsoever vested, to be exercised only during the Existence of a favourable Foreign Exchange. (Gregory 1929, 2: p. 40) The very interesting aspect of the House of Lords' recommendation is the link between fiat issue and the state of the foreign exchanges. It represents a departure from strict gold standard rules where only actual gold flows can be monetized. Under the proposal of the committee it is enough to have the conditions for gold flows to take place, as opposed to actual arrival of gold, for suspension to be allowed. There is money issue, so to speak, on credit. The compromise of lending freely at high ratesthe high rates ensuring in time the arrival of gold and the validation of the fiat issue-is important in that it is a remedy specifically for internal convertibility crises for which Peel's Act had made no allowance. The model of the gold standard developed in section 5.3 above did not make provision for issues of stability and crisis. Portfolio adjustment by the Bank and gold flows in response to interest differentials brought about smooth adjustment. How can the panic of October be accommodated in such a framework? Here it becomes essential to recognize the dominating effect of the currency-deposit ratio. When the reservedeposit ratio of the Bank is sufficiently low so as to reduce confidence in the viability of internal convertibility, then further reduction in the reserve-deposit ratio may bring about changes in the currency-deposit ratio so large as actually to reduce the money multiplier and the money stock. In terms of equation (3), the money multiplier, in that low reserve-deposit-ratio region, responds positively to an increase in the reserve-

259

The Bank of England in the Crisis of 1847

deposit ratio, and therefore the equilibrium interest rate responds negatively. The sign reversal implies the possibility of multiple equilibria as shown in figure 5.9. Point A' is the stable equilibrium studied earlier; point A is another equilibrium in the region where convertibility concerns dominate the money-supply process. Point A is an unstable equilibrium that can be attained only for initial conditions on the GG schedule. Initial conditions below and to the left of GG lead to unstable paths of the reserve ratio that decline to zero. From a point such as C, with low bullion and low reserves, the Bank lowers the reserve-deposit ratio while bullion is rising. The reserve-deposit ratio keeps falling, the currency-deposit ratio keeps rising, and the system must collapse. 5.6

Concluding Remarks

In this paper we examined the operation of the gold standard and the performance of the Bank of England during the crisis of 1847. The key feature of that crisis was its origin: it began with the instability of the real sector rather than from monetary disorder. That crisis highlights the role of confidence in both external and internal convertibility. We presented a simple model that seems to capture the central characteristics of the crisis and that emphasizes the role of international capital flows during the adjustment process. Our analysis suggests that the suspension of Peel's Act, i.e., the collapse of the rigid rules of the gold standard, was the correct and the essential policy required for the restoration of confidence. We return now to the more general question of the gold standard as a frame for macroeconomic stability. The monetary system, to work satisReserve - Deposit Ratio

8=0

' - - - - - - - - - - - - - - - Bullion B

o

Fig. 5.9

Financial model-the unstable case.

260

Rudiger Dornbusch and Jacob A. Frenkel

factorily, would have to satisfy three criteria: (1) assure stability and predictability of the general level of prices and output; (2) separate banking and financial problems, to a large extent, from the macroeconomy; and (3) provide a stable financial framework that facilitates financial intermediation and lending, both domestically and in the world economy. There is considerable doubt now whether the objective of price-level stability was well served by the gold standard. There is evidence of short-term price variability substantially in excess of post-World War II experience. It may also be argued that long-term stability was, to a large extent, accidental-a consequence of fortuitous gold discoveries rather than the systematic operation of the system. But even with these qualifications there can be little question that the gold standard was a system that utterly excluded the extreme monetary instability Europe witnessed, for example, in the 1920s. The system also excluded the accelerating path of inflation that we experienced in the 1970s from the interaction of macroeconomic shocks, inertia, and accommodation. But on a different account the gold standard was disappointing. Until the principle "during crisis, discount freely" was firmly established, the gold standard provided an exceedingly poor framework for financial markets. The presence of a lender of last resort-whether it be the Treasury or the central bank-is essential with fractional reserve banking. The lack of a lender of last resort was emphasized in the 1847 crisis when the Bank of England, in the midst of a banking panic, sold consols and reduced discounts, thus assuring confidence in deposit convertibility (not gold convertibility) at the expense of devastating financial distress.

R

8814 8227 6715 6546 6167 5704 5891 5747 5977 6017 5715 5554 5419 4876 3700 2833 2558 2719 2741 3197 3793 4420 4628 5089 5375 5665 5625

Date

DEC 26 JAN 2 9 16 23 30 FEB 6 13 20 27 MAR 6 13 20 27 APR 3 10 17 24 MAY 1 8 15 22 29 JUN 5 12 19 26

Appendix

18037 17895 15645 15374 15024 14123 13851 14628 14707 15250 15860 16252 16434 16019 15504 16242 13016 11760 11611 11800 13071 14430 15410 15923 16922 17419 17717

D+G

7696 7904 9785 10340 10356 9660 9183 9330 8837 9322 9289 9536 9962 9403 9502 11258 10005 9125 9312 8930 8751 8289 8432 8151 8228 8160 7921

D

15067 14952 14308 13949 13443 12902 12288 12299 12215 12045 11596 11449 11232 11016 10246 9867 9330 9214 9338 9589 9870 9949 10170 10237 10359 10512 10526

B

Pc 93.625 93.875 93.500 92.500 90.375 91.500 91.000 90.500 90.625 90.250 89.625 88.250 88.875 89.250 88.000 85.750 86.375 86.000 87.000 86.625 87.000 87.000 88.375 88.250 88.750 88.500 88.750

i

3.204 3.196 3.209 3.243 3.320 3.279 3.297 3.315 3.310 3.324 3.347 3.399 3.376 3.361 3.409 3.499 3.473 3.488 3.448 3.463 3.448 3.448 3.395 3.399 3.380 3.390 3.380

Weekly Observations, December 1846-December 1847

63.000 64.333 66.833 70.250 73.250 74.917 73.833 71.583 71.583 74.583 74.333 74.166 75.833 77.000 77.083 74.416 74.083 75.833 79.500 81.833 85.166 94.833 102.420 99.833 88.833 91.583 91.833

Pw 20253 20725 21593 21403 21276 21198 20397 20552 20238 20028 19881 19895 19813 20140 20546 21034 20772 20495 20597 20392 20077 19529 19542 19148 18984 18847 18901

C 2.63163 2.62209 2.20675 2.06992 2.05446 2.19441 2.22117 2.20279 2.29014 2.14847 2.14027 2.08630 1.98886 2.14187 2.16228 1.86836 2.07616 2.24603 2.21188 2.28354 2.29425 2.35601 2.31760 2.34916 2.30724 2.30968 2.38619

C/D 0.4887 0.4597 0.4292 0.4258 0.4105 0.4039 0.4253 0.3929 0.4064 0.3946 0.3603 0.3417 0.3297 0.3044 0.2386 0.1744 0.1965 0.2312 0.2361 0.2709 0.2902 0.3063 0.3003 0.3196 0.3176 0.3252 0.3175

R/(D+G)

1.69285 1.77478 1.90170 1.87984 1.91442 2.00846 1.96232 1.98196 1.96267 1.91342 1.94415 1.95036 1.93583 2.06898 2.27602 2.29168 2.44981 2.50084 2.48146 2.41791 2.29815 2.18884 2.14196 2.06186 2.00044 1.95349 1.98007

m

5185 4331 4069 4216 3775 3946 3992 4488 4330 4190 4467 4273 4112 3409 3322 2630 1547 1177 2030 2798 4228 4986 5583 6449 7551 7786

JUL

17707 14550 13200 12830 12820 13457 13872 13762 14299 14514 15147 15934 16932 17291 17129 14171 13347 13607 13796 14304 15086 15968 16241 16667 17370 17479

D+G

7968 9305 8640 8326 8316 7886 7514 6931 7106 6791 6981 7185 7484 7962 7714 8675 8581 8911 8804 8312 7866 8239 8441 8437 8607 8243

D

10397 10086 9919 9770 9331 9253 9287 9240 9140 8960 8915 8880 8782 8565 8409 8431 8313 8439 8730 9259 10017 10533 11033 11426 11991 12237

B

3.380 3.380 3.376 3.395 3.385 3.458 3.483 3.443 3.443 3.433 3.453 3.514 3.483 3.556 3.598 3.709 3.750 3.715 3.609 3.535 3.561 3.540 3.504 3.529 3.535 3.519

i

88.750 88.750 88.875 88.375 88.625 86.750 86.125 87.125 87.125 87.375 86.875 85.375 86.125 84.375 83.375 80.875 80.000 80.750 83.125 84.875 84.250 84.750 85.625 85.000 84.875 85.250

Pc 87.083 82.250 74.000 75.500 77.250 75.416 66.833 66.500 60.333 56.666 51.333 49.500 53.500 56.750 54.166 54.250 55.166 53.500 52.333 53.666 54.250 52.917 52.083 51.917 52.833 52.750

Pw 19212 19755 19850 19554 19556 19307 19295 18752 18810 18770 18448 18607 18670 19156 19087 19801 20766 21262 20700 20461 19789 19547 19450 18977 18440 18451

C 2.41114 2.12305 2.29745 2.34855 2.35161 2.44826 2.56787 2.70553 2.64706 2.76395 2.64260 2.58970 2.49466 2.40593 2.47433 2.28254 2.42000 2.38604 2.35120 2.46162 2.51576 2.37250 2.30423 2.24926 2.14244 2.23838

C/D R/(D+G)

0.2928 0.2977 0.3083 0.3286 0.2945 0.2932 0.2878 0.3261 0.3028 0.2887 0.2949 0.2682 0.2429 0.1972 0.1939 0.1856 0.1159 0.0865 0.1471 0.1956 0.2803 0.3122 0.3438 0.3869 0.4347 0.4454

m

2.06645 2.13112 2.24172 2.22293 2.30519 2.29826 2.33000 2.24915 2.26618 2.32775 2.22126 2.25100 2.25543 2.38484 2.42851 2.51889 2.89761 2.99098 2.72328 2.58983 2.28667 2.10102 1.98880 1.84160 1.67391 1.66536

Sources: Data for R, D + G, D, and B come fron United Kingdom, Parliament 1848a, app. 8, pp. 126-43. All other data are from Hubbard 1848. Notes: R, D+ G, D, and B denote, respectively, the stock of note reserves held by the Bank of England; total deposits including Exchequer; London Bankers and private deposits; and the total amount of bullion. All are measured in thousand pounds sterling. C denotes currency, which is measured as bullion minus reserves + 14,000 (the fiduciary issue); m denotes the money multiplier. Pc denotes the price of3-percent consols; Pw the price of wheat; and i is a hundred times the yield on consols, i.e., (100 x 3)/Pc.

3 10 17 24 31 AUG 7 14 21 28 SEP 4 11 18 25 OCT 2 9 16 23 30 NOV 6 13 20 27 DEC 4 11 18 25

R

Date

Appendix (continued)

263

The Bank of England in the Crisis of 1847

Notes 1. Quoted in Andreades 1924, p. 340. 2. On the banking-and-currency-school controversy, see Mints 1945 and Viner 1937, chap.5. 3. For the derivation see Friedman and Schwartz 1963. Throughout our discussion we abstract from the existence of private banks, their note issue, their deposit liabilities, or their demand for Bank of England notes. 4. The supply shock-a harvest failure----exerted opposing effects on prices and output so that the net effect on nominal income and thereby on interest rates may in fact be disregarded without obvious strain. 5. If a rise in the reserve-deposit ratio lowers the currency-deposit ratio so much as to raise the money multiplier, in contrast with our assumption in equation (3), then a higher reserve-deposit ratio would lower interest rates and thus change the specification in equation (4'). The iJ = 0 schedule might then be negatively sloped at low levels of r, and so may even the ; = 0 curve . We examine this case in section 5.5 below. 6. The defects of the Bank Act were discused forcefully by Thomas Tooke (1844) whose views are analyzed by Laidler (1975). For a further analysis see Morgan 1943. 7. It is interesting to note that the developments during the October crisis were part of the evidence that stimulated Jevons's theory of the "frequent autumnal pressure" (see 1evons 1884, chap. 5). 8. On 23 October, when London bankers held deposits of £1.6 million at the Bank of England, the note reserves were only £1.5 million. Balances of all bankers rose from £1.5 to £2.1 million in the week of 16 October, but they declined to only £1.8 million by 23 October. 9. While we could not trace the exact origin of this maxim, it was referred to by Lionel Robbins in a memorandum submitted to the (Radcliffe) Committee on the Working of the Monetary System (United Kingdom, Parliament 1960, p. 218). Robbins ascribes this maxim to "a practical banker." We are indebted to David Laidler for this reference. It is interesting to note that a similar statement can be found in Leaf 1927, p. 34. There it is described as an "old saying in the City." Leaf then goes on to say "but it takes time to bring it even from Paris." We are indebted to Geoffrey Wood for this reference and to Anna 1. Schwartz for help in the search for the source of the popular maxim. 10. For some evidence, see Howarth 1974, p. 313.

References Acres, W. Marston. 1931. The Bank of England from within. Vol. 2. London: Oxford University Press. Andreades, C. B. E. 1924. History of the Bank of England, 1640-1903. 2d ed. London: P. S. King & Sons. Bagehot, Walter. [1873] 1962. Lombard Street. Reprint. London: Richard D. Irwin. Dornbusch, Rudiger, and Jacob A. Frenkel. 1971. Aspects of adjustment policies and mechanism during the 1847 crisis in Britain. University of Chicago Workshop in Economic History, report 7172-3. Evans, D. Morier. 1849. The commercial crisis 1847-1848 beingfacts and figures illustrative of the events of that important period, considered in

264

Rudiger Dornbusch and Jacob A. Frenkel

relation to the three epochs of the railway mania, the food and money panic and the French revolution. 2d ed. London: Letts, Son and Steer. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Gregory, T. E., ed. 1929. Select statutes, documents, reports relating to British banking, 1832-1928. London: Oxford University Press. Howarth, David. 1974. Sovereign of the seas. New York: Atheneum. Hubbard, John G. 1848. A letter to Sir Charles Wood on the monetary pressure and commercial distress of 1847. London: Longman. Hyndman, H. M. [1932] 1967. Commercial crises of the nineteenth century. Reprint. New York: Augustus M. Kelley. (First published 1892.) Jevons, W. Stanley. 1884. Investigations in currency and finance. London: Macmillan. Keynes, John Maynard. 1932. Essays in persuasion. New York: Harcourt, Brace & Company. Laidler, David E. W. 1975. Thomas Tooke on monetary reform. In Essays on money and inflation, chap. 11. Chicago: University of Chicago Press. Leaf, Walter. 1927. Banking. New York: Henry Holt and Company. MacLeod, Henry Dunning. 1896. A history of banking in Great Britain. Vol. 2 of A history of banking in all the leading nations. New York: Journal of Commerce and Commercial Bulletin. Mill, John Stuart. 1871. Principles ofpolitical economy. 7th ed. London: Parker & Company. Mints, Lloyd W. 1945. A history of banking theory. Chicago: University of Chicago Press. Mitchell, B. R. 1962. Abstract of British historical statistics. Cambridge: Cambridge University Press. Morgan, E. V. 1943. The theory and practice of central banking, 17971913. Cambridge: Cambridge University Press. Tooke, Thomas. 1844. An inquiry into the currency principle. 2d ed. London. Turner, B. R. 1897. Chronicles of the Bank of England. London: Swan Sonnenschein and Company. United Kingdom. Parliament. 1848a. Appendix to reports from the secret committee on commercial distress. Parliamentary papers, 1847-48. Vol. 8, part 2. London: HMSO. - - - . 1848b. First and second reports from the secret committee appointed to inquire into the causes ofthe distress. Parliamentary papers 1847-48. Vol. 8, part 1 (395) (584). London: HMSO. - - - . Committee on the Working of the Monetary System [Radcliffe Committee]. 1960. Principal memoranda ofevidence. Vol. 3. London: HMSO. Viner, Jacob. 1937. Studies in the theory of international trade. New York: Harper and Brothers.

265

The Bank of England in the Crisis of 1847

Comment

J. R. T. Hughes

The papers by Dornbusch and Frenkel, Dutton, and Pippenger cover part of the more than nine decades (1821-1914) when England was on the classical gold standard. The return to gold payments was threatened in 1825 and sequentially in 1835, 1836, 1837, and 1839. Former Governor J. Horsley Palmer's defense of the Bank of England's policy after 1836 and criticism of that defense by Samuel Jones Loyd (later Lord Overstone) led to the legislation of 1844, the Bank of England Act-Peel's Bank Act (see Palmer 1836; Loyd 1837a, 1837b; Horsefield 1953). That act essentially put into statute form David Ricardo's posthumously published Plan for the Establishment of a National Bank (1824). Loyd had been the major propagandist for the Ricardian solution-division of the Bank of England into two parts, a Banking Department to do a strictly banking business, like any commercial bank, and an Issue Department to mechanically exchange Bank of England notes for gold, and gold for Bank of England notes at a fixed gold price.! This arrangement would produce a convertible currency and, as other nations established fixed mint-par values for gold, an international regime of fixed exchange rates within the gold points would come into existence. The Bank of England Act of 1844 remained mainly unchanged2 for seven decades until 1914 when it was "suspended" for the last time. Leaving aside the period before 1844, the British monetary system was to be tested in three very different economic environments: (1) 1844-73, the era Sir John Hicks once described as the "great Victorian boom" (and which hosted the famous crises of 1847,1857, and 1866); (2) 1874-95, a period of puzzling sluggishness in British monetary affairs in which there was never any monetary strain of the sort that had once capped each boom; and (3) 1896-1914, when prices and interest rates rose again. The Bank of England, at the center of the gold standard system, learned to live in these environments by trial and error. Before 1873 the Bank was relatively large, but was surrounded by a huge, privately issued money stock (the inland bill system), a unit banking system (429 "country banks" in 1842, and still 167 banks by 1874), and an economy that periodically expanded with force, straining the monetary system at the seams (Coppieters 1955, pp. 158-59). During the long compression of the interest-rate structure in 1874-96 and stagnating prices, the economy never boomed again; there was no strain on the financial system (although when Barings threatened to suspend payments in 1890 another panic was feared); and Bank policy became conservative to a remarkable degree (Hughes 1968). After 1896 a plateau in gold reserves and an J. R. T. Hughes is professor of economics at Northwestern University, Evanston, Illinois. The comment prepared by Hughes, who did not attend the conference, was read by Lawrence H. White of New York University.

266

Rudiger Dornbusch and Jacob A. Frenkel

expanding economy resulted in the defensive Bank-policy stance noted by Pippenger. There really are two separate questions to answer here: (1) How well did the Bank of England manage the gold standard during these three very different long-term financial climates? (2) How does the gold standard system of 1844-1914 rate as an international financial system compared to what followed it after World War I? Frenkel and Dornbusch give us a close-up of the Bank during the first great financial crisis under the Bank Act of 1844. Their paper explains well the great surprise: Panic hoarding shifted from the earlier emphasis on gold to an excess internal demand for gold's representative, the Bank of England note. They correctly see that the Bank was playing its own game in 1847, not Dutton's rules of the game. When gold flowed out, the Bank had expanded its holdings of securities, replacing the gold with notes in circulation and thus "financing" the gold outflow. But that policy was probably an unintended outcome; the Bank's managers, as representatives of a private, profit-making institution, were competing aggressively for business. They kept Bank rate below the market rate of discount. As C. N. Ward-Perkins (1966, p. 264) put it in his classic paper a century later: The day after the Act became law, on 5 September 1844, the directors, taking Sir Robert's advice to heart, announced a new rate and policy for discounting, a minimum rate of 2V2 per cent being accepted for first-class three-month bills. This was competition indeed; market rate, though it had stood below 2 per cent, was tending to rise, and for the next three years minimum bank rate was consistently lower than market rate. The policy objective was no higher than that, whatever the outcome. The supply of money is defined by Dornbusch and Frenkel, excluding the circulation of private merchants' money-bills of exchange. We would not exclude them now from our various definitions of money. They would at least be part of M4. They circulated hand-to-hand as means of payment in normal times (Ashton 1953), but in a crisis they fell to heavy discounts as interest rates rose. They were dumped in part to avoid capital losses (Hughes 1960, pp. 256-74). In volume the bills were much larger than the Bank of England's own circulation, and in the great crises of 1847, 1857, and 1866 it was their cascading deluge upon the banks and discount houses that made the Bank Act of 1844 an iron lid that later had to be removed by the Treasury letter. William Newmarch's estimates show that the average volume of bills afloat, £54.89 million in the fourth quarter of 1843, had risen by an astonishing 44 percent to £78.93 million by the first quarter of 1847, while the Bank of England's own aggregate circulation (in the public's hands and in the Banking Department's reserve) had remained nearly unchanged at just under £20 million (see

267

The Bank of England in the Crisis of 1847

Tooke and Newmarch 1857, pp. 589-92). The country bank issue of notes in 1847 was about £7 million, so a total liquidation of the private money stock would put some £87 million of private money against the Bank's £20 million in Bank of England notes. In addition, bank deposits totaled £55 million outside the Bank, and the Bank's own deposits were £15 million. That made £157 million to be exchanged for Bank of England notes totalling £20 million, without considering government debt, private issues of stocks and bonds, and other forms of credit sellable or at call (for data see Coppieters 1955, pp. 149-60). The Bank under the act could not normally increase its notes except through an inflow of gold to the Bank. The Bank's gold reserve had meanwhile fallen. When panic came to a head in the autumn, the Bank's oft-denied role as lender of last resort came into play-the Bank was expected by the public and the government to be the banker for the entire nation. And while the Bank profited handsomely from its public duty to discount huge amounts at high rates, the Bank was running out of its own notes even while discounting. In the Treasury letter, the chancellor and prime minister intervened, authorizing the Bank to trade securities to the Issue Department for fiduciary notes (to be used for loans only at the highest discount rates). At the same time they promised a bill of indemnity intended to get bank officials off the hook for their violations of the Bank Act fiduciary-issue restrictions. While it is true, as Frenkel and Dornbusch emphasize, that Sir Francis Baring and others were shocked that the Banking Department's reserve of notes could be exhausted while the Issue Department's reserve of gold held at high levels, the situation was of no surprise or disappointment to Lord Overstone. He believed that credit cycles were inevitable, that the Bank Act protected convertibility, and that the financial massacre created by the dwindling reserve of the Banking Department was a necessity. Overstone's only objection was the Bank's perception that it had any public duties at all, that it had attempted to mitigate the crisis. To Overstone the origin of financial trouble was no concern of the Bank's directors. When asked by the Lords' Committee of 1848 ([1848] 1857, question 1376) about this: "You consider that in its Banking Department the Bank should conduct its Banking Business in the same way whether the drain is for foreign or internal purposes?" Overstone replied, "I apprehend that the Banking Department should not look to the causes of a Drain of Bullion, but simply look to the State of the Banking reserve." If disaster came, so be it. He was asked: "If the causes which naturally and necessarily produce changes [contractions] are actually in operation it is not desirable by any artificial means to endeavor to prevent those violent results. In fact, the patient, however painful the operation may be, must submit to it, if it becomes necessary?" (questions 1395-96). Overstone's answer was simply "Yes." It was to be rules, not authority,

268

Rudiger Dornbusch and Jacob A. Frenkel

all the way. Writing in 1857, a decade later, as Mercator in the Times, Overstone stated his principles about money supplies (quoted in Hughes 1960, p. 231) thus: The great laws which determine the monetary equilibrium of the commercial world assign to this country a certain amount of money. No internal arrangements to which we may resort can alter or suspend the law ... the monetary arrangements of this, as of every other country, must be sub-ordinate to the great principles which regulate the monetary equilibrium of the world. Any attempt to resist or modify the result of these principles can result only in confusion and embarrassment. Overstone was not one for pusillanimous countercyclical monetary policies. When in 1857 a similar sequence of events occurred-a downturn in the real economy followed by mounting financial panic and another huge liquidation of the bill system-history repeated itself, only this time the Banking Department reserve went bone dry and the Treasury letter relieved the markets of a drought of Bank of England notes (Hughes 1967). The effect was not "magical" (Ward-Perkins 1966, p. 266) as it had been in 1847, but the currency remained convertible and economic recovery came quickly enough. As in 1847 the Issue Department's bullion hoard held up nicely while the Banking Department scraped bottom and then went through the floor (Hughes 1960, app. 5). One more great crisis with application of the Treasury letter came in 1866 and then the old-time experience of the Bank under the 1844 Act became irrelevant. In 1873 the British financial crisis was not severe as it was in the United States. The British financial system had already been purged by the crash of Overend and Gurney in 1866. With the British economy facing two decades of mainly falling prices, the Bank followed market interest rates downward, but now kept Bank rate above them (Hughes 1968). Stung by the criticism it had earlier faced for putting Bank rate below market rate in 1844--47, the Bank resolutely kept its reserves high, its discounts low, and achieved the high reserve ratios noted by Pippenger. Friedman and Schwartz (1963, p. 514) would comment on a similar policy pursued by the Federal Reserve system in the 1930s that did or did not contribute to stagnation, depending upon how one views Wicksell's monetary theory (see Hicks 1977, chap 3). To some extent, as Dutton notes, the Bank continued to expand its loans in times of gold outflows, thus sterilizing the gold-outflow consequences on the money supply. When recovery came with the Alaskan and South African gold discoveries after 1896, the Bank of England maintained its conservative

269

The Bank of England in the Crisis of 1847

policies and the British investor went on a binge of foreign acquisitionshe had now become the aging rentier noted by Brinley Thomas. By 1913 more than 80 percent of the new issues in the City of London were foreign (Thomas 1954, pp. 228-30). The Bank of England, now much reduced in size relative to the great (now) amalgamated banks of London, acted defensively to maintain the center of the old financial system until the Germans marched in 1914. Although the Bank continued to maintain a restraining hold on the money market by keeping its Bank rate above the market rate (King 1936, chap. 9), without any return to the old booming economy of the earlier nineteenth century the Bank's quantitative weakness was never tested. The second factor suggested by the work of Frenkel and Dornbusch is worth additional comment. The technique of "crisis management" worked out in 1847 was applied again in 1857 and, routinely, in 1866. Originally the Bank of England itself had suggested that suchan escape hatch be put into the 1844 Bank Act. But Peel left it out of the final legislation (Horsefield 1953, pp. 110-11). Once the London financial world realized that a Treasury letter could always be counted on to be forthcoming, the Bank's crisis-proven management technique conceivably added long-term stability to expectations and thus to the gold standard itself. Conceivably. Many writers, besides Bagehot, urged the Bank to engage in countercyclical operations. Dutton finds some not very strong evidence that the Bank followed such advice, or at least its actions made it appear so. The fact is that from 1866 to 1914 the system was never tested again. The stability of the gold standard in England may well have been due to the failure of the economy to boom again before World War I. The Bank had lived for seven decades under the 1844 Act and the world had lived that long under the gold standard centered on the Bank. The experience of the Bank in those years, apart from its policy of selling government securities during panics to raise cash, evolved into our main notions about central banking. The currency had remained convertible at the center, at the Bank of England. Was the gold standard, thus managed, a good international financial system? It did not generate long periods of inflation. The periods of stagnation were followed by periods of growth. There was vast worldwide economic growth based upon free capital movement as well as the dynamic influences of England's free trade policies. To say such achievements were unimpressive implies that an example of a superior system can be found. Perhaps it was the Bretton Woods system in the two decades before it drowned in dollars. However, Nixon's speech of 15 August 1971 was an international Treasury letter in reverse, and there was no commitment ever again to redeem dollars for gold. The United States needed no bill of indemnity from anyone.

270

Rudiger Dornbusch and Jacob A. Frenkel

Notes 1. Page 1 of Ricardo's pamphlet reads: "The Bank of England performs two operations of Banking ... it issues a paper currency as a substitute for a metallic one; and it advances money in the way of loan to merchants and others.... That these two operations have no necessary connection, will appear obvious from this, that they might be carried on by two separate bodies, without the slightest loss of advantage, either to the country or to the merchants who receive accommodation from such loans." 2. There was no change in the Act of 1844 apart from periodic expansions of the fiduciary issue as country banks gave up their issuing rights over time. There had been 6361icensed issuers of private money as late as 1832. The last private issuer was taken over by Lloyd's Bank in 1921 (see Coppieters 1955, p. 159; Feavearyear 1963, p. 321).

References

Ashton, T. S. 1953. The bill of exchange and private banks in Lancashire, 1790-1830. In Papers in English monetary history, ed. T. S. Ashton and R. S. Sayers, chap. 3. Oxford: Clarendon Press. Coppieters, Emmanuel. 1955. English bank note circulation, 1694-1954. The Hague: Martinus Nijhoff. Feavearyear, Albert. 1963. The pound sterling. Oxford: Clarendon Press. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Hicks, John. 1977. Economic perspectives: Further essays on money and growth. Oxford: Clarendon Press. Horsefield, J. K. 1953. The origins of the Bank Charter Act, 1844. In Papers in English monetary history. See Ashon 1953. Hughes, J. R. T. 1960. Fluctuations in trade, industry, and finance. Oxford: Clarendon Press. - - - . 1967. The commercial crisis of 1857. In Purdue faculty papers in economic history, ed. E. T. Weiler. Homewood, Ill.: Richard D. Irwin. (First published in 1956 in Oxford Economic Papers n.s. 8 [June]: 194-222.) - - - . 1968. Wicksell on the facts: Prices and interest rates, 1844-1914. In Value, capital, and growth: Essays in honour of Sir John Hicks, ed. J. N. Wolfe. Edinburgh: Edinburgh University Press. King, W. T. C. 1936. History of the London discount market. London: Routledge & Sons. Loyd, Samuel Jones. 1837a. Reflections suggested by ... Mr. J. Horsley Palmer's pamphlet, etc. Pamphlet. - - - . 1837b. Further reflections on the state of the currency. Pamphlet. Palmer, J. Horsley. 1836. The causes and consequences of the pressure upon the money market. Pamphlet. Ricardo, David. [1824] 1951. Plan for the establishment of a national

271

The Bank of England in the Crisis of 1847

bank. In The works and correspondence of David Ricardo 4, ed. Piero Sraffa. Cambridge: Cambridge University Press. Thomas, Brinley. 1954. Migration and economic growth. Cambridge: Cambridge University Press. Tooke, Thomas, and William Newmarch. 1857. The history of prices. Vol. 6. London. United Kingdom. Parliament. [1848] 1857. Report from the secret committee of the House of Lords appointed to inquire into the causes of the distress. Reprint. Session 1, Vol. 2, (0.50). (First published as Vol. 8 [565]. ) Ward-Perkins, C. N. 1966. The commercial crisis of 1847. In Essays in economic history, ed. E. M. Carus-Wilson. New York: St. Martin's Press. (First published in 1950 in Oxford Economic Papers n.s. 2.)

General Discussion MCCAULEY noted the authors' suggestion that the suspension of Peel's Act was a good thing. He asked them to elaborate on how behavior after the crisis was affected by the suspension. MCCLOSKEY commented on the rapidity of the economy's adjustment to the crisis. If a crisis of confidence or a harvest-induced financial crisis of this magnitude could be resolved within a month, as Dornbusch and Frenkel suggest in their paper, then why be concerned about the adjustment mechanism under the gold standard? If the adjustment takes place so fast, it may be worthwhile to make the simplifying assumption that interest rates were exogenous to the British economy. DORNBUSCH restated the aims of the Dornbusch-Frenkel paper. The authors' objective was to look at a particular year in which the gold standard was in operation and to ask how the system responded to shocks. The shocks in question were not reckless credit expansion, which was impossible under Peel's Act; rather, in this instance, the disturbance eminated from the real side. Under Peel's Act, there was a presumption that when an external bullion drain took place, domestic currency in the hands of the public would decline as well. An adjustment would be facilitated. In this episode, however, the Banking Department of the Bank of England initially offset the attempt by individuals to run down their cash balances. Therefore no adjustment took-place until the Bank of England decided that it was no longer feasible to continue to intervene, at which time the Bank liquidated its consol portfolio. In a single day it lowered the price of consols, or raised the discount rate, by four hundred basis points. A lot of firms went bankrupt that day. The Bank justified its

272

Rudiger Dornbusch and Jacob A. Frenkel

action by stating that its primary responsibility was to maintain the integrity of its note issue-a comment on the Bank of England's perception of what it meant to act in the public interest. At this time, the Bank's sole responsibility was to protect convertibility of its notes; if the Bank's initiatives created fears of mills being closed or riots in the streets, that was the government's problem. The Bank of England did not think of itself as engaged in countercyclical monetary policy; its interest was to make profits. Whenever there was the risk of an internal run, its concern was to have enough consols to be able to go to the stock market and liquidate them. Dornbusch again restated the paper's conclusion: During the crisis of 1847, the Bank of England did not act as a lender-of-Iast-resort; in fact it did exactly the opposite. When bankruptcies were spreading because of an internal shortage of notes, the Bank of England went into the market and bought notes, selling consols from its portfolio. Its actions were exactly the opposite of those required of a lender-of-Iast-resort. In response to McCloskey, Dornbusch suggested that adjustment was so rapid because the interest-rate increase forced upon the market by the Bank of England was so extreme. For twenty-five years the highest interest rate observed had been 6 percent. Suddenly the interest rate was driven up to 10 percent, and a large number of banks failed. Adjustment was rapid, but there is no presumption that this was a good way of managing monetary affairs. Dornbusch elaborated upon what he meant when he stated that the gold standard was working badly: he meant relative to the post-1847 period when there was an understanding that in response. to an internal drain, the Bank of England would discount freely. If the gold standard is understood to be a regime under which notes can be issued on the strength of gold, then the pre-1847 system worked poorly. The Bank of England certainly wished to avoid the possibility of a reckless credit expansion and wished to contract credit whenever bullion was lost; but any time the exchanges were favorable and a crisis occurred, it was expected that the Bank should print money freely. Thus the lesson from the 1847 episode is the distinction between internal- and externalconvertibility difficulties. GOODHART raised a question about the October crisis. In his opinion, the April crisis was explained perfectly in the paper. The analysis of what happened when the 1844 Act was relaxed raises no unresolved problems. But in Goodhart's view, Dornbusch and Frenkel fail to adequately explain the October crisis. In April a crisis occurs, Bank rate is raised, gold starts flowing in, and there is a recovery of confidence. Yet later the authors write: "During the late summer and especially in early fall, conditions deteriorated, setting the basis for the October crisis."

273

The Bank of England in the Crisis of 1847

Goodhart wondered whether the authors had an explanation for the October crisis. DORNBUSCH explained the October crisis as the culmination of six months of gold outflows to pay for imported grains. The decline in the Bank's reserve ratio led the public to question whether the Bank could recover from the loss of its reserves. WHITE noted the authors' argument that the crisis of 1847 was a real crisis in the sense that the immediate factor precipitating it was a harvest failure. At the time, however, some individuals interpreted the event as a delayed reaction to the overissuance of bank notes starting in 1844, when the Banking Department of the Bank of England was freed to pursue any policy it desired. White wondered if Dornbusch and Frenkel had considered those arguments and whether they found evidence to support them. In response, FRENKEL referred to a quote from John Stuart Mill, who describes the crisis of1847 as a completely exogenous harvest failure that differed fundamentally from prior financial crises. MCCLOSKEY returned to the question of the economy's speed of adjustment. He noted that there were two possible interpretations of the 1847 episode. Either the economy was tremendously agile or the Bank of England was tremendously agile. Either the elasticities that facilitate the adjustment were very high or the violence with which the policy instruments moved was very great. On the basis of the evidence presented in the paper, it is not clear which is true.

PART

III.

International Experience in the Operation of the Gold Standard

6

Canada and the Interwar Gold Standard, 1920-35: Monetary Policy without a Central Bank Ronald A. Shearer and Carolyn Clark

6.1

Introduction

In the literature on the pre-World War I international gold standard, Canada has a prominent place (e.g., Viner 1924); in the literature on the interwar gold standard, Canada is but a footnote. One reason is obvious. Between the wars Canada was only legally "on gold" for two-and-a-half years (1 July 1926-January 1929), and even during this period the government was not strongly committed to the gold standard as a monetarycontrol'mechanism. In practice if not always in spirit, Canada was a flexible-exchange-rate country (Chisholm 1979; Knox n.d.). Nonetheless, at least two aspects of Canada's interwar monetary experience should be of interest to students of the gold standard. On the one hand, the restoration of gold convertibility at the prewar exchange-rate parity with the U.S. dollar was accomplished with none of the controversy and strife that accompanied, for example, the British return to gold. While the smooth return to gold was obviously conditioned by many factors affecting commodity and factor markets and the Canadian balance of international payments, appropriate monetary adjustments were crucial. These were accomplished without a central bank or other effective monetary control institutions. Indeed, Canada's Ronald A. Shearer is professor of economics at the University of British Columbia, Vancouver, Canada. Carolyn Clark is associate professor of economics, Washington State University, Pullman, Washington. The authors gratefully acknowledge financial assistance from the Social Sciences and Humanities Research Council of Canada and from the University of British Columbia President's Research Fund. They are also deeply indebted to Mr. Glenn Wright and the staff of the Public Archives of Canada.

277

278

Ronald A. Shearer and Carolyn Clark

experience might be interpreted as evidence of the irrelevance of a central bank. On the other hand, although the gold standard was reestablished painlessly, at the first sign of stress it was again abandoned. In terms of timing, Canada, not Great Britain, led the world off the gold standard. It has been argued that the 1929 suspension of convertibility was a result of the government's failure to control the money supply. There were institutional gaps in the Canadian financial system that severely limited the effectiveness of available instruments of monetary policy. We conclude, however, that the suspension of the gold standard was not a failure of monetary policy, but a plausible response to the virtual impossibility of operating a gold standard mechanism in the face of serious external stress with imperfect monetary-control institutions and very narrow gold points. 6.2

6.2.1

The Institutions

The Pre-World War I Canadian Monetary System

Canada adopted the gold standard in 1853. By 1913 it had crystallized into what Keynes called a "fixed fiduciary issue" system (Keynes 1930, p. 237). Legal tender was either gold coin or Dominion notes, a government-issued currency. Beyond a basic fiat issue ($22.5 million), these notes were subject to a 100 percent gold reserve. Chartered bank notes circulated alongside Dominion notes, but were not legal tender, and bank deposits were of increasing importance. Banks were required to convert their notes and demand deposits into Dominion notes (or gold) on demand and for this purpose held substantial reserves of both Dominion notes and gold. However, there were no constraining cash-reserve requirements.1 The Bank of Montreal acted as the government's fiscal agent and on rare occasions performed some central-banking functions (for example, during the financial crisis of 1907). However, there was no central bank; indeed, the very concept was anathma to a large part of the banking industry (McIvor 1961, pp. 109-11). 6.2.2 The Suspension of Gold Convertibility, 1914 In early August 1914, following the outbreak of war in Europe and fearing a banking crisis and a run on its gold reserves, the Canadian government released the banks from their statutory obligation to convert bank notes into legal tender and suspended the convertibility of Dominion notes into gold (PAC Finance Records, group 19, box 105, p. 2893). However, the statutory price of gold was not changed, and the goldreserve requirement for Dominion notes issued at the initiative of the

279

Canada and the Interwar Gold Standard, 1920-35

government was not suspended. These provisions meant that legally the government could not directly issue Dominion notes without purchasing their full gold backing, and hence could not directly use monetary expansion to finance government expenditures. While the suspension of gold convertibility relaxed the external discipline of the gold standard (the specie-flow mechanism was not allowed to work), a powerful discipline on direct money creation by the government remained which imparted a deflationary bias to monetary policy when Canada was off gold and thus contributed to the return to gold. We must immediately qualify the statement that the gold-reserve requirement was not suspended. In part it was. The Finance Act permitted the banks to borrow Dominion notes from the Department of Finance, and notes issued at the initiative of the banks under this provision had no gold-reserve requirement. This was the institutional basis for allegations that the gold standard and the Finance Act were incompatible (Courchene 1969; Elliott 1934). 6.2.3

The Finance Act: The Government as Lender-of-Last-Resort

By permitting the minister of finance to make unlimited advances of Dominion notes to the banks, secured only by collateral "approved by the minister," the Finance Act created a discount window, but made no provision for the other functions of a central bank. There was no board charged with comprehensive responsibility for monetary policies and no agency for conducting open-market operations (and no open market in which to operate). Lines of Credit

Administrative arrangements for the discount window varied, but by 1920 the practice of operating within negotiated lines of credit was well established. The practice entitled banks to advances up to a specified limit, on demand, at the posted interest rates. Additions could be negotiated at any time, but all lines of credit expired on 1 May each year. In general, lines of credit were not a constraint on the banking system. Between 1920 and 1935, on average, advances amounted to only 16 percent of aggregate lines of credit, and in 100 of the 180 months, to less than 15 percent. Borrowing reached 25 percent of aggregate lines of credit in only 30 months, and the maximum usage was 39 percent (November 1929). At no time did the banking system have less than 60 percent of authorized lines of credit available for immediate use. The Advance Rate

Under the 1914 Finance Act, the interest rate on advances (the advance rate) could not be less than 5 percent, although a 1917 order-incouncil permitted special advances to the banks, secured by British

280

Ronald A. Shearer and Carolyn Clark

treasury bills, at 3.5 percent per annum. These special advances aside, the advance rate was 5 percent for the duration of the 1914 act. In 1923 a new Finance Act removed the 5 percent floor. The record of interest rates on advances is set out in table 6.1. The rate was not adjusted flexibly. During the twenty-year history of the act, the ordinary advance rate (applicable to loans secured by any eligible collateral) was changed only eight times and provisions were made for special advance rates (applicable to loans secured by special issues of government securities) on only ten occasions. By contrast, over the same period the Bank of England changed its bank rate thirty-eight times and the Federal Reserve Bank of New York its discount rate forty-five times. 6.2.4 Dominion-note Issues The three components of the Dominion-notes issue for the period 1920-34 are plotted on figure 6.1, with periods when the Canadian dollar was at or above par distinguished by shading from periods when the dollar was at a significant discount. The behavior of these series tells much of our story. Table 6.1

Advance Rate, 1914-34 Ordinary Rate 22 Aug 1914-19 Oct 1917 20 Oct 1917-1 Apr 1922 2 Apr 1922-31 Oct 1924 1 Nov 1924-31 Oct 1927 1 Nov 1927-30 Nov 1927 1 Dec 1927-8 Jun 1928 9 Jun 1928-31 Aug 1928 1 Sep 1928-30 Nov 1930 1 Dec 1930-9 Oct 1931 10 Oct 1931-26 Oct 1931 26 Oct 1931-1 May 1932 2 May 1932-13 Oct 1932 14 Oct 1932-31 Oct 1932 1 Nov 1932-2 Nov 1932

5.0 5.0 5.0 4.25 4.0 3.75 5.0 3.75 a 4.5 4.5 4.5 3.0 3.5 3.5 3.5

3 Nov 1932-30 Apr 1933 1 May 1933-31 Jul 1934 1 Aug 1934-31 Dec 1934

3.5 2.5 2.5

Special Rate 3.5

3.75 3.75 3.75

2.5 2.5 3.0b 3.0 3.0 2.0 3.0b

aAdvances outstanding as of 8 June were payable at 3.75 percent; the rate on new advances was 5.0 percent as was the rate for all advances by banks which withdrew gold. b Applicable to advances which the banks were required to take by statute, 1 November 1932, renewed 1 November 1934.

281

Canada and the Interwar Gold Standard, 1920-35 c

'"

GOLD COVERED DOMINION NOTES

'"N g

1!I21

1'122

1!I23

1!I24

1!I25

1!I26

1!127

1!128

1!l29

1930

193)

1932

1933

YEAR

Fig. 6.1

Dominion-note issues, 1920-32.

The "other fiat note" series is comprised of the legal fiat issue and, until May 1926, notes issued as a special wartime loan to Great Britain. While this series shows several important drops early in the period, it is otherwise constant. The major variations are in notes issued as advances and in the gold-covered issue? The difference in the behavior of gold-backed notes and advances under the two exchange-rate regimes is striking. When the Canadian dollar was at a discount, the gold-backed issue displayed little variation, in spite of wide fluctuations in total Dominion notes. By contrast, there were wide fluctuations in advances, including obvious seasonal fluctuations. But, when the Canadian dollar was at or above par, the goldbacked issue fluctuated widely, particularly seasonally, in sympathy with the total issue. Advances were relatively stable seasonally and generally declined over time. The run up of advances in 1927-28 is an important exception which will be of concern later. The interaction between advances and gold-backed Dominion notes for the adjustment of bank reserves and the monetary base, under the two exchange-rate regimes, is interesting but not surprising. When Canada was on gold, reserves were available to the major banks at the relatively trivial cost of shipping gold from New York. Providing that the relevant New York interest rates were lower than the advance rate, this method ensured the least-cost adjustment of the banks' cash positions. Each year, when seasonal pressures pushed the Canadian dollar through the gold import point, major gold inflows occurred. But when Canada was offgold, the effective price of gold was the product of the fixed U.S.

282

Ronald A. Shearer and Carolyn Clark

price (US $20.67 per ounce) and the exchange rate. However, this gold could only be carried on a bank's books, paid into circulation, or converted into Dominion notes, at the fixed statutory price of Can $20.67 per ounce. The addition to Canadian bank reserves would not even pay for the gold purchased in New York. When Canada was off gold, the Finance Act was the least-cost method of adding to bank reserves. 6.2.5

The Instruments of Monetary Policy

Although the Finance Act did not create a central bank, it obviously provided potential instruments for the management of the monetary base. We can distinguish between those instruments affecting the quantity and those affecting the price of base money. On the quantitative side, the most obvious monetary-policy instrument was control over lines of credit, but there were other possibilities as well. Although in principle advances were available on demand under established lines of credit, moral suasion was a useful management tool as was the restriction of the eligible-collateral supply. The fiduciary issue was fixed by law, but the government was not above finding ways around the law; in principle, as long as convertibility was suspended, the goldcovered issue was also subject to management. On the price side, the instrument was the advance rate, at least after 30 June 1923. In the absence of reserve requirements, we find no instruments for managing the money-supply multiplier, apart from moral suasion. 6.3

The Return to Gold, 1920-26

From the deep discount of World War I, the Canadian dollar returned briefly to par in mid-1922, fell to a discount again in 1923, and returned definitively to par in mid-1924, although legal convertibility was not restored until July 1926 (figure 6.2). There can be little doubt that the depreciation during World War I resulted from excessive monetary expansion (Curtis 1931b; Deutsch 1940; Knox 1940). Figure 6.3 suggests that there was also a monetary explanation for the return to parity. In the early 1920s Canada experienced a monetary contraction that was both greater (18 percent versus 9 percent for M2) and more prolonged (twenty-four versus eighteen months) than that in the United States. Perhaps more significant was that while the U.S. money supply recovered sharply from its trough in 1922, the Canadian money supply remained substantially unchanged through mid-1924. No significant change occurred in the ratio of Ml or M2 to the monetary base (figure 6.4). However, with the exception of late 1922 through mid-1923 (the period of renewed depreciation), there was a steady decline in the monetary base during the re'turn to gold (table 6.2 and figure 6.5).3 Although some fluctuations in monetary gold holdings are evident,

283

Canada and the Interwar Gold Standard, 1920-35 o

111

, n

"

II II 1\ 1\

~ --r-_J-J'

\ \

\_, ' -._---------- -

POUND STERLING

CJ· .

1913

Fig. 6.2

]9]4191'519161917

1918

1919 vEAR

1920

192J

1922

~923

192419251926

1927

Discount from prewar gold parities: Canadian dollar and pound sterling, 1913-26.

particularly seasonal fluctuations when the Canadian dollar was at par (late 1922, 1924-26), over the period as a whole the entire contraction occurred in the fiat component of the base and, if we focus on the years 1920-24, almost entirely in advances. There was no central bank to force the monetary adjustment. Did the government nonetheless use the potential instruments of policy to this end? 6.3.1

Quantitative Measures: Lines of Credit

To a bank, an unused line of credit was a costless liquid asset that guaranteed almost instant access to legal tender. In principle, control over lines of credit was like control over cash reserves. Figure 6.6 shows three major reductions in aggregate lines of credit between 1920 and the resumption of legal convertibility (May 1922, 38 percent; May 1923,27 percent; and May 1925, 17 percent), which suggests the possibility of deliberate restrictive monetary policy in support of a return to gold. However, our research suggests not. Examination of applications and authorizations, housed in the Public Archives of Canada, for lines of credit for the years 1920-26, revealed no evidence that the reductions of May 1922 and May 1923 were induced by the government. The reductions appear to have been the banks' reactions to the generally lower level of prices and hence lower demands for credit. The May 1925 cuts were at the government's initiative, but involved only two banks. Each cut was justified on the grounds that the bank's request was vastly in excess of its potential needs, but there is no evidence of a

284

Ronald A. Shearer and Carolyn Clark

c

I/')

U"l

----'- .........

N

M2

UNITED STATES

".---'

-' CANADA

~-+-------.----~--~--~--.,.-----, 1~0

1~]

]~2

1~3

]~4

1~5

!~6

!~7

YERR U"l

N

c c

UNITED STATES

"

'"

'

.....

_---,

",-"-'

,.-"

CANADA

~-+----.----~--~--~--~--~----, 1920

]92]

]~2

1~3

]924

1925

]926

!q27

'!'ERR

Indexes of the money supply, Ml and M2, Canada and the United States, 1920-26 (1920 = 100).

Fig. 6.3

o

.

lQ20

192]

1922

]923

1924

1926

YERR

Fig. 6.4

Ratio of the money supply to the monetary base, Canada, 1920-26.

285

Canada and the Interwar Gold Standard, 1920-35

Table 6.2

Adjustment of the Monetary Base during the Return to Gold, June 1926-June 1926 (millions of dollars) Levels

1914

1920

1922

1924

1926

138.3 46.6 91.7

182.0 91.5 90.5

170.0 84.8 85.2

164.5 68.3 96.2

178.4 86.0 92.4

Fiat Dominion notes 22.5 Legal fiat issue 22.5 Finance Act 0 Imperial Treasury bills 0 Other 0

201.5 63.5 88.0 62.5 25.6

147.5 63.5 44.0 0 44.0

114.1 63.5 19.2 0 19.2

83.3 63.5 19.8 0 19.8

0

50.0

40.0

31.4

0

160.8

383.5

317.5

278.6

261.8

1922-24

1924-26

1920--26

-...Q

+14 +18

-

4 6

- 4

+

2

Gold Banks8 Government b

British issue TOTAL MONETARY BASE

Changes

Gold Banks Government Fiat Dominion notes Legal fiat issue Finance Act Imperial Treasury bills Other British issue TOTAL MONETARY BASE

1914-20

1920--22

+ 44 + 45 - 1

-12 - 7

-17 - 5 + 11

+179 + 41 + 88

-54 0 -44

-33 0 -25

-31 0 + 1

-118 0 - 68

+ 62 + 26

-62 +18

0 -25

0 + 1

-

+ 50

-10

- 9

-31

- 50

+223

-66

-39

-17

-122

- 62 6

8Bank gold abroad included in the monetary base. 1914 = 17.2; 1920 = 17.3; 1922 = 15.1;

1924 = 14.8; 1926 = 14.4. bGovernment excess gold reserves excluded from the monetary base. 1914 = 1.0; 1920 =

5.1; 1922 = 0.3; 1924 = 0.5; 1926 = 2.6.

plan to restrain the borrowing of either bank or to reduce the borrowing capacity of the banking system as a whole. We conclude that deliberate restriction of lines of credit was not an instrument used by the government in support of the 1924 return to parity. We have had no success in our attempts to incorporate unused lines of credit in an econometric analysis of the banks' demand for cash reserves. This result may be reasonable, indicating that the banks regarded lines of credit as free goods about which careful calculations did not have to be made.

286

Ronald A. Shearer and Carolyn Clark

MONETARY BASE

...

GOLD

/\

1\

~'~"'J ';~;:A: //

!

\

1\ I "\ ! \ ....... ! \

Ii i

1\

'-

\ __ J

\

.I

i"",,/I', \ \ ..... "... J

1

FIAT ISSUES

a

Lr

, q~20

1q:21

1q-~2

1923

1q24

1925 1926 vEAR

! 927

1928

! 929

1910

H~31

! 93:2

1q33

The monetary base and its components, 192G-32.

Fig. 6.5

Cl

o

lD

o

o

lf1

LINES OF CREDI T

o U10

0::l"1

5

-l ()

o

1920

Fig. 6.6

192]

192219231924192519261927192819291930 vEAR

Lines of credit and advances, 192G-32.

193]

1932

1933

192~35

287

Canada and the Interwar Gold Standard,

6.3.2

Quantitative Measures: Collateral for Advances

The Floating Canadian Debt

A consequence of the financing of World War I was the creation of a sizable floating debt-nonmarketable, short-term Treasury bills held exclusively by the banks. Bearing interest between 5.5 and 6 percent when the advance rate was 5 percent, these bills were prime collateral for advances under the Finance Act. On two occasions, December 1919 and May 1922, the minister of finance attempted to retire some issues before maturity; at least the 1922 case appears to have been partly a deliberate effort to reduce the fiat issue of Dominion notes. In early December 1919, the government sought the agreement of the banks to retire before maturity $125 million of the $208 million floating debt. The banks agreed to $110 million. The monetary consequences, however, appear to have been trivial. Other collateral was readily available, and the general level of interest rates was well above the advance rate. As a result, repayments of advances amounted to only $6.8 million, a small fraction of total advances outstanding at that time ($113 million on 30 November). The May 1922 effort was much more controversial because the government sought to retire the entire floating debt. Interest rates, however, had dropped sharply and sound lending opportunities had dried up. Treasury bills bearing interest at 5.75 percent or 6 percent were very attractive assets for the banks. Several banks were recalcitrant, and in the end the minister only insisted on the retirement of those bills that had been pledged as collateral for advances. His stated objectives were to reduce "as far as possible that portion of the Dominion-note issue which is not secured by gold" and, as a matter of continuing policy, "to increase our percentage of gold held against Dominion note issues" (PAC Finance Records, group 19, box 101, p. 3510). Only $49.1 million (or 34 percent of the floating debt) were in fact retired, and only a small ($16 million), temporary drop in advances was experienced. The 1922 episode can be interpreted as an attempt to implement quantitative monetary policy, but the measure was restrictive probably because the advance rate was high in relation to other interest rates and in relation to bank-lending opportunities. At best the policy was a modest operation whose effects would have occurred in any case when the bills matured in a few months. The Floating British Debt

More important, quantitatively, was the retirement of the British floating debt, a product of Canada's wartime financial assistance to Britain (White 1921; Deutsch 1940; Curtis 1931b; Knox 1940). The floating debt was in two parts: a $50 million loan of Dominion notes by

288

Ronald A. Shearer and Carolyn Clark

the government of Canada in 1917 (the "British issue"), and a succession of loans by the Canadian banks, aggregating $230 million. The British issue violated the Dominion Notes Act but was legal under the War Measures Act. Upon the expiration of most orders under the latter act on 1 January 1920, the British issue was probably illegal, but its legality was never questioned. In April 1922 the British government began making $5 million monthly payments to the Canadian government to repay the loan and retire the Dominion notes, but the Canadian government used only the first two payments to retire Dominion notes, reducing the British issue to $40 million. The issue remained at this level through February 1924 and was then reduced in two jumps to $31.4 million (March 1924) and $14.4 (December 1925), before being retired in May 1926. Thus, between mid-1920 and mid-1924, the reduction of the British issue amounted to only $19 million, about 5 percent of the 1920 monetary base (table 6.2). Had the entire British repayment been applied to the British issue, the monetary base would have been reduced by 13 percent by late 1922. The government's search for ready cash thus significantly moderated the contraction of the money supply, transferring the pressure to 1925 and 1926, after the prewar parity was again well established. This story is not one about a government doggedly pursuing a restrictive monetary policy in suport of a return to gold. More significant in the early stages of the return to gold was the retirement of the British debt to the banks. These loans were substitutes for direct governmental loans, and in the negotiations the government was a far from a passive intermediary. In effect, the government assumed a variable residual participation, offering to the banks the discount facilities of the Finance Act at a special advance rate. To the extent that this facility was used, the loans had an inflationary effect on the monetary base; but because the bills were not all discounted, the inflationary effect was smaller and less permanent than would have been the effect of a direct issue of Dominion notes of the same amount. Of the $230 million loaned to the United Kingdom by the banks, less than half were discounted at any time. As the imperial Treasury bills were retired, advances fell steadily. Again it must be stressed that the quantitative impact of debt retirement on the fiat issue of Dominion notes was, indirectly, a product of the level of the advance rate. The banks had large lines of credit and abundant collateral. Had the advance rate been in accord with market rates or perhaps below them, some part of the reduced fiat issues would have been replaced through the discounting of other securities. 6.3.3 The Advance Rate In our interpretation, the setting of the advance rate was the key to the monetary contraction necessary for the return to gold, even though it is

289

Canada and the Interwar Gold Standard, 1920-35

unlikely that the government initially conceived of the advance rate as an instrument for monetary regulation. The Finance Act of 1914 was a temporary emergency measure, and although the Treasury Board was given discretion to set a rate above the floor, it is very unlikely that there would have been a legislated floor if the advance rate had been designed as a policy instrument. (The Treasury Board itself was a body responsible for overseeing the government's fiscal position, not for managing monetary policy.)

1914-20 From its introduction in 1914 until 1924 the advance rate was unchanged, except for the introduction and expiration of the special advance rate to facilitate Britain's borrowing in Canada. Almost as soon as the rate was set at 5 percent, interest rates dropped sharply in New York (figure 6.7) so that the Canadian rate became very much a penalty rate. It is not surprising that little borrowing was done in the first few years, and most of that by smaller banks that did not have low-cost access to the New York money market. Through mid-1917, the Finance Act was not an instrument for inflationary expansion of the money supply. However, the 3.5 percent special rate for advances secured by imperial Treasury bills established in October 1917 soon became competitive with the rising New York call-loan rate, and by 1918 even the ordinary advance rate was below New York levels. In 1919 and 1920 the differential between the advance rate and New York rates became very intense, and in April 1920 the minister of finance was advised by his major adviser, the president of the Bank of Montreal, to abolish the 3.5 percent special rate because it was so low as to promote "over trading" (PAC Finance Records, group 19, box 2673). A letter was written to the Canadian Banker's Association

NEW YORK CALL LOAN RATE

ADVANCE RATE (Ordinary)

NEW YORK DISCOUNT RATE

ADVANCE RATE (Special)

o

'+.--r-----r---,----r-"-..,---r-~-__"T""'""-.____~-~__,..-__,_-r_____, 1918

Fig. 6.7

1919

1920

192J

1922

1923

1924 1925 'r'EFlR

1926

1921

1q2B

! 929

1930

1931

1932

1qJ~

The advance rate, the New York call and discount rates, 1918-32.

290

Ronald A. Shearer and Carolyn Clark

on 21 May 1920 announcing the abolition of the special rate (PAC Finance Records, group 19, box 3514). No reply is preserved in the Public Archives, but evidence from the ledger books shows that advances were made subsequently at the 3.5 percent rate, suggesting that the new policy was opposed by the banks and not implemented. Not only was the ordinary advance rate not raised when central banks around the world were raising their discount rates, this modest proposal for restrictive advance-rate policy came to naught. 1920-24

When the imperial Treasury bills were retired, the advance rate was effectively increased from 3.5 percent to 5 percent. More importantly, while the New York call-loan rate remained in the 6-7 percent range through 1920, in 1921 and 1922, like most short-term rates, it plunged to low levels. With the Canadian advance rate fixed at 5 percent, what had been an inflationary interest-rate differential now became a deflationary differential. The advance rate thus contributed to the return to gold, not as a matter of deliberate policy but by the accident of the drafting of the 1914 Finance Act and the technique adopted to finance British wartime purchases in Canada. When the 1923 Act removed the 5 percent floor, the advance rate was not changed. Although we have no direct evidence, it is plausible that the government was attempting to support the again-falling Canadian dollar. Consistent with this interpretation, the rate was not lowered in the fall of 1923 to support the seasonal expansion of the money supply, but it was lowered in 1924 when the Canadian dollar was strong. In this respect, the government behaved like a gold-standard central bank; apparently Canada was learning about monetary policy British-style. However, two points should be noted. First, advances under the Finance Act were at a low ebb in 1923. The 5 percent advance rate would have discouraged banks from taking advances to increase their reserves, but it could not have been intended to induce further retrenchment; no such retrenchment was possible. Second, unlike the British bank rate or the New York discount rate, the advance rate was not directly linked to any openmarket interest rates in Canada. There was no open money market of consequence, and even the broad psychological effects normally attributed to changes in the bank rate could not have occurred. Changes in the advance rate were a private affair between the government and the banks. We have discovered only one instance (September 1928) when the change in the advance rate was noted in the press. 6.3.4 The Gold-Reserve Ratio Although we can find little evidence of quantitative monetary policy in support of a return to gold, we have noted one attempt involving early

291

Canada and the Interwar Gold Standard, 1920-35

retirement of floating debt that was justified as improving the goldreserve ratio. Two other examples of similar policies can be cited that probably had little impact on the return to gold but which suggest the government's general frame of mind. We noted earlier that when the Canadian dollar was at par, the seasonal expansion of bank reserves was effected by the importation of gold. The seasonal demand for money and the seasonal strength of the current account combined to push the Canadian dollar through the gold import point, and gold flowed in. However in 1922 and 1924 before the gold import point was reached, the government used excess funds at its disposal to import gold, making the seasonal gold adjustment for the banks. The government's purpose can only be surmised, but it seems likely that it was attempting to improve the goldreserve ratio (figure 6.8). The actions were probably unnecessary; the normal working of the adjustment mechanism would probably have pushed the Canadian dollar through the gold import point anyway, and the effects on gold reserves were largely temporary. 6.3.5

Overview of the Return to Parity

We have not examined fully why Canada's return to gold was so unremarkable. We have ignored the real factors affecting the balance of payments and the behavior of labor and commodity markets and only asked: How did Canada, without a central bank, make the monetary adjustments necessary for the return to gold? Several points seem significant by way of summary. 1. We have no direct evidence of a conscious "grand design" of a monetary policy to restore the gold standard, although, from time to time, government policies were predicated on improving the gold-reserve ratio. The theme of an eventual return to gold recurred in public and private discussions of monetary affairs, but it was not a serious political issue nor a subject of vigorous public debate. . 2. The major instrument of quantitative monetary policy, control over lines of credit, was not used. Indeed, we can point to only two attempts by the Canadian government to implement quantitative monetary policy (through the early retirement of floating debt), with but modest effects. 3. More important were the monetary consequences of the retirement of the British floating debt, an accidental by-product of the settlement of Canada-United Kingdom wartime financial affairs. 4. Until July 1923, the advance rate could not be lowered. In subsequent months, long after relevant market interest rates had fallen, the government, for unknown reasons, chose not to lower the advance rate. We argue that this was the most important decision affecting the return to gold. Not only was the 5 percent advance rate deflationary in its own right, it created the conditions under which the retirement of the British floating debt would also be, on balance, deflationary.

CJ

CJ Cl ~

CJ

,

1914 1915

Fig. 6.8

(k:

a: N •

o -. ~Ll1

CJ

CJ 1.1:

CJ

~

o

1917

1918

,

1919

,

1920

,

192J

,

YEAR

1922

,

1923

1924 1925

1926

1927

Ratio of official gold reserves to outstanding Dominion notes, 1914-32.

1916

,

1928

,

,

1929 1930

,

193J

,

1932

i

1933

293

Canada and the Interwar Gold Standard, 1920-35

5. Occasionally during the monetary adjustment, the government showed the usual governmental penchant for accepting "easy money," e.g., the funds available by not retiring the "British issue" on schedule. However, because of the 100 percent marginal gold-reserve ratio, the government could not resort to the expediency of issuing Dominion notes to finance expenditures. With the Canadian dollar at a discount, notes issued by the government would not even pay for their own gold backing let alone finance government expenditures. We suspect that this remnant of the gold standard, when Canada was officially off gold, was of profound importance in conditioning the return to gold. It prevented the government from following a directly inflationary policy. 6.4

The Suspension of the Gold Standard, 1928-31 4

Reestablished so easily, why was the gold standard suspended again so quickly? It has been argued that the suspension was a direct consequence of excessive monetary expansion, reflecting in part a "fundamental inconsistency between the operations of the gold standard and the Finance Act," and in part faulty government policy (Courchene 1969; Elliott 1934). However, this argument is misleading. Rather than a failure of Canadian monetary policy, the 1929 suspension was a reaction to a powerful external shock whose severe, immediate monetary consequences could not be otherwise offset or contained. It demonstrated the extreme difficulty faced by a country like Canada, with remarkably narrow gold points, in adhering to the gold standard when the international monetary environment was not tranquil. 6.4.1

The Monetary-Policy Hypothesis

The foundations of the monetary hypothesis can be seen in the behavior of the Canadian money supply, which increased somewhat more rapidly than that of the United States in the late 1920s-or rather took a noticeable relative jump in late 1927 (figure 6.9). Given the fixed exchange rate, however, the money supply was endogenous; its behavior cannot explain what happened. Perhaps more relevant is the behavior of the monetary base, particularly the sharp rise in fiat Dominion notes in 1928 (figures 6.5 and 6.10). A sharp increase in advances apparently displaced gold from the monetary base-advances drove out gold. In considering how this displacement might have happened, however, we must remember that without a central bank or a government directly managing the fiat issues of Dominion notes, the size and composition of the monetary base were also endogenous. The only exogenous factors were interest rates. A monetary-policy explanation of the suspension must focus on the setting of the advance rate.

294

Ronald A. Shearer and Carolyn Clark tJ1 N

M2

,I .- -

_

-

..... _

-

oJ

\\' _

~

'" _

UNITED STATES

"

\

"

\

~-+-------r---,.----.---.,------,---..,

1926

1927

1928

1929

193Q

193J

1932

YEAR o

IJ')

tJ1

N

CANADA

g

.~/~

__ /\r-~'

f, II

---'/"

'.,,_t"\_~

UNITED STATES

t.f)

r-

~+---r------r----Y-----r---r----, 1~6

1~7

1~8

1~9

'~D

1~J

!~2

YEAR

Fig. 6.9

Indexes of the money supply, Ml and M2, Canada and the United States, 1926-32. o

o

N

TOTAL FIAT ISSUES

o VlO 0:::-

5

8

°

1926

1927

1928

1929

1930

193J

1932

YERR

Fig. 6.10

Fiat issues of Dominion notes, 1926-32.

1933

295

Canada and the Interwar Gold Standard, 1920-35

6.4.2 The Advance Rate To explain the sharp increase in advances, Courchene points to the lowering of the advance rate from 4.5 percent to 4 percent in November and to 3.75 percent in December 1927 (table 6.2; figure 6.7). The November adjustment was subsequently defended as necessary to keep the advance rate in line with the discount rate of the Federal Reserve Bank of New York, which had been reduced from 4 percent to 3.5 percent in August 1927 (Hyndman 1928, p. 53), hardly a surprising policy for a gold standard country with a stable exchange market. The December adjustment was probably part of the same policy, but it had another important consequence. To persuade the banks to purchase a special issue of three-year 4 percent Treasury notes, the government made a commitment to discount these notes at the new advance rate of 3.75 percent. This "special advance rate" restricted the government's freedom of action during the subsequent gold crises and initiated a new policy with respect to the advance rate and the Finance Act mechanism. The Advance Rate and the New Floating Debt

We will not explore in detail the new advance-rate policy, but it is useful to note it in passing. Most of the changes in the advance rate during the twenty-year history of the Finance Act occurred after December 1927 (table 6.1). With four exceptions (June 1928, September 1928, May 1932, and May 1933), these changes involved either a new special advance rate or a reduction in the ordinary advance rate coincident with a new issue of short-term debt placed directly with the banks. The advance rate was set to provide an inducement to the banks to purchase the new government securities. The government had discovered the potential of the Finance Act to support government financial operations and used it repeatedly for that purpose. However, the government also made one effort to use it for monetarypolicy purposes. 6.4.3

The First Gold Drain

Advances increased in the fall of 1927, but not substantially, and there was no drain of gold. The major increase in advances and the beginnings of the gold drain occurred later, in mid-1928, after the dramatic rise in the New York call-loan rate (figure 6.7). The Policy Response: June 1928

It has been argued that the government did not understand the problem of the gold drain and did nothing (Elliott 1934). In fact, as the gold drain began, the Department of Finance attempted an almost classic central-bank response involving both a rise in the advance rate and a

296

Ronald A. Shearer and Carolyn Clark

"gold device." To our knowledge, this was the first deliberate centralbanking operation in Canada. The exchange rate went outside the gold export point on 31 May 1928. On 5,6, and 7 June one bank exchanged $7 million of Dominion notes for gold. The next day the advance rate was raised to 5 percent for all new advances secured by ordinary collateral and all new and outstanding advances of those banks, "which hereafter withdraw gold" (PAC, Treasury Board Records, group 55, box 645). Unfortunately, the monetary effect of the new advance rate was muted by the special advance rate of 3.75 percent introduced in December 1927. The government clearly understood the problem but was constrained by the earlier commitment. Given the lack of an active, open money market with international connections and the secrecy governing the change in the advance rate, the increase was clearly not intended to affect international capital flows in the normal way. Indeed, it is better to regard the new advance rate as a "gold device," a method of increasing the cost to banks of shipping gold. As such, it was successful for a time. Although the Canadian dollar remained outside the gold export point, there were only three shipments of gold during the balance of June and early July, and these were by the government. However, in mid-July, when the discount on the Canadian dollar increased, reaching 1/2 cent, private shipments occurred again. Thus, when potential arbitrage profits were very large, advance-rate policy failed. No further policy response occurred or was necessary. A sudden but seasonally normal recovery of the exchange rate stopped the gold outflow. The crisis apparently over, in early September the advance rate was reduced from 5 percent to 4.5 percent. 6.4.4 The Second Gold Drain For two months the Canadian dollar remained strong, but in midSeptember, when it would normally strengthen, it weakened; apart from a brief burst above the gold import point in late November (when $22.5 million of gold was imported), it remained below par but within the gold point. In early December it plunged through the gold export point, and gold began to flow out. From 30 November 1928 to 28 February 1929, $60 million of gold was shipped, although because of purchases of newly mined gold from the Royal Mint, official reserves fell by only $50 million. Then the gold drain stopped. Indeed, private shipments stopped at the end of January 1929. The government's shipment of $18 million to meet foreign commitments during this period, while profitable, is puzzling, given their concern about the gold-reserve ratio. The Canadian dollar remained well outside the gold export point until mid-March 1930, and the discount actually reached 2.5 cents after the stock-market collapse.

297

Canada and the Interwar Gold Standard,

192~35

The Policy Response

The government's initial reaction to the second gold drain was different; the advance rate was not increased, but other gold devices were invoked. In mid-December 1928 the Canadian banks apparently agreed not to ship gold for their own account, leaving open the question of shipments on behalf of U.S. banks. Standard gold devices were used. By law, gold should have been made available in Montreal, the normal shipping point to New York. It was only made available at Ottawa, adding to transportation costs (New York Times 1927a, 1927b). In December 17 percent and in January 70 percent of the gold paid out was bullion or British gold coin rather than U.S. coin. This gold had to be assayed before it could be accepted by the U.S. authorities, adding to transactions costs. When in mid-February 1929 the discount on the Canadian dollar increased so that these gold devices could no longer be relied upon, it seems to have been agreed that the banks would charge prohibitive agent fees for handling the gold transactions of foreign banks (as a matter of standing policy, the government would only make gold available to Canadian banks) (Journal ofCommerce 1929a 1929b). When the discount became so large that the necessary commission rates were implausible, say by mid-March 1929, the banks appear simply to have declined to purchase gold for foreign banks (Commercial and Financial Chronicle 1929, pp. 1107,1978). These gold devices were effective. From the first week of February 1929 until December 1930, there were no private shipments of gold from Canada. 6.4.5 The 1930 Respite The exchange rate returned within the gold export point in mid-March 1930, and, apart from a two-week drop below the export point in early May (when again no gold flowed out of Canada), the rate remained within the gold points until mid-December. Indeed from July on, the Canadian dollar had several brief bursts of strength, approaching the estimated gold import point but never decisively breaking through it. The 1930 Gold Inflow

On most occasions when the exchange rate touched the gold import point, there were sizable shipments of gold to Montreal, but there were also sizable shipments when the rate was well below the gold point. Over the whole period, July through November, Canada received $36 million of gold, almost 60 percent of the drain during the crisis of 1928-29. The responsiveness of gold flows to very marginal, and in some cases negative, profit margins astonished financial commentators. In part the shipments were a response to market incentives created by the government and in part a result of governmental purchases.

298

Ronald A. Shearer and Carolyn Clark

Although we have no direct evidence, there is strong circumstantial evidence (from reports in the financial press) that the government was purchasing gold in New York to improve Canada's gold-reserve position. Such a policy was strongly urged at this time by the Canadian Bankers' Association who were disturbed by the sharp decline of the gold-reserve ratio below "what is deemed adequate . . . in the chief countries maintaining the gold standard" (PAC Finance Records, group 19, box 105). To the bankers, it was a matter of international confidence in Canadian banks with extensive international business; to the government (needing to roll over 44 percent of Canada's funded debt in the next four years), it was a matter of securing "the lowest possible interest charges" (Canada, House of Commons 1930). We estimate that government shipments were between $15 million and $16 million, over 40 percent of the total gold inflow. Policy Measures: The Finance Act

Monetary policies reinforced the effects of gold purchases on the official gold-reserve position, creating market incentives to ship gold to Canada. In spite of a sharp decline in interest rates worldwide, including central-bank discount rates, the advance rate remained constant at 1928 levels, 4.5 percent (ordinary) and 3.75 percent (special). The sharp widening of the differential between the advance rate and New York interest rates (figure 6.7) created an incentive to import gold, either to repay advances or as a substitute for advances to augment cash reserves, in effect temporarily reducing the gold import point. Some sizable shipments were made by U.S. banks, and there were reports of American banks diverting funds from New York to lend in Canada. Advances to the banks fell by $21 million-59 percent of the total gold inflow. This decline was concentrated in July and August, before the seasonal expansion in note circulation and in the demand for cash reserves, and essentially paralleled the gold inflow. A policy implemented 1 May, if maintained, would have strengthened this effect. In an unprecedented action, the government severely curtailed authorized lines of credit. Bank protests were loud and difficult to ignore. The policy was rescinded almost immediately, although threats of similar restrictions were repeated the following year. The government reinforced these measures by using its deposits with the banks to pay for a portion of the gold purchased by the mint, rather than issuing Dominion notes, the normal practice. Through November, $3.6 million of gold so purchased was added to "excess" gold reserves, improving the official gold-reserve ratio without adding to the cash reserves of the banks. The 1930 episode provides strong evidence that as in 1922 and 1924, the

299

Canada and the Interwar Gold Standard, 1920-35

government had a policy designed to improve its gold-reserve position. Three instruments were used: direct purchases of gold in New York, creation of excess gold reserves, and a penalty advance rate. The combined result was a small increase in Dominion-notes outstanding (a normal seasonal adjustment) and a dramatic rise in the gold-reserve ratio, from 37.7 percent on 30 June to 54.1 percent on 30 November. Over the next five weeks, the normal seasonal drain of gold from Canada occurred. With the dollar below the gold export point, $40 million of gold was shipped to New York. A few days after gold shipments in mid-January had drained the last of the gold acquired in the preceding months, the dollar recovered to within the gold points, where it remained until early June 1931. 6.4.6

The Third Crisis

The final episode began on 10 June 1931 when the exchange rate went outside the gold export point and culminated in October with an official embargo on gold exports. In between, the discount on the Canadian dollar ranged from 20/64ths during the summer, to one cent after 15 September, and plummeted to fourteen cents just before the embargo was declared. But, in spite of these powerful market incentives, only $35 million of private gold shipments occurred, $18 million in June. During July and August the outflows substantially stopped, resuming to a limited extent in September and October before the embargo was declared. In the five weeks between 15 September and 19 October when the profit incentive for shipments was exceptional, only $8 or $9 million was exported.

Gold Policies We found no records in the Public Archives relating to the government's gold policies in this period. However, it is unlikely that even the June outflow reflected a policy of unrestricted gold exports, and, again, the cooperation of the banks had to bea key ingredient. But this time the implicit restriction of exports elicited virtually no comment in the financial press. The policies of 1929-30 were probably still fresh in everyone's mind, so a reintroduction of such measures caused no surprise. During September and October, with the market inducements for gold exports exceptionally intense, the government allowed a trickle of gold out of official reserves, maintaining the image, if not the substance, of being on the gold standard. But there was a new element in the problem. For the first time the discount on the Canadian dollar was so deep that it was profitable for the public to convert relatively small quantities of Dominion notes into gold and to demand the honoring of the gold-payment clauses of some bond issues. The government's response ranged from bureaucratic harassment to outright refusal to convert. The convertibility

300

Ronald A. Shearer and Carolyn Clark

of Dominion notes and refusal to honor gold clauses became for the first time political issues both inside and outside the country. The British suspension of gold provided strong and timely justification for a formal Canadian embargo. 6.4.7

Overview of the Suspension of the Gold Standard

The suspension of the gold standard in 1929 was achieved almost as quietly as the resumption in 1924-26-although a significant political uproar occurred later. In summarizing, we would emphasize the following points. 1. The impetus behind the 1929 suspension was the sharp rise in the New York call-loan rate which put severe stress on the Canadian foreignexchange market. 2. Contrary to some suggestions, the government was not unaware of the nature of the problem and initially attempted tried-and-true centralbanking techniques to deal with it-in the first crisis gold devices and advance-rate policy, and in the second crisis more gold devices-before secretly imposing an embargo on gold exports. In subsequent crises the government continued the embargo until it was legitimized by the British suspension. The government had the full support of the Canadian bankers and the tacit approval of the government of the United States. We surmise that the lesson drawn by the Canadian government from this experience was that classical policies could only cope with minor external pressures. Although it can be argued that the measures were half-hearted at best, the situation was very difficult. The Canadian gold points were exceptionally narrow, and some Canadian banks were very alert to potential gold-arbitrage profits. Moreover, there was a serious institutional gap. Canada had no open money market with international connections, so while the response of international capital flows to more vigorous use of the advance rate might have been in the right direction, it would have been sluggish. 6.5

Conclusions

Canada's interwar experience with the gold standard is paradoxical. Without a central bank, but as a result of interest-rate policies put in place by accident, the gold standard was quietly and efficiently restored. However, when the Department of Finance discovered its capacity to act like a central bank, its operations were a failure and the gold standard was suspended at the first sign of significant external stress. Part of the reason for the failure was that it had earlier launched a policy of using the Finance Act to support short-term government financing, which predictably undermined the developing concept of gold standard centralbanking. The major reason for the failure, however, was the rigidity of

301

Canada and the Interwar Gold Standard,

192~35

the gold standard in the Canadian case. In the absence of substantial excess gold reserves, a 100 percent marginal reserve requirement left very little room to maneuver in the event of a gold drain; and the gold points between Canada and the United States were narrower than for any other pair of major industrial countries. Exchange-rate movements that would have had no consequences elsewhere, elicited major gold movements and potentially serious monetary distress. Although from time to time the government formulated policy with an eye to the gold-reserve ratio, even when Canada was formally on gold the government showed no commitment to the gold standard as a monetarycontrol mechanism, other than for seasonal adjustments. The discipline was too severe.

Notes 1. Against deposits there were no reserve requirements; the banks held whatever "prudent practice" and market incentives dictated. There was a 100 percent reserve requirement against any "excess circulation" (over paid-up capital) of bank notes, but econometric evidence suggests that the banks largely treated these reserves as a substitute for reserves they would otherwise have held in their vaults (Clark and Shearer 1975). 2. We define the gold-covered issue as that part of the Dominion-note issue against which 100 percent gold reserves were required, i.e., the total issue of Dominion notes less the fiat issues permitted under the Dominion Notes Act and the Finance Act, and the 1917 "British issue." The gold-covered issue is not quite identical with actual gold reserves because of small holdings of excess gold reserves. 3. We have included in the monetary base gold held by the banks outside Canada on the grounds that this gold was almost immediately available to the banks for use in Canada, if necessary, at trivial transactions costs (Clark and Shearer 1975). On the other hand, we have excluded from the monetary base any excess gold held by the government over and above the legal requirement on the grounds that this gold was not monetized. Alternative definitions of the monetary base, excluding bank gold abroad or including government excess gold, do not affect the general conclusions drawn here. 4. The data on gold flows, daily exchange rates, and gold points (and their interpretation on the basis of material in the Public Archives and the financial press) on which this section is based are fully documented in Clark and Shearer 1980.

References Canada. Dominion Bureau of Statistics. 1949. The Canadian balance of international payments, 1926 to 1948. Ottawa. - - - . 1933. Royal Commission on banking and currency (Macmillan Report). Ottawa. - - - . House of Commons. 1928. Select standing committee on banking and commerce. 21 March testimony by G. W. Hyndman.

302

Ronald A. Shearer and Carolyn Clark

- - - . 1929. Report of debates in the House of Commons. Ottawa. - - . 1930. Budget speech (1 May). Ottawa. Chisholm, D. 1979. Canadian monetary policy, 1914-1934: The enduring glitter of the gold standard. Ph.D. diss., Cambridge University. Clark, C., and R. A. Shearer. 1975. The chartered banks' demand for cash reserves, 1920-1934. Vancouver. Mimeo. - - . 1980. Canada's suspension of the gold standard, 1928-1931. Vancouver. Mimeo. Commercial and Financial Chronicle. 1929. Courchene, T. 1969. An analysis of the Canadian money supply 19251934. Journal of Political Economy 77 (June): 363-91. Curtis, C. A. 1931a. Canada and the gold standard. Queens Quarterly 38 (Winter): 104-20. - - - . 1931b. The Canadian banks and war finance. Contributions to Canadian economics. 3: 7-40. - - - . 1931c. Present position of Canadian dollar largely due to weakness in Finance Act. Monetary Times 67 (16 Oct.): 4, 13. - - - . 1932. The Canadian monetary situation. Journal of Political Economy 40 (June): 315-37. Deutsch, J. J. 1940. War finance and the Canadian economy, 1914-20. Canadian Journal ofEconomics and Political Science 6 (Nov.): 525-42. Elliott, G. A. 1934. Canadian monetary policy-drift, domestic management, and debts. Papers and Proceedings of the Canadian Political Science Association 6 (May): 251-62. Journal of Commerce. 1929a. (21 Feb.), p. 9. - - . 1929b. (23 Feb.), p. 9. Keynes, J. M. 1930. A treatise on money. Vol. 2 of The applied theory of money. London: Macmillan. Knox, F. A. N.d. An introduction to money, banking, and international finance. Kingston, Ont. Mimeo. - - - . 1940. Canadian war finance and the balance of payments, 19141918. Canadian Journal of Economics and Political Science 6 (May): 226-57. McIvor, R. C. 1961. Canadian monetary, banking, and fiscal development. Toronto: Macmillan. New York Times. 1927a. (8 Mar.). - - . 1927b. (9 Mar.). Public Archives of Canada (PAC). Department of Finance Records. Group 19. - - - . Treasury Board Records. Group 55. Viner, J. 1924. Canada's balance of international indebtedness, 19001913. Cambridge: Harvard University Press. White, T. 1921. The story of Canada's war finance. Montreal.

303

Canada and the Interwar Gold Standard,

Comment

192~35

Charles Freedman

Shearer and Clark have written a very detailed, scholarly analysis of monetary policy in Canada in the period between the end of the First World War and the creation of the central bank. As an economist working in a central bank and trained in the type of monetary theory that takes the existence of a central bank as an institutional datum, I found the discussion of the mechanics of a system without a central bank both interesting and useful. The principal weakness of the Shearer-Clark paper is the lack of a carefully articulated theoretical model that would place the analysis of the behavior of the financial sector within a general-equilibrium context and would permit us to contrast such behavior with that implied in the current institutional environment. A particularly important aspect of such a formalization would be to focus the attention of the reader on which variables are endogenous or exogenous in different possible regimes and which exogenous shocks were affecting the system. I would like to begin my comments on Shearer and Clark by sketching out such a model. A fairly standard textbook model of an open economy is comprised of four equations-(l) an IS curve in which output is determined as a function of the interest rate and exchange rate; (2) a price equation in which prices adjust to the gap between output and full-employment output; (3) a balance-of-payments equation in which the sum of the current account, the capital account, and the change in international reserves is set equal to zero;! and (4) an LM curve in which the demand for central-bank money is equated to its supply. The supply of centralbank money in such a model is equal to the sum of domestic assets and net foreign assets (i.e., international reserves) held by the authorities. Because of the importance of the LM curve in the discussion that follows, it is worth writing it explicitly for the case of an economy with a central bank that imposes mandatory reserve requirements. (1)

kD(·)

+ C(·) + RE(') = NDA + NFA + RB(')'

The demand for central-bank money or base is equal to required reserves (kD) plus currency (C) plus excess reserves (RE), and the supply of central-bank money is equal to the net domestic assets (NDA) plus the

Charles Freedman is chief of the Department of Monetary and Financial Analysis, Bank of Canada. The views expressed in this comment are those of the author and no responsibility for them should be attributed to the Bank of Canada.

304

Ronald A. Shearer and Carolyn Clark

net foreign assets (NFA) of the relevant authorities plus the borrowed reserves of the banking system (RB), if any. There are three different regimes that can be analyzed in the context of this simple model. In the pure flexible-exchange-rate system, the level of international reserves, NFA, is held constant and the endogenous variables are the interest rate, output, prices, and the exchange rate. The level of NDA is an exogenous variable. In a fixed-exchange-rate world with sterilization of changes in international reserve holdings, changes in NDA are used to offset changes in NFA and the endogenous variables are the interest rate, output, prices, and the level of international reserves. Finally, in a fixed-exchange-rate world without sterilization, the authorities do not attempt to use NDA to offset changes in NFA and the endogenous variables remain the interest rate, output, prices, and the level of international reserves. Two basic sets of propositions are derived from this type of model. The first relates to the effect on the system of an exogenous shock in the capital flow or the current account. The second focuses on the effects of an exogenous change in the level of net domestic assets held by the monetary authorities. In analyzing the structure of the Canadian economy in the period between World War I and the establishment of the Bank of Canada in 1935, one can treat the equations representing the equilibrium in the goods market, the determination of prices, and the balance-of-payments equilibrium as being roughly similar in structure to those in the post-1935 period. There are, however, a number of important changes that must be made to the equation representing the financial structure of the economy. First, reserves against deposits were not mandatory. Second, notes issued by the chartered banks above a certain level had to be backed 100 percent by either Dominion notes or gold. Third, under the Finance Act the banks could borrow Dominion notes at an advance rate set by the government of Canada~ Fourth, the government required 100 percent gold backing for issues of Dominion notes beyond the fiduciary issue except for those issued under the Finance Act. Fifth, the Canadian money market was very underdeveloped and Canadian banks made extensive use of the New York money market. Ignoring the fiduciary portion of the note issue and focusing only on marginal responses, one can write the equation that shows the equality between the demand and supply of Dominion notes as (2)

DN(·)=GG+FAB(·),

where DN is the demand for Dominion notes, GG is government holdings of gold, and FAB are bank borrowings under the Finance Act. The demand for reserves by the banking system and the supply of reserves to the banking system can be depicted as

305

(3)

Canada and the Interwar Gold Standard, 1920-35

k( )D( 0

0

)

+ BN(

0

)

= GB + DNB,

where BN are bank-notes outstanding, GB are bank holdings of gold, and DNB are bank holdings of Dominion notes.2 Adding the two equations one gets (4)

k( 0)D(·)

+ BN(·) + [DN( 0) - DNB] =GG+GB+FAB(o).

This equation more closely resembles the traditional LM equation in that the sum of the precautionary reserves held against deposits and the currency held outside the banking system is equal to the gold holdings plus borrowings by the banks from the authorities. Note that there is no equivalent to NDA in this system. Writing the LM equation in this form immediately raises a number of questions. First, did the government hold foreign exchange (e.g., deposits in New York banks) against which no notes were issued but which could be transformed into gold if desired? The authors indicate that the government on occasion made direct purchases of gold in New York but do not explain how these were financed. Second, what determined whether a gold inflow was held by the banks or by the government? Third, were the banks prepared to use their foreign-currency liquid assets in New York or their ability to raise deposits from U.S. residents to augment their gold holdings when they were short of reserves? That is, could the banks create a capital inflow by selling U.S. dollar assets and bringing the gold proceeds back to Canada? Rich (this volume, chap. 12) cites a number of sources that indicate the banks did this in the period preceding World War I. All these questions relate to the endogeneity of the reserve base of the chartered banks. A second set of questions relates to the determinants of the desired holding of reserves by the chartered banks. First, what factors influenced the behavior of the ratio of reserves to deposits of the banks? Rich has suggested that in the period before World War I there was a significant relationship between that ratio and income such that when income rose, the reserve requirement was permitted to fall, and the stock of money could increase without an increase in the level of reserves. Shearer and Clark suggest that there was no such relationship in the postwar period, at least in the early 1920s, but their analysis is in terms of the overall money multiplier. The latter can of course be influenced by a switch between notes (with 100 percent reserves) and deposits. Second, what factors entered into the determination of the use by the banks of advances, a device on which the authors focus a great deal of attention in their study? One possibility is the differential between the rate charged on advances and the rate available on liquid assets in New York (as suggested by the authors). Alternatively, the relevant differential could be that between

306

Ronald A. Shearer and Carolyn Clark

the rate charged on advances and that available on assets in Canada, either explicitly in the Montreal money market or implicitly in terms of the marginal return on an increase in bank loans to business. The different assumptions lead to very different conclusions regarding the effect on the variables of the system of a change in the rate on advances. Consider, for example, the effect of a decline in the rate on advances. If the banks used the funds from the now cheaper advances to invest in V.S. shortterm assets, the results would be a capital outflow, a reduction in gold holdings of the authorities, and no change in the Canadian money stock. There would simply be a reduction in gold-backed Dominion notes to offset the increase in Dominion notes issued under the Finance Act. (This result would be equivalent to the effect of an increase of NDA on NFA in a fixed-exchange-rate system with infinitely elastic capital flows.) Alternatively, if the advances were used to buy Canadian liquid assets or to increase loans to Canadian business, the result would be an increase in the Canadian money stock. Thus the specification of the functional arguments in the expression for borrowed reserves can be very crucial in determining the outcome of conceptual experiments regarding a shift in exogenous variables. Shearer and Clark analyze the period leading to the restoration of the gold standard and that leading to the departure from the gold standard. In the former they emphasize the role of advances under the Finance Act, and in the latter they focus on both the balance of payments and the advances under the Finance Act. In their discussion of the period leading to the return to the gold standard, Shearer and Clark suggest that the main factor permitting such a return to the gold standard without undue hardship was the decline of the Canadian money stock both absolutely and relative to the V.S. money stock. Since the multiplier was fairly constant over the period, this decline resulted from a decline in the monetary base. Turning to the components of the monetary base, one can see that gold was fairly flat over the period and that the decline in base resulted from the decline in fiat issues related to advances under the Finance Act. The authors do not address directly the question of why gold holdings were relatively stable. Presumably, this stability resulted from developments in the Canadian balance of payments. Given the emphasis in the later part of the paper on capital flows and their effect on total gold reserves, it would have been useful, if only by way of contrast, to discuss these developments in the first half of the decade. Since it was the decline in the fiat issues that turned out to be the crucial element in the decline of the base, the authors spend the greater part of the first section of the paper discussing possible causes of this decline, including the setting of lines of credit for advances, the availability of

307

Canada and the Interwar Gold Standard, 1920-35

collateral for advances, and the rate on advances. In the main, it was the latter two factors and most especially the rate on advances that were crucial in the reduction of the monetary base. Following the argument made above, it must have been the return on Canadian investments or loans that was the alternative rate relevant to the banks and not a U.S. rate since in the latter case there would have been offsetting movements between gold-backed not~s and those issued under the Finance Act. I think that the authors are correct in concluding that there was no conscious grand design in monetary policy to restore the gold standard. George Watts, a long-time student of Canadian monetary policy, assures me that there was no body of principles generally recognized by the Canadian authorities at the time that permitted them to act in such a way as to ease the return to gold. In their discussion of the later period, the authors focus on both the advance rate set by the authorities and on the changes in international reserves resulting from balance-of-payments flows. Again, one must raise the question of the considerations that influenced the banks in their use of advances. If it were the comparison with U.S. rates, then there would have been a link between the rate on advances, capital outflows, and hence the gold stock, without any change necessarily taking place in the money stock. However, to the extent that the banks responded to the difference between the rate on advances and interest rates in Canada, a relatively low rate on advances would lead to an increase in the Canadian money stock and hence increases in output, prices, and a gold outflow. The authors focus mainly on the former scenario although recognizing that there was also an increase in the money stock. An alternative hypothesis that would be consistent with the data is that a key element in explaining the money-stock increase in Canada in the late 1920s was the Canadian expansion of this time which derived from the North American boom of the period. The increase in income caused increases in both loans and money demand which were satisfied by the chartered banks. The latter, in turn, were able to satisfy their desire for reserves by borrowing under the Finance Act at rates that were very attractive. To distinguish between the various hypotheses would require a careful articulation of a general-equilibrium framework. There are a number of other interesting elements discussed in the Shearer and Clark paper, such as the relationship between the price of gold and the relative costs of borrowing in New York and borrowing under the Finance Act, the use of gold devices, and the response by the authorities to the gold flows at the end of the 1920s and in the early 1930s. The detailed and scholarly analysis of these and other aspects of Canadian monetary policy in this period will prove very useful to the student of Canadian history between the wars.

308

Ronald A. Shearer and Carolyn Clark

Notes 1. By specifying the arguments carefully, this equation can be consistent with either stock models or flow models of external transactions. Furthermore by setting certain partial derivatives equal to infinity, this equation can also represent a world of infinitely elastic capital flows. 2. The assertion of Shearer and Clark (note 1) that the banks behaved as if their reserves against excess note issues were a substitute for cash reserves they would otherwise have held can be captured by either putting a coefficient between 0 and 1 on BN or putting BN as an argument in the k function.

General Discussion CAGAN was struck by the authors' statement that Canada returned to the gold standard by accident, for the same statement has been made about the United States in 1879. In the United States the government said it was going to go back on gold, and, with the implementation of no particular policy, the dollar just went back. Perhaps if by accident a country can get close enough to equilibrium, then the market will do the rest of the work, because no one has an incentive to ask for gold as long as people are convinced that the government will always be able or willing to take the steps necesssary to maintain convertibility. In other words, once the economy is within a certain distance of equilibrium, it snaps back to equilibrium, and any subsequent adjustment takes place gradually over time. WHITE drew attention to a number of references in the paper to the "imperfect" development of monetary mechanisms. White was curious to have the authors specify exactly what was lacking. SHEARER responded by citing the difficulty of controlling the monetary base. Canada had no central bank to worry about the availability of credit. There was no provision for an agency to conduct open-market operations. There was no money market, and virtually no market in which short-term securities were traded. There was a discount rate fixed at minimum by statute, but the discount rate had no signaling effect, at least to the economy as a whole. It wasn't even published. On only one occasion in the period studied did newspapers comment on a change in the discount rate. BORDO asked about the role of the Bank of Montreal and other large Canadian commercial banks in this period. He thought that in some ways the Canadian experience may have been similar to the experience of the United States in the early part of the nineteenth century, e.g., the Suffolk banking system. In particular, having a large commercial bank committed to the maintenance of convertibility may be sufficient to ensure the stability of the system.

309

Canada and the Interwar Gold Standard, 1920-35

SHEARER noted that the Bank of Montreal was the government's fiscal agent. It held the government's bank account and was a source of advice on monetary matters. However, there is no evidence that it performed any central-banking functions. Shearer returned to the issue of the paper's exogeneity assumptions. Essentially, the authors assume that two variables were exogenous: the level of British indebtedness to Canadian banks and the level of interest rates. Shearer reported the results of econometric work on banks' demands for advances and reserves. The banks' demands for advances appear to have been determined on the basis of a profit-maximizing decision. The determination of the demand for bank reserves, on the other hand, is very complicated indeed. Variations in the level of notes and deposits outstanding seem to have very different effects on the level of reserves. Moreover, reserve behavior appears to have depended on whether Canada was on or off the gold standard. Finally, Shearer commented on the role of gold in the early period. Why was its level constant? The answer is that Canada was on a flexible exchange-rate system and there was an embargo against the export of gold. Little gold moved out of either the banks' or the government's reserves.

7

Operations of the German Central Bank and the Rules of the Game, 1879-1913 Paul McGouldrick

Allegedly, the "rules of the game" for central banks before 1914 prescribed that central banks expand or contract their portfolios of earning assets in some proportion to gains and losses in gold and other reserve assets. This paper confirms and enlarges on Arthur Bloomfield's finding that at least one important central bank did just the opposite: Over cycle phases from troughs to peaks and from peaks to troughs, the German Reichsbank consistently expanded its portfolio when losing gold and contracted its portfolio when gaining gold. Since the former occurred during cyclical upswings and the latter during cyclical downswings, a reasonable implication might seem to be that German business fluctuations were exacerbated thereby. And yet, I find that gold inflows had no cyclical pattern and that both the monetary liabilities of the Reichsbank and all high-powered (base) money in the hands of the banks and the nonbank public moved countercyclically. Swings in Reichsbank money issues, backed by the Reichsbank portfolio, were more than offset over cycle phases by opposing swings in Reichsbank money issues, backed by gold. As a consequence, the German banks and nonbank public-not foreign sources-withdrew gold during upswings and returned gold to the Reichsbank during downswings. Paul McGouldrick is professor of economics at the State University of New York at Binghamton. The research underlying this paper was started on a Fulbright research fellowship in West Germany and was aided by a research grant from the Deutsche Bundesbank. The paper has benefited from criticisms by Heywood Fleisig, Knut Borchardt, Richard Tilly, Michael Bordo, Anna J. Schwartz, members of the Economics Workshop (State University of New York, Binghamton) and participants in the NBER Hilton Head conference of March 1982 at which a preliminary version was delivered. Of course, all errors of commission and omission remain the responsibility of the writer.

311

312

Paul MeGouldrick

I hypothesize that the rules of the game have been misinterpreted for Germany at least. The true rule and the only one (apart from "structural" rules such as acting as a lender-of-Iast-resort) was to target on the exchange rate with London, which meant, under classical gold standard conditions, targeting on the price of gold. Success in this exercise is indicated by the absence of systematic cyclical fluctuations in German gold imports. The instrument of policy was the bill discount rate, which was changed over German business cycles so as to eliminate procyclical fluctuations in German imports of short-term capital. By purchasing unlimited quantities of bills offered at the prevailing discount rate, the Reichsbank made bills such close substitutes for Reichsbank money itself that errors of policy were minimized by the dampening effect of changing Reichsbank money and a close liquidity substitute in opposite directions when bills were purchased. The results of the tests I made, I believe, validate the theoretical explanation of Reichsbank policy sketched above. Moreover, the demand for Reichsbank money as estimated from quarterly data, using the average bill discount rate and estimated German permanent income as explanatory variables, is highly stable. Finally, German cycles were mild compared either to those in other countries at the time or in West Germany after 1950. I conclude that Reichsbank policy was successful in reducing, although certainly not eliminating, German business cycles in the 1879-1913 era. Perhaps the most interesting implication of this study is that in the pre-1914 era, the sources of high-powered money, as well as its volume and fluctuations, mattered.

7.1

Reichsbank Structure and Operational Features

The Reichsbank was established by the German parliament (Reichstag) in 1875 and began operations in 1876. At the time, Germany was converting from the many currencies of the former thirty-one sovereign states to one currency unit (the Reichsmark) and also changing from a silver to a gold standard. The 1870s were also years of pronounced cyclical fluctuations due to an exogenous event: The receipt of a huge war indemnity from France causing first a boom and then a severe and prolonged depression. Technical problems of currency and metallic changes occupied the Reichsbank during the first three years of its operation. To eliminate that period dominated by political and institutional change, this study begins with a cyclical trough year, 1879, and concludes with a cyclical peak, 1913, just prior to World War I. The legislative founders of the Reichsbank modeled it on the Bank of England, with modifications they regarded as improvements. Since many readers are familiar with the Bank of England's structure and operations

313

Operations of the German Central Bank, 1879-1913

before 1914, the salient features of Reichsbank structure and operations are perhaps best explained by contrasting them with the English model. 1. Like the Bank of England, the Reichsbank was subject to goldreserve rules. Specifically, it had to hold gold and subsidiary money equal to at least one-third of its note liabilities. Like the Bank of England, the Reichsbank also had a second and more stringent rule: notes in circulation could not normally exceed in value the sum of reserves plus a fixed issue covered by the portfolio (raised at irregular intervals after 1876). However, an outstanding difference was that the Reichsbank was authorized to issue additional notes routinely upon payment of a federalgovernment tax equal to 5 percent at annual rate of the additional notes. In Reichstag debates, this feature was advocated as ensuring flexibility of the currency while imposing a cost penalty on the Reichsbank for additional note issues. 2. Like the Bank of England, the Reichsbank had monetary liabilities additional to its notes. But while the Bank of England held bank deposits like those of modern central banks, deposits at the Reichsbank were related to the development and operation of the only nationwide system for "noncash" payments-the giro system of transfer. Functionally, the giro system was a checking system, but one operated by the central bank instead of by individual private banks; a German bank would not clear giro transfers with other banks but would make and receive all transfers within Germany (with few exceptions such as an intracity one in Hamburg) through its Reichsbank giro account. The routing of the giro system was just the opposite of a checking system once the payer had started action; the payment went directly to the payee's bank instead of to the payee and then the payee's bank. However, the function was the same as that of an English or American checking system. Like the Bank of England, the Reichsbank was not subject to legal restrictions on reserves against its non-note liabilities; the gold-cover rules held only for notes. (The Reichsbank was also not split between Issue and Banking departments.) 3. Both central banks were subject to disclosure rules. The Reichsbank had to publish a weekly balance sheet as well as annual reports. In addition, the Reichsbank had to disclose weekly the amount of uncovered notes outstanding subject to the tax.! 4. With respect to instruments of policy, the two central banks were broadly similar. Like the Bank of England, the Reichsbank used as its primary policy instrument the discount rate but, unlike the Bank of England, used it nearly exclusively. (It is commonplace in the literature that the Bank of England frequently resorted to credit rationing as a supplementary instrument.) Both central banks were obligated to buy and sell gold in unlimited quantities at fixed prices; but in fact, both banks

314

Paul McGouldrick

used gold devices to stack the deck (usually in favor of gold sellers). Yet, as will be shown below, it is likely that the Reichsbank was less apt to transgress in this respect than did the Bank of England. 5. De jure, the Bank of England was a private body, while the Reichsbank was a mixed public-private enterprise (private stockholders contributed the capital and received dividends, but the federal government chose the management, had all voting seats on the board of directors, and made all members of management and operating personnel federal civil servants, as well as receiving 40 percent of profits in excess of 4.5 percent of share capital). De facto, both were mixed enterprises, and the Reichsbank had a feature emphasizing profitability incentives for top- and middle-level managers (part of their salaries was made proportionate to profits). That feature was inserted in the enabling bill by Parliament so as to discourage bureaucratic behavior (Flink 1930, pp. 11-15). 6. Like the Bank of England, the primary constituent of the Reichsbank portfolio of earning assets was bills of exchange. The Reichsbank also made collateral (Lombard) loans and purchased and sold securities, but the latter item was de-emphasized (weekly balance sheets show only a small and irregular item comprising securities, in which swings were tiny compared to those in bills). The Reichsbank did assist federalgovernment security placements, but again this was a minor item. Lombard loans were of somewhat greater importance but much smaller than bills in levels and changes, and the Lombard lending rate was always pegged a a hundred basis points above the bill rate. Thus, open-market policy was not important to any extent in Germany, whatever its role was in Great Britain. 7. Both central banks acquired almost a monopoly of note issues, the exceptions being issues of Scottish banks in England and issues of a very few banks in Germany. (During the 1870s, many issuing banks gave up their note issues; by 1914 the four survivors were South German and Saxon state banks with very small and stable issues.) In Germany, base (high-powered) money consisted of gold coin and bullion, a small and fixed issue of Treasury fiduciary notes (Reichskassenscheine), Reichsbank money liabilities (notes and giro deposits), the small and dwindling stock of note issues by other banks, and subsidiary coinage. As in England, central-bank money was designed for business and larger transactions; until 1906, the minimum note denomination was a hundred marks, and even afterwards twenty- and fifty-mark notes did not pass into wide circulation. Giro accounts were also only for banks and large nonbank enterprises; small transactors could make remittances by direct payment of single amounts at very low fees; and the number of accounts did not ever exceed three thousand. Gold remained competitive with Reichsbank money throughout; its proportion to Reichsbank notes fell

315

Operations of the German Central Bank, 1879-1913

secularly until the 1890s but then rose secularly until, by 1913, it was almost back to the 1880 ratio (Richard Tilly, an economic historian, quoted in Borchardt 1975, p. 27). In short, the formal differences between the Bank of England and the Reichsbank appear to be minor, with the exception of the tax on noteexcess issues. But as shown in a later section, certain institutional and procedural features did impress a unique stamp on Reichsbank targets, instruments, and operating policies. 7.2

Rules of the Game and Reichsbank Behavior over Business Cycles

According to a careful and influential student of the pre-1914 gold standard, central banks "were supposed to reinforce the effects of these [gold] flows on commercial bank reserves, not merely not to neutralize them." More concretely, "a discount rate and credit policy geared primarily to movements in central bank reserves was supposed . . . to have the effect of increasing central bank holdings of income-earnings assets when holdings of external reserves rose, and of reducing domestic assets when reserves fell" (Bloomfield 1959, p. 47). A second rule was that a central bank should raise the discount rate when losing gold and do the opposite when gaining gold (pp. 27-32). For Germany, Bloomfield found that the Reichsbank nearly always obeyed the second rule (though his measure of reserve tightness or ease was not gold but the ratio of reserves to money liabilities, introducing some uncertainty as to interpretation). Using annual data, he found that the Reichsbank violated the first rule more often than not. In the majority of years between 1880 and 1913, the Reichsbank raised income-earning assets (the portfolio) when reserves were falling and lowered assets when reserves were rising. My analysis, using quarterly, seasonally adjusted averages of weekly Reichsbank data2 over the same time span, strongly confirms Bloomfield's finding for Reichsbank gold and portfolio policy over businesscycle phases as well. As shown by tables 7.1 and 7.2, the portfolio expanded more algebraically than did reserves in every business upswing, and in three of the six, the portfolio grew even as reserves declined absolutely. During all 1879-1913 business contractions, the portfolio rose less algebraically than did specie reserves, and in three of the five contractions, it declined as reserves increased? Averaging overall upswings and downswings separately yields the following matrix of annual percent changes during cycle phases (see tables 7.1 and 7.2). Reserves Portfolio

Upswing -1.7 6.8

Downswing 12.4 - 0.6

316

Paul McGouldrick

Table 7.1

Percent Changes at Annual Rates in Total Portfolio of Reichsbank and Its BiU Component during Cyclical Upswings and Downswings, 1879-1913a Total Portfolio (1) Upswings Feb. 1879-Jan. 1882 Aug. 1886-Jan. 1890 Feb. 1895-~ar. 1900 ~ar. 1902-Aug. 1903 Feb. 1905-Jui. 1907 Dec. 1908-Apr. 1913

Bill Component (2)

4.5 6.2 11.0 0.6 12.8 5.6

5.8 12.2 10.8 2.9 15.0 9.9

~ean change Downswings Jan. 1882-Aug. 1886 Jan. 1890-Feb. 1895 ~ar. 1900-~ar. 1902 Aug. 1903-Feb. 1905 Jui. 1907-Dec. 1908

6.8

9.4

4.8 -0.6 -1.2 2.3 -9.4

1.0 0 -2.0 -3.6 -18.6

~ean

-0.8

-4.6

change

Sources: Burns and ~itchell 1946, p. 79 for turning-points. See also note 3 of text. Reichsbank weekly balance sheets are available in Deutsche Reichsbank 1901,1925; U.S. National ~onetary Commission 1911c; and Reichsbank annual reports. aFrom quarterly averages of monthly seasonally adjusted averages of weekly total portfolio and bill (Wechsel) component of portfolio levels.

Moreover, there is evidence that it was the intention of Reichsbank policy to raise the portfolio when reserves declined, rather than market forces overpowering weak Reichsbank attempts to follow the Bloomfield-hypothesized rule. First, the Reichsbank had more control over the bill portfolio than it had over Lombard loans, securities, and other assets. As noted earlier, the Lombard loan rate was always pegged at a fixed spread over the bill discount rate, enabling supply-and-demand forces peculiar to this market to counteract changes in discount rates. Operations to assist Reich government financing plus the absence of open-market policies of modern types distorted changes in the securities component of the portfolio. By contrast, the Reichsbank had complete freedom to set the discount rate for bills and viewed the bill portfolio as the instrument of bank policy. Hence, the best index of what the Reichsbank intended to do was the bill portfolio, not the total portfolio. Column (2) of table 7.1 shows much larger procyclical changes in the former than the latter.

317

Operations of the German Central Bank, 1879-1913

Table 7.2

Percent Changes in Reichsbank Specie during Cyclical Upswings and Downswings, 1879-1913 (quarterly seasonally adjusted data) Percent Change

Upswings Feb. 1879-Jan. 1882 Aug. 1886-Jan. 1890 Feb. 1895-~ar. 1900 ~ar. 1902-Aug. 1903 Feb. 1905-JuI. 1907 Dec. 190B-Apr. 1913 ~ean change Downswings Jan. 1882-Aug. 1886 Jan. 189~Feb. 1895 ~ar. 1900-~ar. 1902 Aug. 1903-Feb. 1905 JuI. 1907-Dec. 1908

~ean

change

Total (1)

At Annual Rates (2)

Beginning Level (million Rm) (3)

4.13 12.42 -24.32 -6.60 -17.64 8.89

1.3 3.5 -4.9 -4.4 -7.1 1.7

514.3 710.4 1,052.3 999.3 1,050.6 1,158.0

-3.85

-1.7

32.65 31.78 25.46 12.57 33.84

7.3 6.4 12.7 8.4 27.1

27.26

12.4

535.5 798.6 796.4 933.2 865.2

Sources: See table 7.1. Notes: As in all other tables using quarterly data for months when business cycles peaked or troughed, a small error arises from using calendar, not centered, three-month periods. The latter could be constructed from the author's worksheets and printouts.

Even more persuasive is table 7.3 showing the change in the spread between the Berlin open-market rate for high-grade bills and the Reichsbank discount rate over business upswings and downswings. In theory, the Reichsbank could have attempted to restrain growth in the bill portfolio during business upswings by raising the discount rate more sharply than the open-market rate. If bill supply by borrowers rose vigorously enough, the bill portfolio would still have moved procyclically, but the Reichsbank could be credited with the intent to restrain. But as table 7.3 shows, the pattern was just the opposite. When open-market rates rose during business upswings, so did the discount rate, but by less than the rise in the former. Consequently the spread between the higher discount rate and the lower open-market rate narrowed, encouraging borrowers or bill holders to switch sales to the Reichsbank. In recessions Reichsbank behavior was symmetrical; it encouraged the decline (or below-normal rise) in its portfolio by reducing its discount rate by less than open-market rates fell. And this "perverse" behavior was remark-

318

Paul MeGouldrick Cyclical Changes in the Spread between the Reichsbank Discount Rate and the Berlin Open-Market Rate, 1879-1913 (in basis points)

Table 7.3

Beginning Spread Upswings Feb. 1879-Jan 1882 J\ug. 1886-Jan. 1890 Feb. 1895-~ar. 1900 ~ar. 1902-J\ug. 1903 Feb. 1905-Jui. 1907 Dec. 1908-J\pr. 1913 ~eans

Downswings Jan. 1882-J\ug. 1886 Jan. 189O-Feb. 1895 ~ar. 190o-~ar. 1902 J\ug. 1903-Feb. 1905 Jui. 1907-Dec. 1908 ~eans

Ending Spread

Change

120 115 135 123 110 125

70 70 83 75 90 78

-50 -45 -52 -38 -20 -47

121

78

-43

70 70 83 75 90

115 135 123 110 125

45 65 40 35 35

78

122

42

~orgenstern 1959, chart 50 and are therefore approximations. The positive spread is due to the fact that the Berlin open-market rate was for the highest-quality paper while the Reichsbank rate was for all commercial paper satisfying minimum requirements and Reichsbank staff criteria.

Note: Levels in above table were read by eye from

ably consistent; there was never an upswing in which the spread rose nor a recession when it declined. Thus the Reichsbank's obedience to the Bloomfield discount-rate rule was only formal; in reality, it contravened that rule by inadequate discount-rate adjustments. So far, Bloomfield's tests have been confirmed and his conclusions strengthened, with the addition of evidence that it was conscious Reichsbank policy to achieve that result rather than an ineffectual effort on the bank's part (or that the bank was overwhelmed by events). However, the Bloomfield tests are seriously inadequate with respect to both the Reichsbank's own interpretation of the rules and the validity of the theoretical rules as optimal or descriptive of actual German central-bank policies during the classical gold standard period. The supposed rule of a fixed link between central-bank reserves and the portfolio would have been consistent with the· second rule-a central bank should raise (lower) the discount rate when gold is leaving (entering) the country-only if excess demand for base money by banks and the public also had produced a balance-of-payments deficit and excess supply of base money, a balanceof-payments surplus. In the excess-demand case, gold would have moved from the central bank into both domest~c and foreign hands, making a

319

Operations of the German Central Bank, 1879-1913

contraction of the portfolio necessary for both domestic and balance-ofpayments stability. But if excess demand for base money domestically were accompanied by a balance-of-payments surplus, one rule conflicted with the other by necessity. Raising the discount rate to eliminate domestic excess demand for base money would have aggravated the externalpayments surplus and hence generated an inflow of gold. In turn, the inflow would have counteracted discount policy. According to the monetary approach to the balance of international payments, domestic excess demand for money is consistent with balance-of-payments surplusespresumably the normal case under classical gold standard conditions. The results of table 7.4 should therefore not be surprising. Although the Reichsbank portfolio moved procyclically, its total money liabilities moved countercyclically, averaging a greater percentage rise during recessions than business upswings. In two out of six upswings, Reichsbank money actually declined, and the largest increase was only 6.9 percent at an annual rate. Moreover, even this and another still-moderate rate of increase (5.7 percent) occurred at the beginning of the 1880-1913 period when lack of experience might explain them. As experience was gained and the Berlin money market became more closely connected with money markets of other leading centers, Reichsbank money during cyclical upswings either declined or rose at below-trend rates. And in every recession, Reichsbank money rose and by not inconsiderable annual rates; the lowest was 3.9 percent, the highest, 8.0 percent. Intentionally or not, the Reichsbank thus reached a goal that has eluded modern central banks at different periods, namely, avoiding procyclical movements in its money liabilities. Unlike William McChesney Martin or Arthur Burns, the Reichsbank succeeded in leaning against the wind! Since this conclusion is important in what follows, I tested to be sure that it is a fact and not a statistical artifact due to data imperfections. Like the data underlying tables 7.1 and 7.2, the Reichsbank money levels from which table 7.4's percent changes were calculated are quarterly averages of monthly balance-sheet items that I seasonally adjusted. Two possible sources of bias can be eliminated. The seasonally unadjusted quarterly and monthly averages (in unpublished tables) also bear out the conclusions reached above. Panel 2 of table 7.4 furthermore shows that over the whole 1879-1913 period, the conclusions are not changed, with an extended-base adjustment of Reichsbank money that eliminates the effect of other note banks giving up note issues that the Reichsbank assumed.4 During the classical gold standard era, base money consisted of specie outside the central bank as well as central-bank money. The moderate countercyclical movement of Reichsbank money was due to losses of reserves outweighting portfolio expansions during business upswings and gains in reserves outweighing portfolio contractions during recessions. But if drains and refluxes of gold were internal ones, total base money

320

Paul McGouldrick

Table 7.4

Changes in Monetary Liabilities of Reichsbank during Cyclical Upswings and Downswings, 1879-1913 (percentages based on quarterly averages)8 Percent change

1. Actual monetary liabilities Upswings Feb. 1879-Jan. 1882 Aug. 1886-Jan. 1890 Feb. 1895-~ar. 1900 ~ar. 1902-Aug. 1903 Feb. 1905-Jul. 1907 Dec. 1908-Apr. 1913 ~ean

Total (2)

At Annual Rate (3)

794.0 1,095.2 1,621.8 1,813.4 1,902.9 2,266.3

16.73 23.63 -2.27 -1.91 8.25 10.62

5.7 6.9 -0.5 -1.4 3.4 2.5

change

Downswings Jan. 1882-Aug. 1886 Jan. 1890-Feb. 1895 ~ar. 1900-~ar. 1902 Aug. 1903-Feb. 1905 Jui. 1907-Dec. 1908 ~ean

Beginning Level (million Rm) (1)

change

2.8

926.7 1,356.0 1,585.0 1,778.7 2,059.9

18.2 19.6 14.4 7.0 10.0

4.0 3.9 7.2 4.7 8.0 5.6

held by banks and the nonbank public could still have moved procyclically and thus contributed to business cycles. Table 7.5, which estimates specie held in Germany outside the Reichsbank, rejects this possibility.5 Such specie did move procyclically, although mildly so, averaging a 6.7 percent annual rate of increase during expansions and 5.2 percent during recessions. But specie plus Reichsbank money (column 4) moved countercyclically because the stronger movements in Reichsbank money dominated the weaker movements in specie. Unfortunately, reliable estimates of monthly net specie inflows into Germany (U. S. National Monetary Commission 1911c) are available only for cycle phases between 1895 and 1908, but inferior data (Hoffman 1965, table 241) confirm this finding for specie (a mild procyclical movement swamped by the Reichsbank money swings) for the 1879-95 and 1908-13 business-cycle phases. (The largest expansion of specie held by the public in 1879-95 and 1909-13 periods-an estimated 3.9 percent at annual rate from December 1908 to April 1913-was smaller than the 4.0 percent expansion during the February 1905-July 1907 upswing and the 6.7 percent expansion during the 1903-5 recession). Thus, the conclu-

321

Operations of the German Central Bank, 1879-1913

Table 7.4 (cont.) Percent change Beginning Level (million Rm) (1)

Total (2)

At Annual Rate (3)

840.0 1,148.2 1,660.8 1,835.4 1,903.9 2,266.3

16.16 21.67 -1.80 -2.43 8.35 10.63

5.5 6.3 -0.4 -1.7 3.5 2.5

2. Extended base concept b Upswings Feb. 1879-Jan. 1882 Aug. 1886-Jan. 1890 Feb. 1895-~ar. 1900 ~ar. 1902-Aug. 1903 Feb. 1905-Jul. 1907 Dec. 1908-Apr. 1913 ~ean

change

Downswings Jan. 1882-Aug. 1886 Jan. 189Q-Feb. 1895 ~ar. 190Q-~ar. 1902 Aug. 1903-Feb. 1905 Jui. 1907-Dec. 1908 ~ean

change

2.6

975.7 1,397.0 1,631.0 1,790.7 2,062.9

17.67 18.89 12.53 6.32 9.86

3.9 3.7 6.3 4.2 7.9 5.2

Sources: See table 7.1. 3Beginning and ending quarters are those of the month in which the cycle phase began and ended, respectively. Quarterly averages of monthly seasonally adjusted items are taken to reduce irregular variation. However, unpublished percentage changes based on month-only beginning and ending data do not show appreciably different results; in particular, rank orders are never affected. b~oney liabilities are calculated as if the Reichsbank had been the sole issuer of notes throughout, except for notes of banks still issuing them on December 1910. Thus, this panel eliminates "noise" due to banks giving up note issues (and the Reichsbank assuming them) during the entire 1879-1913 period. See note 4 of text.

sions are not upset by rough estimates for earlier and later periods, although caution is indicated. Accordingly, base money held by banks and the public (in contrast to their specie holdings) grew at a remarkably stable pace over German business cycles, and what deviations there were from absolute stability were in a countercyclical direction.6 What the Reichsbank actually did was to substitute internal for external drains and refluxes of specie over phases of German business cycles, by techniques explained later. Its success is indicated by the gold imports net of gold exports for the business-cycle phases starting in January 1890 and ending in December 1908, shown in table 7.6. In the 1890s imports were virtually the same

322 Table 7.5

Paul McGouldrick German High-Powered-Money Approximation, January 1890-December 1908: Levels and Percent Changes during Cyclical Upswings and Downswings (million Rm) Specie Outside Reichsbank a

Upswings Feb. 1895-Mar. 1900 Mar. 1902-Aug. 1903 Feb. 1905-Jul. 1907

(1)

Total HP moneya (2)

1,699.4 2,427.9 2,966.5

3,303.9 4,276.2 4,870.8

Mean percent change Downswings Jan. 1890-Feb. 1895 Mar. 1900-Mar. 1902 Aug. 1903-Feb. 1905 Jul. 1903-Dec. 1908 (Memo: Dec. 1908 level) Mean percent change

1,603.0 2,234.2 2,579.3 3,655.0 3,805.6

2,954.6 3,863.2 4,355.9 5,707.9 6,085.0

Percent Changes at Annual Rate Outside Specie (3)

HP Money (4)

6.2 4.4 9.6

3.4 1.2 6.9

6.7

3.8

1.2 4.3 10.0 3.3

2.4 5.3 6.9 5.3

5.2

5.0

Sources: High-powered-money (HP) approximation equals monetary liabilities of the Reichsbank plus estimated monetary specie in Germany outside the Reichsbank. Other components, such as Treasury notes (Reichskassenscheine) and notes of state banks, were very stable and are excluded. Specie outside the Reichsbank was estimated by cumulating on Hoffman (1959, table 240 Metal/geld benchmark for end of 1889) annual changes for 1890 and 1891, and specie-imports net of exports, by month, in U.S. National Monetary Commission 1911c, less monthly changes in specie at the Reichsbank. Notes: Both the net specie imports and Reichsbank levels from which changes are derived are not seasonally adjusted. No adjustment is made for industrial uses of specie, so a cumulative upward bias exists in the trend of the series. aBeginning of cycle phase.

regardless of cycle phase, and from 1900 through 1908 the very mild fluctuations at annual rates were countercyclical, not procyclical. The flows are consistent with the hypothesis that the Reichsbank operated successfully to stabilize gold inflows and thereby stabilized the monetary base. Indeed, an upward trend and noncyclical swings dominate the table 7.6 series. So far, the discussion has been exclusively cast in terms of base money, its components, and Reichsbank portfolio changes over cycle phases from troughs to peaks and peaks to troughs. What about shorter periods? As tables 7.4 and 7.7 show, the remarkable stability in cycle-phase percent changes at annual rates in Reichsbank money contrasts sharply

323

Operations of the German Central Bank, 1879-1913

Table 7.6

Net Gold Imports into Germany during Cyclical Upswings and Downswings, 1890-1908 Total Net Inflow (million Rm rounded)

Upswings Feb. 1895-Mar. 1900 Mar. 1902-}\ug. 1903 Feb. 1905-Jul. 1907 Downswings Jan. 189{}-Feb. 1895 Mar. 190{}-Mar. 1902 Aug. 1903-Feb. 1905 Jul. 1907-Dec. 1908

(1)

Net Inflow (at annual rate) (2)

254 68 462

50 48 191

259 408 541 314

51 204 361 222

Sources: For 1892-1907, U.S. National Monetary Commission 1911c, pp. 247-50. Estimated for 189{}-95 and 1907-8 phases from Hoffmann 1965, table 240, with arithmetic interpolation from previous calendar year for January 1890 and February 1895.

with high quarter-to-quarter instability. In 140 quarters between January-March 1879 and April-June 1913, 48 had plus or minus changes greater than 10 percent; and the absence of any decline whatsoever during cycle phases must be juxtaposed against 49 quarter-to-quarter declines. But these quarterly changes were also unsystematic, having the appearance of a random walk. Table 7.8 shows regressions of current on lagged percent changes in Reichsbank money; none of the regression coefficients for lagged changes are statistically significant at the 5 percent level and the coefficient signs are generally negative, not positive. The Durbin-Watson h ratios for the three equations also permit us to infer randomness of short-period changes in the equation residuals. This inference does not imply that money demand was not stable; as I show below, it was. But that stable demand explains only a minor portion of quarterly changes in Reichsbank money. The systematic, mildly countercyclical movements in Reichsbank money over cycle phases, in contrast to a picture of a random walk superimposed on an upward trend, are explained theoretically in a subsequent section of the paper. Intermediate-term instability or stability in Reichsbank money can be assessed by dividing the cycle phase into two halves. Changes over half-phases indicate instability during upswings; in all of them, the larger change was more than twice the size of the smaller change without regard to sign, and the sign changed during four out of six upswings (McGouldrick 1982, table 7). The changes also indicate little or no intention or success, on the part of the Reichsbank, in raising its money liabilities by

0 0 0

Less than -30 (1)

1 0 1

-30 to -20.1 (2) 6 6 12

-20 to -10.1 (3) 7 4 11

-10 to -5.1 (4) 10 15 25

-5 to -0.1 (5) 16 15 31

0 to 4.9 (6) 14 11 25

5 to 9.9 (7) 13 14 27

10 to 19.9 (8)

4 2 6

20 to 29.9 (9)

1 1 2

30 to 39.9 (10)

Size Distribution of Quarterly Changes in Monetary Liabilities of the Reichsbank (percent changes at annual rate)

Sources: See table 7.1. Note: As in other tables, monetary liabilities are quarterly averages of seasonally adjusted monthly Reichsbank-balance-sheet items.

---

1879-1896 1897-1913 Whole period

Table 7.7

72 68 140

Total Quarters (11)

325

Operations of the German Central Bank, 1879-1913

Table 7.8

Eq. No. a

The Relationship between Current and Lagged Values of Reichsbank Money Percent Changes, II 1892-IV 1907

Constant

(2)

.60b (2.10) .52b

(3)

(1.76) .53 b

(1)

Lagged One Quarter - .154 (1.23) -.134 (1.06)

(1.74)

Lagged Two Quarters

Average Change from First to Third Lagged Quarters

Durbin h Ratio -1.28

.133 (1.05) -.019 (0.80)

-1.14

Note: Absolute t-values are in parentheses. aRegressions are ordinary least squares. bSignificant at the 5-percent level. C Applies only to the first independent variable.

more near the trough than near the peak of the business cycle, an obvious goal of modern stabilization policies of a nonmonetarist type. In four out of six expansions, the change was algebraically greater during the second than the first half. The modern stabilization goal was more nearly approximated during downswings (gunning money liabilities as the economy slid further into recession); the percent change was greater in four out of six second halves of contractions. Changes in both halves of recession phases were all positive. It may be that the strong upward trend in German output made the Reichsbank more sensitive to unduly large expansions. However, the basis for such a conjecture is slight since the proportion of higher second-half-recession percentage changes to the total of percentage changes is not particularly large. The bank's portfolio was a better indicator of Reichsbank policy than was Reichsbank money. The bank could affect the portfolio directly by discount-rate policy while its gold stock (and thus money liabilities) was subject to undesired changes attributable to the public's demand for gold and international specie flows. However, half-phase movements of the portfolio show just about the same degree of instability as do the money liability half-phase results-algebraically greater percent movements during the second than the first half in four out of six business expansions, and greater percent movements in the second half of three out of five contractions. Neither Reichsbank money nor its portfolio thus offers any evidence that the bank (or market forces) was systematically providing more resistance to deviations from long-term trends as the German economy moved mo"redecisively to peaks or troughs.

326

Paul McGouldrick

In sum, Reichsbank money moved countercyclically over whole business-cycle phases (trough to peak and peak to trough), measured by average phase percent changes at annual rates. If phase changes are compared with those in adjacent phases, one exception to countercylical behavior is observed: the 1886-90 expansion rate was higher than for the preceding or succeeding contraction (table 7.4, col. 3). The higher rate can be explained by the very rapid development of the Reichsbank's giro-account business during these years, a Schumpeterian innovation that induced wealth holders to substitute Reichsbank money for other assets. Countercyclical swings in money occurred despite procyclical swings in the portfolio because changes in gold reserves more than offset portfolio changes. Gold flowed out during upswings and returned during downswings, but the flows were internal, not external. Gold movements into Germany show an upward trend but no cyclical pattern. Thus, the ex post data tend to show that Reichsbank policy was directed towards stabilizing gold flows rather than portfolio growth (and certainly not linking portfolio changes to gold-reserve changes, according to the Bloomfield interpretation of the rules of the game). Moreover, the "perverse" behavior of the portfolio was due to Reichsbank action rather than market forces, as is shown by the invariant practice of the Reichsbank in raising its average lending rate by less than the Berlin openmarket rate rose during business expansions and cutting its average lending rate by less than declines in the open-market rate during recessions. Within cycle phases, money liabilities show randomlike movements on a quarterly seasonally adjusted basis and highly irregular movements as between first and second halves of cyclical contractions and expansions. Before offering a theoretical explanation for these patterns, I first turn to the question of the relative stability of the growth of the German economy from 1879 to 1913. 7.3

The Relative Stability of German Growth, 1879-1913

First of all, the imperial German economy grew vigorously. The growth rate of real GNP from 1881 to 1913 was topped only by that of the United States, and German GNP growth exceeded that of France and Great Britain by a large margin. (Russia probably did better from 1900 to 1913 but not nearly as well from 1881 to 1913.) The principal engine of growth was vigorous industrialization, concentrated especially in output of investment and production goods. As a consequence, investment in fixed capital and residential housing was high relative to net social product, ranging from 8 to 17 percent and usually in excess of 12 percent. Germany was blessed with relatively stable growth. We can compare the stability of German growth with that of other leading countries,

327

Operations of the German Central Bank, 1879-1913

specifically that of the United States and Great Britain. (We rule out France from the comparison because its economic growth was slow and its industrial sector was small.) One measure of instability is fluctuations in the average annual growth rate of real GNP. German real growth proceeded at a much more even rate after than before 1879-1 exclude the 1870s because of the distorting factor of the receipt of the 6-billion-franc French indemnity after the Franco-Prussian War (Borchardt 1976, chart of Hoffmann real-netsocial-product estimates, p. 21). Measured by standard deviations of annual percent changes, the same holds true of intercountry comparisons. German instability in 1880-1913 was considerably lower than that of the United States (though greater than that of the British), and even a little lower than the instability in Germany in 1951-68, the golden age of the economic miracle (table 7.9). Moreover, the impression of stability is further heightened by examination of fluctuations in industrial production, in respect of which German relative stability is even more striking. Imperial Germany ranks at the bottom of the instability league before 1914 and below economicmiracle Germany and the United States in 1951-68 (table 7.10). The instability of the United States in the gold standard era stands out. Even when measured for the 1897-1913 period, to exclude the pre-1897 alleged

Table 7.9

Standard Deviations of Annual Percent Changes in Real GNP, 1880-1913 and 1951~8

Germany Great Britain United States

1880-1913

1951-68

3.1 1.8 5.2

3.5 1.4 2.7

Sources: For 1880-1913 from Hoffmann 1965; Deane 1968; Gallman (unpublished U.S. estimates); for 1951-68, United Nations Yearbook.

Table 7.10

Standard Deviations of Annual Percent Changes in Indexes of Industrial Production, 1876-1913 and 195~8

Germany Great Britain United Statesa

1876-1913

1950-68

3.3 4.4 11.1

6.6 2.8 5.7

Sources: For 1876-1913, Hoffmann 1965, tables 10, 13; Mitchell and Deane 1962, pp. 271-72; U.S. Bureau of the Census 1960, series P-13. For 1950-68, United Nations Yearbook and Survey of Current Business. aOnly for 1897-1913, to exclude period of monetary instability before 1897.

328

Paul McGouldrick

destabilizing effects of free-silver agitation, instability of the United States is more than three times as high as that of Germany. However, the case for relative German stability is not established by these statistical comparisons. Differences among countries in data availability and measurement techniques in the construction of industrial production and GNP cast doubt on measures of instability. The Hoffmann index of industrial production, for example, is based in part on industrial employment, which is more stable cyclically than the outputs of sensitive materials that figure largely in the Frickey index (see table 7.10). Additional evidence, however, may be consulted. The high stability of German base-money growth (indeed, its mild countercyclical pattern) implies a high stability of growth of German broad money (defined as base money plus all deposits at German commercial banks exclusive of savings banks). Unpublished estimates of broad money that I constructed do not disappoint this expectation. Table 7.11 shows no single year in the 1878-1913 period in which broad money declined and only four out of thirty-six years in that period in which broad money rose by less than 2 or more than 10 percent. Such a moderate stability of broad-money growth implies a low incidence of bank and financial intermediary failures, and in this respect the German record was outstanding. A few private, unincorporated banks failed in 1891, but 1901, when two banks and two insurance companies went under, was the single year of failure of large institutions. To find whether there were any ripple effects familiar in American economic history, I read weekly issues of the German counterpart of the London Economist (der Deutsche Oekonomist) from the crisis weeks to six months afterwards. The weekly periodical simply reported the crisis and was thereafter silent. It made no mention of stock-market turmoil, any policy change by the Reichsbank (the discount rate was not raised), any drains of reserves from individual banks or groups of banks, or any Frequency Distribution of Annual Percent Changes in Broad a ~oney, 1878-1913

Table 7.11

Growth Rates

less

0.1 to 2.0

2.1 to 4.0

4.1 to 6.0

6.1 to 8.0

8.1 to 10.0

0

3

7

12

6

7

o or Years

Over 10.0

Sources: See table 7.1 for Reichsbank money; Hoffmann 1965 for specie and credit bank deposits; Kaiserliches Statistisches Amt. 1879-1914 for other data. aBroad money equals base money plus all deposits (including current-account ones paying interest) at German credit banks less interbank deposits. Not included are deposits at savings banks (Sparkassen) and at financial intermediaries other than credit banks.

329

Operations of the German Central Bank, 1879-1913

sustained reaction of open-market interest rates (the Berlin rate jumped briefly but soon returned to 2.5 percent, low even by pre-1914 standards). Long-term interest rates on German imperial bonds were also stable or declining, week-by-week and month-by-month. Still another quantitative indicator, which is also statistically independent of real GNP and output time series, is the rate of unemployment. That reported by the German trade unions for all members was low and fluctuated only between 1.5 and 3 percent (Borchardt 1976, p. 31). Since union membership was more concentrated in cyclically sensitive than insensitive industries, this percentage testifies to unemployment stability greater than in Great Britain (where trade-union reports showed higher levels and greater changes) and certainly greater than in the United States (Lebergott 1964). In addition, nonquantitative evidence is available. The lack of discussion of business-cycle problems by contemporary German economists, social scientists, intellectuals, and politicians was certainly not due to a lack of awareness of social and economic problems. What is suggested, however, is that the economy was stable. This was the age when the Social Democratic Party became the largest party in the German parliament and when the professional organization of German economists bore the title the Association for Social Policy (Verein der Sozialpolitik). Complaints and investigations abounded, according to historians such as Holborn (1969, chaps. 6-8) and Mann (1968, parts 7-8). Welfare and social insurance, high tariffs on foodstuffs, the plight of small farmers, structural problems in heavy industry and cartelization, low wages and the overweening power of capital, the alleged power of the Jews or of East Elbian junkers, militarism and heavy military spending, even the subjugation of tavern keepers to the brewers-these themes resound through the writings of moderates like Gustav Schmoller and Max Weber as well as radicals of the right or the left. But there is one theme that is an exception: business instability of a macroeconomic type is scarcely mentioned. Scholars and publicists are scarcely apt to neglect a serious problem if one is around. Contrast German neglect with the concern in the United States, where pre-1914 cyclical instability led to the foundation of the National Bureau of Economic Research. And finally note the general satisfaction with the gold standard in imperial Germany. The sole interest group that made an issue of the gold standard prior to the turn of the twentieth century was that of estate owners dissatisfied with falling prices of wheat and rye, who agitated for a bimetallic standard. Tariff protection and the upturn in world prices after 1896 led to a discontinuance of the estate owners' monetary agitation. The Social Democrats defended the gold standard staunchly throughout; according to an article in Vorwaerts (9 July 1896), "the German working class and its representatives,

330

Paul MeGouldrick

the Social Democrats, will always be found in favor of the gold standard during our monetary controversies which are breaking out" (Borchardt 1976, p. 39, quoting Max Schippel). 7.4 An Interpretation of the Reichsbank's Influence on the Money Stock It is impossible to avoid the conclusion that there was a link between the Reichsbank's operations and the relative stability of German-output growth prior to 1914. Taken by itself, the mildly countercyclical behavior of high-powered money was scarcely decisive; Cagan (1965, charts 6-7, pp. 103-4) found that in the much-less-stable American economy, prior to the founding of the Federal Reserve system in 1914, high-powered money did not in general move procyclically. While the German countercyclical pattern was more uniform than that for the United States, the difference was not huge. This section maintains that the Reichsbank limited German cyclical fluctuations in two ways. First of all, the centralization of gold reserves in the Reichsbank increased the information available to banks and the public as to bank-reserve positions. In the United States, by contrast, individual bankers, businessmen, and farmers had only the vaguest knowledge of the aggregate liquidity or illiquidity of the banking system taken as a whole. This structural superiority could be found in any country with a central bank under pre-1914 gold standard rules. In addition the composition of German high-powered money mattered a great deal as opposed to its level. By systematically lowering the spread between the discount rate and open-market rates during business expansions and raising the spread during contractions, the Reichsbank achieved two objectives, one external and one internal. The external objective was to put downward (upward) pressure on short-term openmarket rates and thereby discourage (encourage) inflows of short-term capital during business upswings (downswings). The result contributed to stabilizing gold inflows so that they did not behave procyclically. The internal objective was to induce German banks not to vary their reservedeposit ratios in the procyclical manner Cagan found characterized U.S. banks. The technique was to make bills of exchange eligible for discounting such a close substitute for high-powered money that Reichsbank portfolio operations, which added to (reduced) high-powered money, simultaneously withdrew (added to) the highly liquid, eligible-bill substitute from bank and nonbank portfolios during business upswings (downswings). As a result, true bank liquidity did not vary procyclically or did so by little, inducing German credit and savings banks to maintain loans, investments, and deposits free of pronounced swings of a cyclical nature.

331

Operations of the German Central Bank, 1879-1913

Thus the composition and sources of high-powered money made a difference, as Tobin (1965, pp. 467,469; also Tobin and Brainard 1963, pp. 383-84, 398-400) argued? Additions of Reichsbank money liabilities had less of a stimulative effect on portfolio decisions by banks and nonbank wealthholders than did additions to the latter's holdings of gold. The external and internal policy objectives did not have to be pursued consciously but resulted automatically in pre-1914 Germany from the overriding target of Reichsbank policy-to keep the mark stable in gold value. Finally, the supply of bills from borrowers did not vary in such a way as to undercut the dampening effect on total liquid-asset holdings from exchange of one very liquid asset-Reichsbank money liabilitiesfor another-bills of exchange eligible for sale to the Reichsbank-in normal Reichsbank operations. What were the avowed targets of Reichsbank policy? One was structural-to make gold freely available within Germany by adherence to gold standard rules and therefore to make it a near-perfect substitute for Reichsbank money in bank and public holdings. Robert Franz put this strongly: "It has always been the Reichsbank's policy to satiate the channels of circulation with gold as much as possible, with the result that the per capita gold cir.culation in Germany is much larger than in any other country" (U.S. National Monetary Commission 1911a, p. 54; 1910a, p. 147). No obstacles were placed in the way of internal conversions of notes or deposits into gold, and seigniorage charges on gold coin were kept so low that jewellers in Pforzheim, Germany melted down coin to obtain gold instead of purchasing bullion (U.S. National Monetary Commission 1910c). This practice contrasts sharply with the lack of domestic convertibility when the gold standard was formally reestablished in Germany in 1925-26; the Reichsbank then limited convertibility to foreign transactions. In Britain, likewise, the minimum conversion under the post-1925 gold bullion standard was four hundred ounces of gold (Moggridge 1969, p. 60). For international transactions, the picture is a little different. The Reichsbank used gold devices in addition to discount policy; at times it granted interest-free advances for importation of bullion, paid out gold only in Berlin instead of in port cities such as Hamburg, and may have used moral suasion in 1907 to persuade banks to reduce demands for gold. But this was all; 1881 was the only time purchase and sale. prices were varied to encourage purchases and discourage sales. The effects of the interest-free advance and Berlin-only sale policies were minimal indeed. Even if one assumes a 7 percent interest rate and a generous twelve-day transit period for gold from London to Berlin, the implicit devaluation of the mark was only 21/100ths of 1 percent. Since Berlin was less than two hundred miles from Hamburg, the freight disadvantage of taking deliveries in Berlin was negligible.

332

Paul McGouldrick

How influential was moral suasion? A quantitative index might be the violations of gold export and import points in exchange rates; export points, in particular, were heavily bunched during business upswings, but the deviations were extremely small (Morgenstern 1959, tables 33 and 53). If we accept Morgenstern's estimate of median gold import and export points as 20.34 and 20.505 marks per pound sterling, the maximum deviation was two pfennings on the low side and three pfennings on the high side, except for November 1907 when the deviation reached five pfennings. Even the last case represented a violation equal to only 24/100ths of 1 percent-scarcely an incentive for a German to load his rucksack with coins and wander into Switzerland. If we take Morgenstern's estimate of maximum gold points (an estimate because shipping, insurance, and other costs of moving gold varied), the one breach of gold points during the whole 1879-1913 period occurred in November 1907. I have belabored the Reichsbank's near-purity in observing the goldavailability rule of the game under the pre-1914 gold standard in order to reject a competing hypothesis-direct management by gold devices-a thesis originating with Sayers (1936) and mentioned quite extensively in the historical literature on the pre-1914 era. Instead, my explanation of Reichsbank operations runs as follows: 1. Pre-1914 German business cycles originated in the real, not the monetary, sector of the German economy. An earlier section established that base money moved countercyclically. In addition, turning points of Reichsbank money do not precede cyclical peaks or troughs (unpublished charts). More often than not, the aggregate reserve-deposit ratio of German credit and savings banks moved countercyclically, not procyclically, quite unlike the case for the United States (see section 7.5 below). Thus the general picture is one of a "Keynesian," not monetarist, explanation of the mild German business cycles that occurred. Keynesian cycles, originating in fluctuations in business investment, are not a residual explanation (because autonomous monetary fluctuations are excluded as a cause). The German economy grew more rapidly than any other advanced economy except that ofthe United States from 1880 through 1913. Vigorous growth is frequently associated with high uncertainty about future rates of return on real assets. Even more decisively, German growth was peculiarly associated with an explosive development of the investment and basic-production-goods sectors (iron and steel, chemicals, electrical goods, machinery of all types), which required large capital investment per unit of output. Hence, cycles due to fluctuations in expected rates of return on durable capital are credible. 2. As argued in the preceding paragraphs, Keynesian cycles originated in fluctuations in business investment. They were not due to movements in the current account. The latter actually moved countercyclically, the positive balance rising in recessions and falling in business upswings

333

Operations of the German Central Bank, 1879-1913

(Hoffmann 1965, table 241 excluding gold movements). On the other hand, swings in business investment were large and procyclical (Hoffmann 1965, tables 248,249). Germany was always a net lender on total capital account, but its aggregate lending declined during business upswings and rose during contractions (thus offsetting swings in the current account). But-and this is crucial-the behavior of short-term capital was very different. The persistently higher open-market interest rates in Berlin and in London and Paris testify to long-term capital exports but short-term capital imports as the normal state of affairs for business upswing and downswing alike. 3. A Bloomfield theoretic rule of the game would have generated procyclical inflows and outflows of gold, because the Bloomfield-posited counterfactual type of central bank would not have attempted to influence domestic short-term interest rates. During a business upswing (contraction), the strong cycles in business investment would have caused the capital balance to rise (fall) by more than the current-account balance declined (rose), as German interest rates rose (fell) relative to those in London and Paris .. As a consequence, the balance of payments would have moved procyclically, causing gold inflows during upswings and outflows during recessions. But as table 7.6 shows, this movement did not occur. The Reichsbank exercised a dampening influence on German shortterm interest rates within German business cycles in order to avoid procyclical movements in"the mark exchange rate and in gold inflows and outflows. During business upswings, the bank raised its discount rate but by less than open-market rates were rising (table 7.3) and conversely lowered them by less than open-market rates during recessions. Structurally, the discount rate was always higher than the measured open-market rate in Berlin; the latter applied only to prime bills of exchange while the Reichsbank discount rate was for all classes of bills; hence the Reichsbank held lower-class bills (Whale 1968, p. 113). But this structural difference did not preclude a considerable elasticity of substitution between prime and other bills in asset-holder portfolios. Hence the failure of the Reichsbank to raise (lower) its discount rate by as much as prime rates rose (fell) indicates downward (upward) pressure on open-market interest rates during business upswings (downswings). As a result of this countercyclical pressure on internal interest rates, German net exports of securities did not rise (decline) enough in upswings (downswings) to do more than counteract the accompanying worsening (improvement) of the current-account balance. Hence gold did not flow in or out during upswings and downswings, respectively. Put another way, internal drains of gold from the Reichsbank reserves took the place of gold flows from abroad during German business upswings. (As table 7.5 shows, the stock of gold held in'Germany outside

334

Paul McGouldrick

the Reichsbank moved procyclically.) But were interest rates the target of monetary policy, if not the ratio of res~rves to monetary liabilities? No. Jacob Riesser, perhaps Germany's leading expert on money and banking, correctly described the target as follows in testifying before the Bank Inquiry Commission in 1908: The rate of private discount is made at least in a general way in conformity with the ratio of supply and demand existing in the market. On the other hand, the rate of bank discount, fixed by the Reichsbank with an eye to the regulation of credit transactions and the maintenance of the gold standard, depends in the foremost place on the favorable or unfavorable condition of the total balance of payments, whose primary expression is to be seen in low or high exchange rates. (U.S. National Monetary Commission, 1910c, pp. 305-6) The exchange-rate target was far better than an interest-rate target for stabilizing gold inflows (because of the high rate of economic growth and the continued preference of the public for the gold component of highpowered money, cycle phases always showed gold inflows, never outflows). The exchange rate was known immediately and summarized all influences on the external demand for and supply of German money. On the other hand, reliable interest-rate information for Germany and other countries embraced only bills and securities of the highest quality; if there is so much uncertainty today on international elasticities of demand and supply for bills and securities, how much greater was the lack of knowledge then! This policy undoubtedly stabilized gold inflows, but why did it also tend to stabilize the economy? The answer is twofold. First, the Reichsbank had a structural bill-purchase policy to make bills of exchange (strictly, bills eligible for discount-a very large proportion of all bills) such close substitutes for high-powered money that the procyclical swings observed in table 7.1 for the Reichsbank bill portfolio did not generate procyclical portfolio decisions by German banks and spending decisions by the nonbank public. Second, the demand for money by banks and the nonbank public was highly stable. The first is the major point, since stability of money demand did not matter so much for a single country under gold standard rules (an excess demand for high-powered money could be eliminated fairly promptly by an induced balance-of-payments surplus). What the Reichsbank did to make bills extremely close to money in liquidity characteristics was to express a willingness to discount all bills presented within cyclically invariant standards of eligibility. A Hamburg banker stated this clearly in testimony before the bank inquiry of 1908: The Bank of England reserves to itself, in fact, the right to discount or not to discount, a system not practiced with us by the Reichsbank. If

335

Operations of the German Central Bank, 1879-1913

the quality of the bill satisfies the Reichsbank, it takes any amount that is sent to it, at the official rate. This is a wonderful safety valve for traffic in general, not only for the banks but for all the patrons of the Reichsbank. (testimony of banker Max Schinckel, U.S. National Monetary Commission 1910c, pp. 377-78) Whale (1968, pp. 163-65 and 125-31), who also noted this policy, stressed that it involved accepting very large seasonal and irregular swings in Reichsbank credit and money, and that German-bank cash reserves were much smaller than English-bank reserves. Robert Franz (editor of the Deutsche Oekonomist, Berlin), referred to a qualitative impact of the Reichsbank policy on bank balance sheets; cash was highpowered money only and not interbank deposits (U.S. National Monetary Commission 1911a, p. 83). A German banker sgreed-banks only kept till money to satisfy daily transactions needs (U.S. National Monetary Commission 1910c, p. 186). Such extremely small bank reserves of base money were encouraged by another salient feature of Reichsbank policy: it did not adhere to the real-bills doctrine in the sense of attempting to screen applications for discount and rejecting bills not related to the "needs of trade." This notion deserves emphasis. In the Reichsbank's own first-quarter-century history only one paragraph espoused the needs of trade doctrine in rather perfunctory terms (U.S. National Monetary Commission 1910a, p. 78). Flink (1930) found no such screening prior to 1908. In practice, the credit banks generated large amounts of finance paper, supposedly the abomination of abominations for real-bills theorists, but the Reichsbank was ready to discount them. The Reichsbank's "most important task [was] to grant all credit required" (Flink 1930, pp. 25,27). A source quoted by Flink agreed: "In times of money scarcity, surprisingly large amounts of commercial paper have been thrown into the portfolio of the Reichsbank which the latter could not prevent" (Flink 1930, quoting von Lumm, p. 27n). Franz quantified the departure from real-bills theory; between 1905 and 1907, money and banking bills were between 49 and 53 percent of all Reichsbank bill assets (U.S. National Monetary Commission 1911a, pp. 63-64). Therefore, eligibility of bills for sale to the Reichsbank only concerned formal characteristics. Bills had to be private, nongovernment, have at least two signatures, and, we may infer from the Reichsbank firstquarter-century history, sellers had to have an ongoing business relationship. No one could w'lpder in off the street and sell a packet of bills. But that was all. The number of persons, firms, and banks discounting was over sixty thousand. After 1908, the Reichsbank made some attempt to screen bills; before then, bills were as good as cash to the German commercial (Kredit) bank system. One policy restriction held, however:

336

Paul McGouldrick

with almost no exceptions, the Reichsbank purchased only bills of very short maturity, forty days or less. As a result, eligible bills of less than forty-days maturity were virtually equivalent to base money when held by the German banking system. The Reichsbank was ready to purchase unlimited quantities from the banking system, and the de facto refusal to screen bills meant that there was little or no uncertainty that specific bills might be ineligible as long as the two-signature, less-than-thirty-to-forty-days effective-maturity characteristics were present. Bank holdings of bills were indeed subject to risk of decline in value (if the discount rate rose) or appreciation of value (if the rate declined). But the interest income from bills, combined with very short maturities, made eligible bills quite attractive vis-a-vis gold or Reichsbank money holdings. The consequence was that Reichsbank expansion during business upswings and contraction during recessions had only a small net effect on the true reserves of the banking system. During upswings, banks gained base-money proper from Reichsbank portfolio expansion but lost eligible bills; during recessions, banks lost base-money proper from the same policy but gained eligible bills as nonbank borrowers turned from the Reichsbank to the banking system. (Such nonbank borrowers did have access to the Reichsbank-see U.S. National Monetary Commission 1911a, p. 63.) As a result, base money inclusive of eligible bills did not change, so that bank lending and money creation were discouraged during business upswings and encouraged during recessions. However, cyclical stability of gold, Reichsbank money, and eligible bill holdings by banks and the public required one other condition as well. The supply of bills by borrowers had to be reasonably stable as well. Table 7.12 is a test of that requirement. The table shows annual average percent changes over business-cycle phases of estimates of thirty-day-maturity bills held by banks and the nonbank public added to high-powered-money proper for the years 1879 to 1900, compared to the annual average percent changes of high-powered money. As the reader can see, the rates of change of the two concepts of high-powered money differed si~nificant~y durin~ only one out of five cycle phases. Evidently, and with the one exception noted, Reichsbank operations to dampen cyclical fluctuations in short-term interest rates did not elicit perverse procyclical movements in bills supplied and not taken up by the Reichsbank. Such movements would have destroyed or seriously weakened the stabilizing effects of Reichsbank policy. A fairly high degree of stability of demand for Reichsbank money liabilities by banks and the public is also indicated by the results of multiple regressions shown in table 7.13. The dependent variable in both equations is changes in Reichsbank money. The functions are based on a Koyck distributed-lag theory of adjustment of actual to desired Reichs-

337

Operations of the German Central Bank, 1879-1913

Table 7.12

Level of Extended High-Powered Money at Trough or Peak Years, and Annual Average Percentage Changes between Turning Points Compared to Those of High-Powered Money Proper, 1879-1900 Annual Average Percent Change between Turning Points

Year

Trough or Peak (1)

Extended HP Moneya (2)

Extended HP Money (3)

HP-Money Proper (4)

1879 1882 1886 1890 1895 1900

trough peak trough peak trough peak

3,582 3,595 3,583 3,936 4,522 5,524

0.1 -0.1 2.4 3.0 4.4

0 0.2 2.6 3.2 2.9

Sources: Same as those for table 7.5, plus estimates of thirty-day bills (Deutsche Reichsbank 1901, table 35, stock estimate by Reichsbank) divided by 3 (to obtain thirty- from ninety-day bills) minus annual average of Reichsbank bill holdings (see table 7.1). aExtended high-powered (HP) money is high-powered-money proper (see table 7.5) plus estimates of bills of exchange of thirty days' maturity held outside the Reichsbank.

bank money holdings by German banks and the public. The results show desired money holdings determined by quarterly German permanent income (having a positive effect) and the Reichsbank average lending rate 8 (having a negative effect); other variables thought to affect desired money holdings had no statistically significant and/or sizable impact, as the first row of coefficients and the t-statistic values indicate. The coefficient for the lagged stock of Reichsbank money is given in the last column of table 7.13. If transformed to the magnitudes appropriate for levels, the lagged-stock coefficients of both equations indicate that over one-half (55 percent) of the discrepancy between the desired and the actual level of Reichsbank money was made up in the quarter when a disturbance caused a difference between actual and desired levels. During the following three quarters, nearly all the discrepancy between desired and actual Reichsbank money was eliminated. Thus, adjustment lags were not long at all. When evaluated at the means of table 7.13, the results indicate a decided stability of demand for money with respect to permanent income and the average Reichsbank lending rate (differing from the discount rate from 1892 to 1897 because the Reichsbank had preferential lending rates for some discounts in these years). A 3.7 percent rise in permanent income eventually generated a rise in demand for Reichsbank money of 1.2 percent, i.e., the income elasticity of demand was roughly one-third of one.9 A one-hundred-basis-point rise in the Reichsbank lending rate

338

Paul McGouldrick

Table 7.13

Regressions Explaining Absolute Quarterly Changes in Reichsbank Money, II 1892-IV 1907 Independent Variables

Equation ConNo. stant

Permanent Income

Reichsbank Reichsbank Lending Liquidity Rate

London Bill Rate

Lagged Reichsbank Dummya Money

(l)b

.018 (4.22)* .017 (2.94)*

-24.6 (2.77)* -26.1 (3.22)*

-205.8 (1.02)

-4.2 - (0.46)

-9.5 (0.56)

(2)C

442 (2.55)* 546 (1.57)

Adjusted R 2 Standard error Durbin h ratio

Equation (1) .347 31.8 .42

- .450 (4.72)* -.454 (4.77)*

Equation (2) .344 31.9 .42

Notes: Permanent income is scaled in millions of marks; the dependent variable and lagged

Reichsbank money in millions of marks; the Reichsbank and London interest rates in percentages; and Reichsbank liquidity (the ratio of its reserve to money liabilities) as a ratio. Absolute t-values are in parentheses. The means of the dependent and independent variables are as follows: change in Reichsbank money (millions), 8.986; Lagged Reichsbank money (millions), 1,694; Permanent income (millions), 30,110; Reichsbank lending rate, 4.078; London bill rate, 2.638; Reichsbank liquidity, .534; Dummy variable, .6349. aDummy variable, to test for effects of alleged shift in Reichsbank monetary policy, coded 0 for II 1892-IV 1897, 1 for I 189B-IV 1907. bOrdinary least squares. CTwo-stage least squares, with Reichsbank lending rate determined in the first stage by German and foreign short-term interest rates. *Significant at 1 percent level.

reduced demand for Reichsbank money eventually to a rounded 25 million marks, equal to slightly less than three-quarters of average quarterly growth. But evaluated at the means of the Reichsbank lending rate and the stock of Reichsbank money, this relation translates only to an arc elasticity of - .06, meaning a highly interest-inelastic demand for Reichsbank money. Evidently, then, making bills virtually equivalent to Reichsbank money in their liquidity characteristics does not imply a large cross-elasticity of demand between bills and Reichsbank money with respect to the interest rates on eligible bills. The Durbin h ratios permit us to infer that all systematic influences have been captured by the two equations shown. The stable demand for money and stable supply of bills of exchange, with respect to cyclical movements in output, permitted Reichsbank policy to be highly effective. By making gold freely available at a fixed price, the policy made gold and Reichsbank money close substitutes, and the policy of taking unlimited quantities of eligible bills at the prevailing

339

Operations of the German Central Bank, 1879-1913

discount rate made such bills and Reichsbank money also close substitutes in wealthholder portfolios. But while the substitutability of bills and Reichsbank money (and, by inference, also gold) corresponds with the views of James Tobin, the very low interest elasticity of demand for Reichsbank money corresponds more with the views of Milton Friedman. Low interest elasticity of demand for Reichsbank money enabled the German central bank to moderate interest swings (so as to keep the balance of payments in balance) over German cycles.

7.5

Testing the Interpretation of the Reichsbank's Influence

Tests of the foregoing interpretations can be set up by comparing relevant statistics for Germany and the United States. Both countries grew at very rapid rates prior to 1914, and both had high and strongly fluctuating ratios of gross-investment-in-plant-and-equipment to GNP (the United States) or net-social-product (Germany). Germany had a central bank, while the United States did not. The view presented in this study is that by keeping fluctuations in the discount rate smaller than fluctuations in open-market interest rates over pre-1914 business cycles, the Reichsbank succeeded in dampening German cyclical fluctuations in interest rates. A test of this hypothesis is to compare the cyclical behavior of the spreads between the German and London and between the New York and London open-market rates of interest. Since the basis of interest-rate spreads for both is London, noise from British cycles is the same in both series, and therefore does not affect comparisons between the two. Each spread is a proxy for shortterm capital flows: The higher is the rate of one country relative to the London rate, the greater should be the short-term capital inflow. (Unfortunately, I did not have access to intrayear data on gold inflows and outflows for the United States, and U.S. and German data on short-term capital movements are either faulty or completely lacking.) The New York-London spread widened during U.S. business upswings and fell during downswings in a least twelve out of eighteen U.S. cycle phases from 1879 to 1913. But the Berlin-London spread behaved quite differently; it rose during German upswings and fell during German downswings, in only three out of eleven German cycle phases (inferred form Morgenstern 1959, chart 18). A second test of the hypothesis is to compare differences in fluctuations of gold held by banks and the nonbank public in Germany and the United States. If the Reichsbank succeeded in dampening German business cycles, German gold fluctuations should have been smaller adjusted for the scale of the economy. The test is less satisfactory than the first one since I had access only to annual estimates of gold outside the Treasury for the United States (and for Germany after 1907 and before 1890), and

340

Paul McGouldrick

the brevity of American business cycles blurs comparisons using annual data. Still, the standard deviation of annual percent changes in gold held by banks and the public in the United States was 12.D-huge relative to the German equivalent standard deviation of 4.58 for the 1879-1913 period. Even if the years 1879-81 are excluded from the test, to eliminate disturbances due to the resumption of specie payments by the United States in 1879, the standard deviation declines only to 7.55-65 percent greater than the German figure of 4.58.10 A more satisfactory test is to compare German and U.S. cyclical movements of reserve-deposit ratios. If it is correct that the Reichsbank dampened cyclical swings in the sum of high-powered money and an asset close to high-powered money in its liquidity characteristics--eligible bills-by withdrawing one even as it supplied the other, reserve-deposit ratios should not have fallen during business upswings and risen during downswings. If they did occur, such swings should have been smaller than those in reserve-deposit ratios of U.S. banks. Like the U.S. banks, German banks could control their loans and investments and therefore their reserve-deposit ratios. If eligible bills were close to high-powered money as effective reserves, the Reichsbank's withdrawal of bills as it expanded the portfolio during business upswings, and converse behavior during downswings, should have changed effective bank reserves less than measured reserves, so that the creation of bank money would have been stable. Table 7.14 compares German reserve-deposit ratios of the credit and savings banks at cycle peaks and troughs from February 1879 to April 1913 and shows annual average percent changes in the ratios from trough to peak and peak to trough. By the test of comparing given-phase change with that in the preceding phase (see table note for specification), the German ratio moved countercyclically in six out of ten available phases (1882 to 1913), procyclically in only four out of ten. By contrast, the same test applied to U.S. reserve-deposit ratios (1879-1913) shows procyclical swings in changes in eighteen of nineteen cycle phases (Cagan 1965, table 27). By an alternate test (direction of change during cycle phase), the U.S. ratio behaved procyclically in all but two of sixteen cycle phases while the German ratio behaved countercyclically in five out of eleven phases.ll One interpretation casts an even more favorable light on the relative stability of German-bank reserve-deposit ratios. The cyclical contraction from 1890 to 1895 was the one exception to the Reichsbank's usual powerful influence on bill rates of interest and to the usual volume of bills that it absorbed. During those years, open-market bill rates reached lows not seen again until the 1930s in the United States. For reasons not explained in its official quarter-century history, the Reichsbank chose not to follow open-market rates down in setting its discount rate but to

341

Operations of the German Central Bank, 1879-1913

Table 7.14

Reserve-Deposit Ratios of German Credit and Savings Banks: Levels and Annual Average Percent Changes Between Turning Points, 1879-1913

Cycle Turning Point Feb. 1879 Jan. 1882 Aug. 1886 Jan. 1890 Feb. 1895 Mar. 1900 Mar. 1902 Aug. 1903 Feb. 1905 Jui. 1907 Dec. 1908 Apr. 1913

trough peak trough peak trough peak trough peak trough peak trough peak

Levels

Annual Average Percent Change Between Turning Points

0.0520 0.0530 0.0392 0.0447 0.0508 0.0416 0.0430 0.0397 0.0371 0.0355 0.0332 0.0282

+0.6 -5.78 +4.1 8 +2.7 8 -3.6 +1.7 -5.4 -4.4 -1.88 -4.68 -3.5 8

Source: Reserves and deposits at German credit and savings banks (Kreditbanken and Sparkassen), Hoffmann 1965, tables 202, 207. Notes: Both reserves and deposits were interpolated to cycle peak and trough months by changes in the preceding year (an alternate method-prorating by given-year changesproduced just about the same results in terms of ratio changes). To allow for the declining trend in the series, the test cited in the text is as follows. A ratio movement reinforcing business-cycle phases is algebraically larger than the change in the preceding phase for a business contraction, algebraically smaller for a business expansion. A change offsetting (dampening) a cycle phase has the opposite pattern. 8Contracyclicai. Other entries in the column are procyclical, except for the first ones.

practice price discrimination by discounting some bills at a preferential rate (Privatdiskontsatz). As a consequence, to some extent bills lost their high liquidity since banks and others were uncertain about their rates of return. Accordingly, the normal mechanism of Reichsbank control became partly inoperative. If this contraction and the following expansion are excluded (the latter because the phase starting point was distorted), the count runs: countercyclical changes in the reserve-deposit ratio in six cycle phases, clear procyclical movements in only two (using the test specified in the notes to table 7.14). A final test uses fluctuations of rates of exchange for the German mark and other currencies (the pound, the French franc, and the dollar). If the Reichsbank's monetary policy targeted on the German rate of exchange and operated successfully thereby, two results should have followed. Violations of gold points, i.e., mark price changes, which made it profitable to ship gold in settlement of claims rather than purchasing or selling marks, should have been rare. In addition, fluctuations of the foreignexchange value of the mark within the gold points should have been

342

Paul McGouldrick

randomlike instead of being correlated with interest rates, interest-rate spreads, or real activity in the German economy. In terms of violations of gold points, the Berlin-London exchange rate shows much better performance than three other exchange rates not involving Germany (New York-London, Paris-New York, and ParisLondon), according to Morgenstern (1959, table 56). The former shows only one monthly deviation from maximum gold points from 1880 to the outbreak of World War I, compared with four (New York-London), thirty (Paris-New York), and sixteen (Paris-London). By another criterion-percent deviation from median gold points-the average percent deviation was 0.06 for the German-London rate compared with 0.16 percent for both New York-London and Paris-New York and 0.09 percent for Paris-London. Concerning the key target, the Berlin-London exchange rate, the pound price of a hundred marks rose in two German business expansions and fell in four, while it rose in three and fell in two contractions. This performance permits a reasonable, though intuitive, inference that this exchange rate was not related to German business cycles, or if related, the association was weak. To the naked eye, the movements in the chart of the series resemble nothing so much as a random walk with respect to German business cycles (Morgenstern 1959, chart 21). The mark price does show longer-run swings, rising from 1880 to late 1880s and early 1890s, and thereafter falling to about 1900, after which the trend was stable. But these longer swings were not cycle-related. To summarize, the evidence is consistent with the hypothesis that the German central bank targeted exclusively on the mark exchange rate. Moreover, the policy was successful. The bank never suspended specie payments. In the one episode when such a suspension was threatenedthe 1907 crisis-it was clearly due to an exogenous event: American flotation of large loans in Europe and massive gold purchases following the American suspension of specie payments earlier in 1907 (U.S. National Monetary Commission 1910c, pp.306-7, 35, 341, 363, and 624). There is clear evidence that the Reichsbank used moral suasion to avoid large gold shipments; this episode is the only one in thirty-four years of pre-1914 operation when maximum gold points were violated. But a 7.5 percent discount rate turned the corner, and the October-November period of gold-point violation was succeeded by monetary ease. Nor did the real sector of the economy suffer unduly; in 1908 industrial output was only 0.9 percent lower than in 1907 and 6.9 percent higher than in 1906 (Hoffmann 1965, tables 10, 13). The foregoing explanation for Reichsbank success in monetary policy-eoncentration on convertibility into gold and British pounds as the overriding target-is not the only possible one. Another could be found in terms of information available to bankers and the public. In the United

343

Operations of the German Central Bank, 1879-1913

States, each banker knew his own reserves of high-powered money at all times but had only the vaguest idea of current reserves available to other banks. In Germany, the centralization of gold reserves in the hands of the central bank made the reserve position of all banks (in terms of ultimate liquidity) known each week when a statement was published. This information facilitated better coordination of lending and investing decisions by the banking system. In addition, the continued use of gold as a medium of exchange and store of value by the nonbank public served to police central-bank behavior. As stated earlier, the proportion of gold to Reichsbank money did not decline over 1879-1913, and each business upswing was accompanied by an internal drain of gold from the central bank, forcing the Reichsbank to p':!y attention to the gold value of its currency. 7.6

Reasons for Reichsbank Policy

Maintenance of a stable gold value of the mark, accordingly, was the true rule of the game and changing the discount rate as the proportion of reserves to money liabilities varied was a false rule. The Reichsbank indeed influenced domestic short-term interest rates. This rule rejects two older views about central-bank behavior prior to 1914, for Germany at least. One of extreme laissez-faire provenance was the Konstatierungstheorie: The central bank never determines the appropriate, marketclearing discount rate but only finds it (Bopp 1944). The second is that prior to 1914, central banks used gold devices, foreign-exchange holdings, credit rationing, and open-market policies to insulate central banks from the rigors of following market rates (Bopp 1944; Sayers 1936). Reichsbank policy also did not conform with the mechanical link of central-bank money liabilities to gold espoused by the currency school in England prior to Peel's Bank Act. Such a link would have been valid only if the velocities of central-bank money and gold were identical (Viner 1937, pp. 221-22), and modern experience permits the inference that this would have been impossible if gold and central-bank money had been used in different proportions by the household, government, and business sectors of the German economy (McGouldrick 1962). Such a difference in proportion is highly likely; the German government, for example, was bound by law to make all payments of civil-servant salaries in gold coin. But the convertibility target satisfied the rule of at least one member of the English currency school, Lord Overstone, who specified only that a proper system should "secure with respect to a paper currency that regulation of its amount which in a metallic currency necessarily results from its intrinsic value" (Viner 1937, p. 389). But why did the Reichsbank act as it did? A plausible reason was profitability considerations, understood as the selection of the desired

344

Paul MeGouldrick

point on the Reichsbank's expected-rate-of-return-risk curve. For one thing, the Reichsbank imported nearly all gold for German nonmonetary as well as monetary uses, and it was obliged by law to sell gold to private buyers or the mint at a fixed price without seigniorage. Stabilizing the exchange rate therefore reduced risk on its gold operations. Another profitability consideration was that of earnings on the Reichsbank's portfolio. Letting the portfolio rise and fall as interest rates did was consistent with profit influences. On the other h~nd, the apparent success in avoiding overexpansion when rates were high or rising and thereby a depreciation of the mark suggests that profits or losses from gold operations offered a happy counterweight to profits from the portfolio and so assured that the central bank followed the convertibility target of policy. Another reason for the choice of that target was the uncertainty surrounding any other. The offer to purchase all bills offered at the prevailing discount rate helped to reduce bank reserves to very small proportions of deposits, as can be seen by comparing the reserve-deposit ratios of table 7.14 with the far higher U.S. ones at the time. However, huge seasonal and irregular fluctuations in demand for Reichsbank money liabilities necessarily occurred, as banks carried out frequent and large switches between eligible bills and money. Frequently the maximum note issue within a year was twice the size of the minimum note issue. (See the large quarterly percentage changes in Reichsbank money liabilities shown in table 7.7.) Hence the market, not the central bank, had to determine the appropriate quantity of central-bank money, given the unlimited-availability rule. Targeting on interest rates was also impossible because of the sensitivity of German to foreign rates and questions about which interest rate was appropriate. Macroeconomic information was absent or available with long lags. The exchange rate on London, the world's leading financial center, was left as the only feasible target. Profit-maximizing within the constraint of exchange-rate targeting also appears plausible from the structure of the central bank's management. It is true that stockholders had no voice in the voting assembly of the Reichsbank; their interests were confined to directors who had the right to be present at meetings and to advise and express opinions but not to cast votes. The directorate that managed the Reichsbank was composed exclusively of civil servants, and stockholders received only four-tenths of all profits in excess of a 4.5 percent of share capital. But, as noted earlier, the salaries of all higher officials were made proportionate to Reichsbank profits (except for a fixed component) in order to avoid "bureaucracy" (Flink 1930, pp. 11-12), and dividends varied as profits did. These facts are quite consistent with profitability as a subsidiary target. Given the overriding constraint of keeping the mark in a stable relationship to gold, expanding the portfolio as demand for money rose and

345

Operations of the German Central Bank,

1~79-1913

contracting it as that demand fell may well have been optimal behavior from a social point of view. The behavior did not result in procyclical movements in central-bank money liabilities as more "altruistic" rules of central banks did in the post-1945 era. The convertibility target allowed the public and the banks to determine the quantity of central-bank money that it desired to hold and avoided indirect manipulation of desired money holdings by the central bank. As shown earlier, the demand for central-bank money was remarkably stable and business cycles were mild. If profitability played a role (we cannot be certain for lack of information-the surviving archives of the pre-1914 Reichsbank are located in East Germany and are inaccessible to Western scholars), the results may not have been undesirable in the slightest. 7.7

Concluding Observations

The hypothesis that the sources of high-powered money affected the impact of the aggregate on economic activity and liquidity may not be applicable only to imperial Germany prior to 1914. In their celebrated Monetary History of the United States, Friedman and Schwartz (1963) ascribed the stability of velocity from 1947 until the outbreak of the Korean War to federal fiscal policy and expectations that prices would later fall. But a contributing factor might also have been the Federal Reserve bond-price-support program which "liquified" U. S. government securities of all maturities in much the same way that the pre-1914 German Reichsbank "liquified" private bills of exchange. Either should have been inflationary, the United States program much more so than that of the Reichsbank because the latter could and did change the support price (the discount rate) rather frequently and also because Germany was constrained by unlimited convertibility into gold. But American velocity and prices were astonishingly stable; the latter even declined in 1948-49, the one and only time that prices fell in a recession during the postwar era. Quite posibly therefore, the bond-price-support program may have reinforced the effect of stable price expectations by neutralizing errors of the central bank in estimating the U. S. demand for money and acting on these estimates. Just as the Reichsbank removed one source of liquidity (bills) as it provided another (money) by bill purchases, the Federal Reserve did the same with government securities. (Of course, stable or falling price expectations were also necessary for this result). It is ironic that implementation of the 1951 accord between the Federal Reserve and the Treasury was followed not by stable but by rising prices and velocity. The foregoing hypotheses and conclusions do not necessarily have applicability to current policy problems and specifically to whether the United States should attempt to return to some type of gold standard.

346

Paul McGouldrick

The pre-1914 system had evolved over centuries and was embedded in a culture emphasizing sanctity of contracts and government restraint in economic intervention. Whether substitute preconditions for confidence in a hypothetical gold standard of the future can be created is more than doubtful. At the same time, however, the relative stability of the German monetary system must be described as remarkable. German economic growth was centered in areas-iron and steel, chemicals, electrical and residential construction-that are frequently alleged to impart a high degree of instability to economies undergoing economic growth. Even more to the point perhaps, German credit banks were highly venturesome, investing in long-term, illiquid capital of industrial enterprises and underwriting securities issues on a considerable scale (Whale 1968). Political historians such as Mann and Holborn, cited earlier, have also emphasized the political and social conflicts of the era. The contrasting relative stability of the monetary realm therefore stands as strong testimony to the advantages of the pre-1914 gold standard when properly ruled by a central bank.

Notes 1. The titles of the balance-sheet categories remained constant throughout the period. The only serious confusion as to the character of assets and liabilities seems to have been the inclusion of foreign-exchange assets under "other assets" instead of under reserves (Metall). Such assets were small relative to metallic reserves, as noted later. 2. Weekly balance sheets were aggregated into monthly and quarterly averages by the writer and then seasonally adjusted with the Census X 11 program, component by component. The usual caveat applies that after seasonal adjustment, balance sheets do not add up exactly as they did before. 3. All cycle phases are measured from trough to peak and peak to trough. The dates of German troughs and peaks were selected by Burns and Mitchell (1946, table 16). 4. By the extended base adjustment, Reichsbank money is calculated as if it had always included the note issues of German banks giving them up prior to 1913. Thus, if a bank gave up its note issue and the Reichsbank assumed it in year t, Reichsbank money is increased by this amount for all years prior to t, eliminating distortions due to a transfer of note liabilities from a bank to the central bank. 5. Industrial use of gold has not been subtracted from net inflows. However, this qualification serves to make the recorded changes in tables 7.5 and 7.6 more procyclical (or less countercyclical) than they probably were in fact, since it is reasonable to assume that industrial use of gold rose more proportionately than money use during business upswings and fell more proportionately than money use during downswings. Hence the true stocks and flows of monetary gold would show less of a procyclical movement than the series in tables 7.5 and 7.6. 6. Base (high-powered) money also included Treasury fiat money (Reichskassenscheine), omitted from the levels of table 7.5 from which the percentage changes over cycle phases were computed. However, the amount of fiat issued by the Reich Treasury was fixed after 1876, so the estimated extremely small fluctuations in fiat holdings of banks and the public (unpublished worksheets) reflect only randomlike movements. Therefore, exclusion

347

Operations of the German Central Bank, 1879-1913

of this component affects only the estimates of levels in table 7.5 and not the annual average percent changes. 7. A monetarist interpretation would be that by setting price supports for eligible bills, the Reichsbank made them as good as high-powered money and hence justified their inclusion in the stock of high-powered money. The interest rate on the bills compensated holders for the small losses caused by adjustments of the support price, i.e., the discount rate. 8. Preferential rates, prior to 1897, were limited to bill purchases by the Reichsbank at rates below the discount rate. This discriminatory pricing of bills purchased was said to have been due to the very low level of short-term interest rates in the 188(}-96 period. The real motivation behind the pricing was probably profit maximization. To account for the deviation from the announced discount rate, the interest rate used in the regressions shown in table 7.13 is the average lending rate of the Reichsbank on all bills of exchange. 9. The low income elasticity of demand is explainable by the rapid development of German credit and savings banks from 1879 to 1913. These institutions offered superior substitutes for Reichsbank money. In addition, gold more than maintained its place as a form of high-powered money; estimates of gold outside the Reichsbank increased by 111 percent from 1879 to 1913, while Reichsbank money rose by only 62 percent (unpublished worksheets). Rising international tensions might possibly explain the increase in gold six years before World War I broke out; statements made to the bank inquiry of 1908 already predicted a suspension of specie payments in case of war. 10. U.S. percentage changes computed from Cagan 1965 (table F-7, col. 1); German percentage changes computed from Hoffmann 1965 (table 240, column Metallgeld). The former series excludes, the latter includes, subsidiary coinage of silver and base metals. While Cagan stresses that U.S. high-powered money changes were less procyclical during the pre-central-bank era than they were after the Federal reserve system started operations in 1914, the comparison here is not between different periods in the United States but between the United States and Germany for one period, that prior to 1914. In a footnote Cagan acknowledges some possibility that output fluctuations prior to 1914 induced accommodating gold flows (Cagan 1965, pp. 245-46, n. 7). 11. Legal or central-bank nonprice rules cannot be invoked in the German case as a cause of good-banking performance. Throughout the 1872-1914 period, German banks were not subject to reserve requirements (banking regulations covered ratios of capital to assets and types of permitted transactions only). At the very end of the period, Reichsbank President Walther Havenstein attempted to persuade banks to carry money reserves equal to 15 percent of deposit liabilities, but he did not succeed (Flink 1930, p. 35).

References Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard, 1880-1914. New York: Federal Reserve Bank of New York. Bopp, Karl. 1944. Central banking at the crossroads. Proceedings of the American Economic Association 4 (Mar.): 260-77. - - a 1954. Die Tiitigkeit der Reichsbank von 1876 bis 1914. Weltwirtschaftliches archiv. Bd. I. Borchardt, Knut. 1976. Wiihrung und Wirtschaft bis zum I. Weltkrieg in der Deutschen Bundesbank. In Wiihrung und Wirtschaft in Deutschland 1876-1976. Frankfurt/Main: Verlag Fritz Knapp.

348

Paul McGouldrick

Burns, Arthur F., and Wesley Clair Mitchell. 1946. Measuring business cycles. New York: National Bureau of Economic Research. Cagan, Phillip. 1965. Determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. Deane, Phyllis. 1968. New estimates of gross national roduct for the United Kingdom, 1830-1914. Review ofIncome and Wealth 14 (June): 104-7. Deutsche Reichsbank. 1901. Die Reichsbank 1876-1900. Berlin: Reichsdruckerei, Kommissions Verlag von Gustav Fischer, Jena. - - . 1925. Die Reichsbank 1876-1925. Berlin: Reichsdruckerei. Flink, Salomon. 1930. The German Reichsbank and economic Germany. New York: Harper & Brothers. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Gallman, Robert E. 1965. Annual estimates of gross national product in the United States, 1834-1909. University of North Carolina. Mimeo. Hoffmann, Walther. 1965. Das Wachstum der deutschen Wirtschaft seit der Mitte des 19. lahrhunderts. Berlin: Springer Verlag. Holborn, Hajo. 1969. A history of modern Germany. Vol. 3. New York: Alfred Knopf. Kaiserliches Statistisches Amt. 1879-1914. Statistiches lahrhbuch fur das Deutsche Reich. Annual volumes. Berlin: Verlag von Puttkammer & Miihlbrecht. Lebergott, Stanley. 1964. Manpower in economic growth: The American record. New York: McGraw-Hill. McGouldrick, Paul F. 1982. "Rules of the game" and operation of the Reichsbank under the gold standard. Paper delivered at the National Bureau of Economic Research conference, A Retrospective on the Classical Gold Standard, 20 March 1982, at Hilton Head, South Carolina. - - - . 1962. A sectoral analysis of velocity. Federal Reserve Bulletin 48 (Dec.): 1557-70. Mann, Golo. 1968. The history of Germany since 1789. London: Chatto and Windus. Mitchell, B. R., and Phyllis Deane. 1962. Abstract of British historical statistics. Cambridge: Cambridge University Press. Moggridge, D. E. 1969. The return to gold, 1925. Cambridge: Cambridge University Press. Morgenstern, Oskar. 1959. International financial transactions and business cycles. Princeton: Princeton University Press. Riesser, Jacob. 1911. See U.S. National Monetary Commission 1911b. Sayers, Richard S. 1936. Bank of England operations, 1890-1914. London: P. S. King and Sons. Sommer, Albrecht. 1931. Die Reichsbank unter Herman von Dechend. Berlin: Heymann's Verlag.

349

Operations of the German Central Bank, 1879-1913

Tilly, Richard. 1973. Zeitreihen zum Geldumlauf in Deutschland 18701913. Jahrbacher far Nationaloekonomie und Statistik 187 (Heft 4). Tobin, James. 1965. The monetary interpretation of history. American Economic Review 55 (June): 464-85. Tobin, James, and W. C. Brainard. 1963. Financial intermediaries and the effectiveness of monetary controls. American Economic Review Papers and Proceedings 53 (May): 383-400. U.N. 1950 68. Yearbook, vols. 4-22. New York: United Nations. U.S Bureau of the Census. 1960. Historical Statistics of the United States, Colonial Times to 1957. Washington, D.C.: GPO. U.S. Department of Commerce, Office of Business Economics, 1950-68. Survey of Current Business, monthly issues. U.S. National Monetary Commission. 1910a. The Reichsbank, 18761900. In The Reichsbank and renewal ofits charter. Vol. 10. Translated from the German. Washington, D.C.: Government Printing Office. - - - . 1910b. Renewal of Reichsbank charter. In The Reichsbank and renewal of its charter. See U.S. National Monetary Commission 1910a. - - - . 1910c. German bank inquiry of 1908, stenographic reports. Vol. 12. Washington, D.C.: Government Printing Office. - - - . 1911a. Articles on German banking and German banking laws. Vol. 11. Washington, D.C.: Government Printing Office. - - - . 1911b. German great banks and their concentration. Vol. 14. Prepared by Jacob Riesser. Washington, D.C.: Government Printing Office. - - - . 1911c. Statistics for United States, Great Britain, Germany, and France. Vol. 21. Washington, D.C.: Government Printing Office. Viner, Jacob. 1937. Studies in the theory of international trade. New York: Harper and Brothers. Whale, Philip. 1968. Joint stock banking in Germany. London: Frank Casso

Comment

Heywood Fleisig

Several yardsticks could be used to measure the contribution of this paper. We might ask, in the framework of a historian's "model," how the paper contributes to understanding the goals and instruments of Reichsbank managers as they operated their institution under the gold standard. Or we might ask questions answerable in the framework of economic Heywood Fleisig is an economist in the Economic Analysis and Projections Department of the World Bank, Washington, D.C. The author is indebted to Gertrud Fremling-Lott, Lance Girton, Catharine Hill, and Joanne Salop for several helpful comments.

350

Paul McGouldrick

models: How did Reichsbank operation affect German growth and macroeconomic performance? Does examination of the Reichsbank help us distinguish between the Viner version of the specie-flow mechanism and the perfectly arbitraged version of gold standard adjustment set forth in the McCloskey-Zecher paper? Or, more generally, does the paper help choose among competing contemporary theoretical models? The last question might cast light on the desirability of adopting the gold standard today, a question of contemporary interest considered in the Report of the Commission on the Role of Gold (U.S. Congress 1982), released shortly after this conference was held. By these measures, the paper falls short of delivering the evidence we need in order to choose among competing historical and economic views of the operation of both the Reichsbank and the gold standard. The Paper's Contribution to Resolving Historical Issues A series of questions might interest historians and economists concerned about the German case study that wouldn't require rich economic tools to answer: What did Reichsbank operators intend? How did they achieve their intentions? What were their larger objectives, if any, for the German economy? How did they see the gold standard furthering or impeding these objectives? What did they perceive as the relation between Reichsbank operations and the major economic changes occurring in Germany and the world? The paper more narrowly circumscribes its inquiry, though, finding that: Reichsbank holdings of bills increased with domestic German economic activity. Reichsbank lending rates fell relative to market interest rates as domestic German economic activity expanded. The money stock moved inversely with German economic activity because giro accounts fell as interest rates rose. And gold drained internally from the Reichsbank as German income rose. An historian of the gold standard would probably find these facts useful, but they don't move us very far along in understanding what Reichsbank policymakers intended and how they executed their actions. The author sets forth manifestations of intent, but his quantitative measures are consistent with unlimited numbers of stories about the intentions and perceptions of Reichsbank operators. What we see in the pattern of Reichsbank lending rates, for example, might have represented its operators' intent, might have occurred despite their best efforts, might have been the unintended and-to them-the inconsequential side effect of something else they were attempting, or might have been-in their view-exogenous or unavoidable. The paper doesn't present documented hypotheses about the behavior of Reichsbank operators, so its quantitative evidence doesn't permit choosing among these hypotheses.

351

Operations of the German Central Bank, 1879-1913

Historians, like economists, can't escape the positivist assumption underlying science, as expressed starkly and perhaps somewhat extremely by Koopmans: No useful measurement occurs without theory. Explaining Reichsbank behavior requires causal statements. Cause is defined by the model, not observed in the world. Hume is surely the last philosopher we should neglect at a conference on the gold standard. Explanations, therefore, require models. Historians construct the functional equivalent of economists' models with webs of conjectures about motive, means, and goals, constructed by examining the complete historical record: documents, letters, speeches, debates, diaries, memoirs, and newspapers. From these facts and conjectures historians produce candidate explanations consistent with these facts and with generally held beliefs about human behavior. Then they devise verbal and quantitative tests to separate false and true explanations. This difficult, but achievable, standard marks the best monetary histories of the interwar years: those of Chandler (1970), Clarke (1967), Friedman and Schwartz (1963), Kindleberger (1973), and Wicker (1966). These histories all digest large volumes of qualitative material, pose hypotheses about behavior, and examine them with qualitative or quantitative tests. Their hypotheses often reflect both older and present views of how these economic links work, so that we understand how past policymakers operated, not only in terms of their own "views of the world" but as their actions and as events confirm or disconfirm our own current views of how the world works. Sometimes the results are stunning, as when Chandler and Clarke spelled out the motivation for the stabilization loans of the mid-1902s; sometimes the results are provocative, as when Friedman and Schwartz documented the Federal Reserve's puzzling lack of concern over halting the bank failures in the face of the Federal Reserve's clear mandate to operate as the domestic lender-of-Iast-resort; and sometimes the results are ambiguous, as when we try to determine whether Warren or Rogers had the most influence on Roosevelt's 1933 decision to devalue when both Warren's gold-price theory and Rogers's desire to maintain sterling parity conformed to Roosevelt's ultimate action. Little of this historical hypothesis building and testing occurs in the paper. Rather, the paper proposes to circumvent that procedure by deducing policymakers' motives from the movements of variables they might control. But such a procedure can no more succeed here than in economics, where econometric reduced-form tests can never distinguish between two different models with the same reduced form. The paper frequently asserts its support for this "historical reduced form," but by failing to independently determine the motives of the Reichsbank operators, this method preserves an elemental vagueness about what the

352

Paul McGouldrick

operators were doing. The same methodology, after all, would permit concluding that the Federal Reserve desired the rising volume of bank failures witnessed in its first twenty years. The Paper's Contribution to Resolving Economic Issues The paper needn't answer these historical questions, however, if it aims instead at evaluating the performance of an economy operating under the gold standard. Evaluating German macroeconomic performance does not require knowing the motives of those operating the Reichsbank. It requires knowing only the movements of Reichsbank instruments. Our present understanding of how those instruments affect the economy permits evaluating the effectiveness of Reichsbank operation. Toward that end, the paper produces several measures of German economic performance: Growth rates of GNP and industrial production showed less year-to-year variation between 1879 and 1913 than between 1951 and 1968, unemployment rates were "low," price stability was "reasonable," and interest rates were "low." But what permits linking Reichsbank "operation"-and I use this terminology to emphasize that so far we have no evidence of "policy" at all-with the behavior of any of these macroeconomic series? Claiming these macroeconomic consequences follow from Reichsbank actions requires causal statements, and causal· statements require models that define cause. Except in this case the models are not the vague, multidisciplinary, heuristic models of the historian, but rather the more formal, optimizing, behavioral models of the economist. And here this paper, like several others at this conference, goes badly astray: It presents tests that can't distinguish among models with dramatically different implications concerning how the Reichsbank could affect the money stock, and how money could affect real activity. In one rationalization of the specie-flow mechanism, for example, mpney neutrality operates with some lag; gold flows can alter national money stocks and change some combination of price levels, terms of trade, real trade volumes, and real output. Money and output are correlated, and the Reichsbank can use its financial operations to influence real activity. But in a Mundell-Fleming gold standard world, perfect capital mobility would ensure that no change in monetary instrument could affect real activity. Rather, only real changes can affect real activity: investment incentives, government spending, taxes, or tariffs. Since money demand is always satisfied, we again observe money correlated with output; but money is not a policy instrument and the Reichsbank cannot affect real activity. Or, in the McCloskey-Zecher (1976) version of the monetary theory of the balance of payments, money neutrality is always satisfied, product substitution is high, and arbitrage is perfect. Monetary policy cannot change relative price levels, the trade balance, or real output-it can only

353

Operations of the German Central Bank, 1879-1913

cause gold movements. Money and output are again correlated, but Reichsbank operations have no effect on the real economy. This paper uses a specie-flow model that essentially assumes that economies interact only through the movement of gold and are in other respects closed. Using a closed-economy monetary model to describe an open economy is an understandable error in the Friedman and Schwartz monetary history, written on the eve of "Mundell-Fleming" (1968; 1971) before twenty years of professional ferment on the question of how open economies interact. But what can be the use of replicating correlations between money and output when we now know they cannot distinguish among competing underlying models? My own priors, weak for reasons I set forth below, are that nineteenthcentury central bankers were probably right in believing they had little control ~ver national price levels and real output. I will now present some theoretical issues that might help distinguish among these various versions of gold standard operation and that might give hints about how changes in the Reichsbank's economic environment might have changed its ability to operate during the period considered in the paper. And I will consider some other theoretical issues that raise questions about other functions that central banks might have served under the gold standard. Capital Mobility

In an early Mundell-Flemirig result, perfect capital mobility under fixed exchange rates and money neutrality combined to ensure that central-bank operations had no effect on domestic output. In that world, the effectiveness of domestic monetary change, as found in the Friedman and Schwartz monetary history, could depend on capital-market imperfection, or on substitution between assets denominated in different currencies being sufficiently imperfect to permit domestic interest rates to diverge from foreign interest rates and differentially affect the home economy. Without such barriers to capital mobility or imperfect substitution among assets, U.S. interwar monetary policy could have caused-the U.S. Great Depression only if U.S. monetary policy so dominated other countries' monetary policies that it produced recession abroad at the same time or even before it produced recession within the United States. The latter, of course, was Keynes's own view of the origins of the Depression. In this Mundell-Fleming world, countercyclical movements of monetary aggregates in small economies can affect the real economy only when barriers impede capital flows. The history of monetary policy becomes the history of the integration of one country's capital market with those of other countries. Whether integration was sufficiently small or changed enough to permit the Reichsbank to affect relative interest rates by enough to affect real output would be an interesting question. But aside

354

Paul McGouldrick

from an inconclusive regression on U.K. interest rates and a conjecture that direct foreign investment might have financed the trade deficits that were positively correlated with German income movements, the paper does not discuss how changing asset substitutability and interest-rate differentials might have combined to change Reichsbank control over the real economy. Goods-Market Integration and Price-Level Determination

Integration of goods markets, homogeneity of commodities, and neutrality of money imply no difference in national price levels, measured in gold, so long as countries remain on the gold standard; relative national aggregate price levels, moreover, cannot change if the exchange rate is fixed. The central bank performs no function besides guaranteeing gold convertibility (I will discuss default below). If the central bank attempts to. change the domestic money stock in order to change the externally imposed price level, gold will flow in or out until the attempt stops. This prospect explains the extreme view of the monetary theory of the balance of payments, where the central bank determines only gold flows, not the price level or real activity. The current international-finance literature largely views the perfectsubstitute, short-run money neutrality version of the monetary theory of the balance of payments as inadequately characterizing international economic adjustment. In most current views, national price levels do diverge and purchasing-power parity does not hold at every instant. National price levels and real terms of trade change-not by accident or randomly as some papers here suggest-to equilibrate both goods and asset markets (discussed below). One determinant of the effectiveness of national monetary policy, therefore, lies in how much the terms of trade can change, which depends on changing substitutability among different national outputs and the speed with which long-run money neutrality obtains. An investigation of this issue might stress the shift from noncommercial agriculture to commercial agriculture facing world markets, the rise of product-differentiating manufacturing industries, changes in market organization such as cartel formation, evolution of national market power, or changes in commercial policy. Each of these changes occurred in Germany during this period. Dramatic changes in potential terms-of-trade adjustment can occur over a thirty-five-year period. The adjustment will differ substantially among countries and may explain changes in the apparent performance of the gold standard as well as national differences in its operation. But the paper I discuss, like several others presented here, addresses such issues only peripherally because such questions have no place in the paper's theoretical framework.

355

Operations of the German Central Bank, 1879-1913

Portfolio Balance and Asset-Market Equilibrium

Under the gold standard, domestic economic agents collectively hold at least the money, gold, and securities of their own countries. Yet these assets have different traits and cannot be perfect substitutes because, among other reasons, there is some chance of uncorrelated defaults (a point recognized in the paper). The effectiveness of monetary policy could be greater than I expressed in my weak prior, or it could increase over time, because different financial instruments exist or come into being. Such instruments could give the Reichsbank more ability to influence interest rates and interest-rate differentials, since imperfect substitution would have changing rates equilibrating the foreign demand for Reichsmark-denominated interest-bearing assets and the domestic German demand for money. Bits and pieces of this story appear in the paper: federal debt rises to 10 percent of social product by 1913, an unspecified mechanism determines the volume of bills of exchange, the Reichsbank might be holding foreign securities at the end of the period, inflows of direct foreign investment might finance payments imbalances (in unspecified form), but short-term lending might as well. In newer hypotheses about international adjustment, monetary changes alter portfolio equilibria and generate the changes in the real terms of trade required for adjustment. So in one version of the story, relative price-level adjustment would be determined entirely by the pieces of the gold standard story this paper holds offstage. Default Conditions

In one polar view of the gold standard, central banks serve no national monetary stabilization function. Price levels in individual countries remaining on the gold standard rise and fall system-wide since national monetary policy has no uniquely national effect; national policies spill out and affect the whole system. In any country, for any nominal value of gold, default conditions determine the volume of paper claims convertible to gold-at some volume of such redeemable claims, the promise to redeem becomes implausible, and a run on the country's gold stock forces it off the gold standard. These default conditions determine the stock of assets convertible to gold and the price level in each currency and, thereby, ~he world price level associated with a certain set of national currency gold prices. The central bank can't affect the price level or real activity; only different real factors produce differences in national economic outcomes. In such a world, the only important role of the central bank is maintaining gold convertibility-managing currency to avoid internal and external

356

Paul McGouldrick

drains large enough to produce runs. The correct model for the Reichsbank would emphasize the same issues as would the solvency of any individual borrower-the relation between the stock of gold and the stocks of currency and national-debt instruments convertible to gold. Or, in a less polar view of gold standard operation, a different default risk might, by itself, guarantee the imperfect asset substitution required to give the Reichsbank some influence over interest rates and real variables. Imperfect substitution could arise from the different probabilities of default on the pledge to redeem in gold different assets like gold coin, Reichsbank notes, federal-government debt, and private debt. Centralbank operations in that portfolio could then change relative rates of return and real activity. These issues have been discussed extensively in the bank-run, speculative bubble, anq debt-default literature. But to answer these questions we need to know such things as the expected rate of GNP growth, the savings rate, the maximum-sustainable current-account surplus, and the variance of these magnitudes. Here, as before, the paper presents no such data and measures no such concept because the paper's underlying model ignores these elements in explaining the link between Reichsbank actions and German macroeconomic performance. Finite Resource Economics Long before OPEC drove Hotelling's (1931) result in finite-resource economics into every corner of the profession----even to the gold standard with the work of Flood and Garber (1981)-the "finiteness" of gold was featured as an attractive part of making it money. Now we know, though, that equilibrium in the gold market requires that the marginal profit from holding gold rise with the rate of interest. Under some cost conditions, the price of gold must then rise as well. But if the nominal price of gold is fixed, its real price can rise only if the general price level falls. In such a world, general price stability exists only when serendipitous changes in gold refining and discovery shift the rising real-equilibrium gold-price path downward to offset its rising equilibrium price path. The older, itself precarious view, was that Providence, in its wisdom, would permit the gold stock to rise exogenously at about the same rate that exogenous labor-force growth and technical progress continued to raise output and increase the transactions demand for money. To this view we now must add the requirement of unexpected increases in the gold stock; expected ones can't suffice because they will be factored into the initial price path, determining the initial level but not its rate of increase. Conclusion It is always difficult to balance the unfairness of attacking the author for failing to write the paper you would have liked to read with the need to

357

Operations of the German Central Bank, 1879-1913

point out where the paper falls short of contributing to our knowledge. Moreover, this paper certainly does not stand alone at this conference in neglecting the issues raised in this comment. As a history paper, the author presents some measures of Reichsbank operation that may be consistent with any number of bank objectives including, as the paper favors, nothing more than the desire to maintain gold convertibility. The evidence presented in the paper, though, proves no one Reichsbank objective more important than any other. As an economics paper, all the evidence presented is consistent with the simple null hypothesis I started with: independent German Reichsbank manipulation of monetary aggregates could have no important independent effect on German prices, interest rates, or output between 1879 and 1913; all important events in the real German economy occurred because of other events in the real German economy or because of world movements in monetary variables. As I have indicated, I hold this view only weakly because other factors could have made national monetary policy effective; but the paper does not cast light on this possibility.

References Chandler, Lester V. 1970. America's greatest depression, 1929-1941. New York: Harper & Row. Clarke, Stephen, V. O. 1967. Central bank cooperation. 1924-31. New York: Federal Reserve Bank of New York. Fleming, J. Marcus. 1971. Essays in international economics. Cambridge: Harvard University Press. Flood, Robert P., and Peter M. Garber. 1981. Gold monetization and gold discipline. International Finance Discussion Papers no. 190, Board of Governors of the Federal Reserve System. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Hotelling, Harold. 1931. The economics of exhaustible resources. Journal of Political Economy 39 (Apr.): 137-75. Kindleberger, Charles P. 1973. The world in depression, 1929-1939. Berkeley: University of California Press. McCloskey, Donald N., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance of payments, ed. J. Frenkel and H. G. Johnson. Toronto: University of Toronto Press. Mundell, Robert A. 1968. International economics. New York: Macmillan. U.S. Congress. 1982. Report of the commission on the role of gold in the domestic and international monetary systems. Washington, D.C.: GPO.

358

Paul McGouldrick

Wicker, Elmus R. 1966. Federal Reserve monetary policy, 1917-1933. New York: Random House.

General Discussion PIPPENGER commented on Fleisig's remarks concerning the portfoliobalance approach to international adjustment. The portfolio-balance view suggests that there should be systematic patterns in the real exchange rate. Pippenger suggested that real exchange rates apparently performing a random walk are inconsistent with the asset-market approach to the balance of payments. In respect of the behavior of the Reichsbank, Pippenger suggested a very simple model. Portfolio balance is an inappropriate assumption to use to model the behavior of the Reichsbank, a profit-maximizing institution. Since the Reichsbank's stock was privately owned, it almost certainly had an interest in profit. Such a bank would have had a desired reserve ratio, which would fall as market interest rates rose. As the market rate rose, the Reichsbank would increase the rate it charged for loans, but by less than the market rate, narrowing the spread between market rates and loan rates, increasing its holdings of securities, and reducing its reserve ratio. MCGOULDRICK responded by noting that his paper mentions as a possibility the hypothesis of profit-maximizing behavior. The problem with the profit-maximizing hypothesis is that it implies that the mark should fall in value relative to other currencies during an upswing. However, we do not observe the mark falling in value, even within the gold points, during business-cycle upswings, and, symmetrically, we do not see the mark consistently rising in value during business downswings. While acknowledging that the Reichsbank had private shareholders, McGouldrick noted that Reichsbank officials were Reich civil servants, paid according to the civil servant's scale. One does not usually think of bureaucrats as intimately concerned with the level of dividends paid to shareholders. But the question remains open. FRENKEL commented on various discussants' different conceptions of the monetary approach to the balance of payments. He argued that the monetary approach is merely an analytical framework that suggests to the extent a central bank conducts open-market operations, it can affect the composition of the monetary base but not its level, at least in the long run. The speed of adjustment is a separate consideration. Frenkel argued that it may not be illuminating to pose various approaches in the context of a horse race, where one must win and others must lose. A reasonable approach must simply be consistent with the fact that a deficit in the balance of payments cannot be sustained unless the central bank is both

359

Operations of the German Central Bank, 1879-1913

willing and able to consistently intervene in such a way as to supply more money than is demanded. EICHENGREEN raised a point in connection with the discussion of the proper way to model the international transmission mechanism. He noted the absence from the conference of a paper that presents evidence concerning covered and uncovered interest parity for the major participants in the classical gold standard. Fleisig's comments could be taken to suggest that there existed a risk premium that would have rendered assets denominated in different currencies imperfect substitutes for one another, permitting interest parity to be violated. Another way of modeling the financial sector is to assume the absence of a risk premium and, since interest parity holds, to concentrate on the margin of substitution between money and bonds. The first approach would suggest the adoption of a fully articulated portfolio-balance model; the second would suggest a simple model of the money market. Thus, a crucial question is the importance of currency-specific risk. There are a number examples in the late nineteenth century of Latin American countries that defaulted on their foreign debt. Whether there are examples of default or similar sources of currency-specific risk close to the center of the gold standard system would seem to be a crucial consideration in guiding choice of specification for modeling the financial sector. RICH questioned McGouldrick's assertion that the German monetary base actually moved countercyclically. He argued that one must also take into account gold coin held outside banks, which moved procyclically and may have more than offset the countercyclical movement of Reichsbank liabilities. MCGOULDRICK pointed out that his table 7.5 contradicts Rich's statement. He found that high-powered money grew on average by 3.8 percent a year during upswings but rose by 5.0 percent on average during business-cycle downswings. In other words, high-powered money moved countercyclically. Specie alone did have a mild procyclical movement, its annual growth rate averaging 6.7 percent during upswings but only 5.2 percent during downswings. Putting the two components of highpowered money together indicates that the countercyclical movements in Reichsbank liabilities more than outweighed the mild procyclical movement in specie in the hands of the public. Thus, total high-powered money moved countercyclically.

8

Swedish Experience under the Classical Gold Standard, 1873-1914 Lars lonung

8.1

Introduction

Sweden adopted gold as the basis for its monetary system in 1873. The Swedish currency, the krona, remained tied to gold at a fixed rate for about forty years until the outbreak of World War I. The prewar gold standard represents the longest lasting monetary regime in Swedish history from the establishment of the Riksbank in the second half of the seventeenth century to the present. Actually, World War I did not mark the demise of the gold standard in Sweden. During the war the krona was intermittently below, above, and at par with the prewar gold rate. In 1922 Sweden was the first country in Europe to return de facto to gold at the prewar parity and in 1924 de jure. The interwar gold standard lapsed in September 1931 when Sweden followed Great Britain's departure from gold. Swedish experience stands out in four respects. First, the country's economic growth, if not the highest, certainly was one of the highest of any country in Europe during the classical gold standard era from the 1870s to the outbreak of World War I. In the second half of the nineteenth century, Sweden began a process of industrialization financed Lars Jonung is associate professor of economics at Lund University, Sweden. The author is greatly indebted to Ingemar Dahlstrand at the Lund computer center who has skillfully and patiently worked with the data base to produce the figures and tables. The author benefited from generous comments by Arthur Bloomfield, Michael Bordo, Thorvaldur Gylfason, Lennart Jorberg, Charles Kindleberger, Frank Kirwan, and Peter Lindert. He would also like to thank seminar participants at the Institute for International Economic Studies, University of Stockholm and at the Department of Economic History at the University of Lund. A great debt is owed to Frank Kirwan who has skillfully improved the author's English. Financial support from the Bank of Sweden Tercentenary Fund is gratefully acknowledged. This paper builds upon data and results from an ongoing study of Swedish monetary history.

361

362

Lars Jonung

by considerable capital imports which transformed a basically agrarian country into an industrialized society. Second, this rapid economic development was accompanied and supported by a swift expansion of the Swedish monetary system. Commercial banking spread, measured by the number of banks, number of branch offices, and volume of deposits. Third, Denmark, Norway, and Sweden formed a monetary union in the 1870s based on a common gold currency, the Scandinavian krona. Eventually Scandinavian bank notes were accepted at par by the three central banks. In this way the Scandinavian countries maintained a close monetary association during the gold standard era. Finally, comprehensive monthly data covering the Riksbank and the complete commercial banking system are available from the early 1870s. The data make possible a detailed examination of Swedish monetary experience during the prewar gold standard era. 8.2 Issues to be Examined The Swedish gold standard (1873-1914) has not previously been exfrom a monetary standpoint applying received economic theory.1 The purpose of this study is to fill the gap.2 The Swedish monetary system and the behavior of money, velocity, prices, and real income during the pre-1914 gold standard era are briefly described in section 8.3. This account also serves as background for an examination of the factors behind gold holding by the nonbank public, the commercial banking system, and the Riksbank (section 8.4). The domestic distribution of gold is intimately related to the role of gold and the role of the Riksbank in the money-supply process. That process is the framework for a discussion of the following questions in section 8.5: (1) Did the Riksbank follow "the rules of the game", i.e., did it respond to a gold outflow (inflow) by reducing (expanding) its domestic lending, or did the bank sterilize gold flows by offsetting operations? (2) Did the ·Riksbank use gold or other currencies like the British pound and the German mark for settling international transactions? (3) What role did short-term and long-term capital flows play in Sweden during the gold standard era? Another important issue concerns the way the' domestic economy adjusted to international and domestic disturbances. Two major theories are of special interest in this context: first, the price-specie-flow mechanism according to which arbitrage in gold-gold flows-maintained balance-of-payments equilibrium and long-run common-internationalprice-level behavior, and second, the monetary approach to the balance of payments according to which arbitrage in traded goods--eommodity rather than gold flows-played a central role in the adjustment mechanism. The law of one price-a corollary of the monetary approach-serves to am~ned

363

Swedish Experience, 1873-1914

establish close international covariation of prices both in the short and in the long run. According to the monetary approach, the Swedish money supply was determined by the demand for money expressed by surpluses or deficits in the balance of payments. Swedish evidence on these issues is considered in section 8.6. Finally, the results of the study are summarized in section 8.7. An appendix contrasts the views of Wicksell and Cassel on pre-World War I price experience. 8.3

8.3.1

The Swedish Economy during the Pre-1914 Gold Standard Era

The Monetary System

The Riksbank, founded in 1656, holds the central position in Swedish monetary history. Shortly after the establishment of the bank, due to an overexpansion of its note issue, it was taken over by Parliament, the Riksdag, which has since owned and controlled it. As a consequence, the bank was a source of subsidized financing for various favored activities of the Parliament during most of the eighteenth and nineteenth centuries. The bank has supplied the Swedish public with bank notes since its establishment, so at an early stage of Swedish monetary history the public became accustomed to paper currency. Following bankruptcies in the 1810s of a number of private commercial-bank undertakings, the Riksbank became the sole bank in existence in Sweden in the 1820s. A royal proclamation in 1824 authorized the establishment of privately owned and operated banks organized as joint partnerships with unlimited liability. The partners had to assume full financial responsibility for the conduct of the bank-"one for all and all for one." A charter issued by the Crown was required to start a new bank. The charter clearly stated that the new bank should not "expect to receive any support from public funds." Experience with bank runs and failures of private banks led the government to shun financial involvement with any new private banks. The law of 1824 became the legal foundation of the Swedish commercial banking system. The first commercial bank, chartered in 1830, immediately began to issue non-interest-bearing certificates of deposit of low denominations, payable on demand to the bearer, i.e., private bank notes. The notes competed with Riksbank notes even though the law apparently prohibited private note issues. No legal steps against the private notes, however, were taken. Instead additional private commercial banks were chartered and established. Due to usury laws the private banks relied primarily on note issues and kept deposit activities to a minimum. The usury laws were repealed in the 1860s and private banks thereafter could compete freely for funds. At the same time, new legislation authorized the establishment of joint-stock banks (aktiebanker) that

364

Lars Jonung

were not allowed to issue notes. That right was reserved for banks with unlimited liability, the enskilda banks. After these legislative changes in the 1860s, the Swedish commercial banking system underwent rapid growth. The enskilda banks, the jointstock banks, and the Riksbank competed in the market for deposits, and the enskilda banks and the Riksbank competed in the market for notes. The private banks were successful in both markets, particularly in the market for notes, in spite of strong political pressures to limit use of the notes, manifested in various attempts by the Riksbank to reduce the competitiveness of the private notes and to strengthen the position of the Riksbank, the bank of the Riksdag.3 Sweden enjoyed a highly developed and efficient banking system prior to World War I. The system's expansion occurred primarily during Sweden's adherence to the prewar gold standard (see table 8.1). The number of commercial banks almost trebled in the period 1871-1913. New branch offices were established by the note-issuing banks as a way of increasing the circulation of their notes, so the number of commercialbank offices increased from 141 to 630. The number of accounts and the size of deposits per capita exhibited a striking increase. The ratio of notes to the money stock, an indicator of the development of the banking system, fell from roughly 40 percent in 1871 to about 10 percent at the outbreak of World War I. In an international comparison of banking and economic growth, Sandberg (1978, p. 680) concluded that no other country, except possibly Switzerland, "could match the Swedish system" in terms of efficiency and growth during that period. The growth of the commercial banking system was also accompanied by a rapid expansion of the savings bank system. The Riksbank, basically a commercial bank during the entire nineteenth century, developed gradually into a central bank, partially due to its inefficiency in the former role. After a long political debate and various government investigations, the Riksbank obtained a monopoly Growth of Commercial Banking in Sweden, 1871-1913

Table 8.1

1871 1880 1890 1900 1913

Number of Commercial Banks (1)

Number of Commercialbank Offices (2)

Number of Accounts Per Capita (3)

Volume of Deposits Per Capita (in kronor) (4)

Ratio of Notes to the Money Stock (percent) (5)

28 44 43 67 75

141 205 190 269 630

0.01 0.03 0.03 0.09 0.25

20 53 72 144 296

42.2 21.1 19.6 13.8 9.9

Source: Jonung 1980.

365

Swedish Experience, 1873-1914

over the note issue in 1897. Private notes were prohibited and eventually taken out of circulation in 1903. Simultaneously, the commercial banks were given the right to rediscount bills with the Riksbank. The facility had previously not been automatically available to them as the Riksbank viewed the commercial banks as competitors. Instead some large commercial banks had earlier in the nineteenth century served as central banks to smaller private banks. The Swedish monetary system during the gold standard era was a stable one. Neither the Riksbank nor the private banks experienced bank runs, thanks to several factors. Public trust in the solvency of the enskilda banks was fostered by the unlimited liability of the owners of the banks. The period 1880-1913 under the international gold standard was one of international political and economic stability, which induced stability in the monetary sector of the small open Swedish economy. The Swedish monetary system was closely connected to the international financial system during the pre-1914 gold standard because of the country's large export and import sectors. The flow of capital to and from Sweden was unrestricted. In the 1850s the Swedish government began large-scale borrowing abroad of long-term capital, primarily from Great Britain and France. Prior to World War I, about 80 percent of long-term government debt was raised abroad (Flodstr6m 1912, pp. 812-15). Several commercial banks were engaged in international financing. Other financial institutions also borrowed abroad, although a domestic capital market gradually developed. In brief, Sweden enjoyed a highly efficient monetary system during the classical gold standard era, primarily due to a legal framework that fostered competition in financial markets.4 8.3.2

Money, Velocity, Prices and Real Income

This section, based on the quantity equation in its income version, briefly describes the behavior of money, velocity, prices, and real income in Sweden during the classical gold standard era. Table 8.2 presents magnitudes of the four components of the quantity equation in selected years 1871-1913, and table 8.3, average annual rates of change of the components for the period 1871-1913 and various subperiods. The money stock (M2) increased secularly during the gold standard era by an average annual growth rate of 6 percent. It exhibited, however, substantial secular and cyclical fluctuations during this period. Relatively high growth rates were registered in the 1870s and during 1890-1913, while the rate was low during the 1880s, which was a decade of deflation. Velocity, measured by the ratio of nominal income to the money stock, consistently experienced a secular decline. The ratio fell from 6.8 in 1871 to about a third of this level in 1913 (see Jonung 1983 for a detailed analysis of this decline). The price level had a slight positive growth rate over the full period,

366

Lars Jonung

Table 8.2

Money Stock (M), Income Velocity of Money (V), the Price Level (P), and Real Income (y): Selected Years, 1871-1913 y

M

1871 1880 1890 1900 1913

(million kronor) (1)

V

P

(ratio) (2)

(1913 = 100) (3)

(million 1913-kronor) (4)

147.9 308.5 430.3 856.7 1854.8

6.77 4.32 3.51 2.64 2.23

84.0 90.0 83.0 88.0 100.0

1192 1482 1821 2569 4128

Sources: Col. (1), Jonung 1975, annual averages of end-of-month data for sum of the public's holdings of Riksbank and private bank notes and demand and time deposits; col. (2), Johansson 1967, table 55 divided by Jonung 1975; col. (3), Johansson 1967, table 55 divided by table 56; col. (4), Johansson 1967, table 56. Rates of Growth of Money Stock (M), Income Velocity of Money (j): Selected Periods, 1871-1913

Table 8.3

(V), Prices CP), and Real Income

M Period

V

Y

P

(average annual percentage rates of change) (1) (2) (3) (4)

1871-1913

6.0

-2.6

0.4

3.0

1871-1879 1880-1889 1890-1899 1900-1913

7.4 3.5 6.5 5.9

-4.7 -2.5 -2.7 -1.3

0.1 -1.2 0.5 1.0

2.6 2.1 3.2 3.6

Secular deflation: 1874-1896

3.0

-1.9

-1.1

2.2

Secular inflation: 1897-1913

7.2

-2.3

1.4

3.5

Source: Jonung 1976. Note: Growth rates are continuously compounded.

with the rate of decline during the secular deflation of 1874-96 more than offset by the rate of rise during the secular inflation of 1897-1913. These secular movements closely coincide with long-term international price movements (see the Appendix on the views of Wickselland Cassel on the determinants of changes in the secular price level). Sweden experienced a rapid transformation from an agrarian economy to an industrialized country during the prewar gold standard era. Gross domestic product rose from a level of 1192 million kronor in 1871 to 4128 million kronor in 1913 measured in 1913 prices (see table 8.2), representing an average annual real growth rate of about 3 percent-among the

367

Swedish Experience, 1873-1914

highest per capita growth rates in Europe during the period, if not the highest of any European country (Bairoch 1976). Sweden was basically an agrarian country in the second half of the nineteenth century. About 80 percent of the total population was employed in agriculture in 1870. By the turn of the century the share of industry in gross domestic product surpassed that of the agricultural sector, although at the outbreak of World War 1,60 percent of the total population was still employed in agriculture (see Johansson 1967, table 58). The process of growth and industrialization was primarily export-led. Lack of developed domestic markets and poor domestic communications hampered domestically oriented industries relative to export-oriented ones in the early stages of industrialization (Jorberg 1970, p. 81). However, beginning in the 1890s, domestic industries experienced rapid expansion, partly owing to large-scale building of railroads. The main export items were timber products, paper and pulp, engineering products, and iron and steel. Swedish export of raw materialsrepresenting 40-50 percent of Swedish exports during the prewar gold standard period-met a growing demand from countries beginning to industrialize, so terms of trade improved for Sweden under the gold standard (Jorberg 1970, p. 88). Exports represented roughly 20 percent, imports a slightly higher share of gross domestic product. Great Britain was by far Sweden's most important export market, absorbing more than 50 percent of Swedish exports in the 1870s and about 30 percent prior to World War I. Most Swedish imports of goods originated in Great Britain and Germany-the British share fell from 30 percent in the 1870s to 24 percent priorto 1914, while the German share rose from 23 percent to 34 percent (Historical Statistics of Sweden 1972, tables 5.1, 5.2). Nearly all Sweden's foreign trade was European. 8.4 8.4.1

Use of Gold and the Legal Framework Legal Framework

In the 1860s representatives from Denmark, Norway, and Sweden discussed the establishment of a monetary system based on gold that would replace the silver standard and encompass the three countries. In May 1873, a monetary union was formed by Sweden and Denmark. Norway joined in October 1875. The Scandinavian Monetary Union introduced a common currency unit, the Scandinavian krona, which equaled a 1/2480 kilo of gold. Consequently, the following gold-parity rates were established: one British pound = 18.16 kronor, 100 Reichsmark = 88.89 kronor, and later, one U.S. dollar = 3.73 kronor. Gold coins minted according to the Scandinavian monetary agreement were

368

Lars Jonung

legal tender in the three Scandinavian countries irrespective of where the coin had been minted originally. Common subsidiary silver and copper coins were also introduced and made legal tender within Scandinavia. The founding of the Scandinavian Monetary Union per se did not result in significantly closer monetary cooperation than would have been the case if each country had adopted the gold standard by itself. A common currency system to a minor extent reduced transaction costs and had symbolic value attesting to Scandinavian cooperation. However, because of the union the three central banks accepted each other's notes at par as part of a clearing system they established. The Bank of Sweden informally accepted Danish and Norwegian notes at par in the 1870s. The Bank of Denmark and the Bank of Norway executed a formal agreement to this end in 1901. Consequently a considerable exchange of notes occurred among the three countries with notes partially replacing gold and other foreign reserves as international means of payment.5 During the entire pre-1914 history of the Scandinavian Monetary Union, the parity rates remained the officially registered rates, while other exchange rates like that of the pound and the Reichsmark fluctuated above and below par.6 The convertibility of the notes of the Riksbank into gold was guaranteed by the Swedish constitution. Laws pertaining to the gold standard could only be changed by two identical decisions by Parliament with an election in between. The constitutional status of the gold standard is explained by Sweden's earlier turbulent monetary history during which at several times Riksbank notes became inconvertible. The constitutional guarantee turned out, however, to be an ineffective protection for the gold standard. When World War I broke out, the Riksbank decided unilaterally to make its notes inconvertible. This unconstitutional stepwhich was never challenged-ended the prewar gold standard in August 1914 (Heckscher 1926, pp. 13-14). 8.4.2 Use of Gold by the Public The longest lasting metallic standard in the history of the Riksbank was the pre-World War I gold standard (Jonung 1976). Despite the roughly forty-year link between the krona and gold, there was never any significant domestic monetary use of gold by the Swedish nonbank public. According to contemporary commentators on the Swedish monetary system before World War I-Rosenberg (1878, p. 44), Davidson (1896, p. 25), Wicksell (1896, p. 38; 1904, p. 86), and Heckscher (1926, p. 11)-gold coins or gold in other forms were employed infrequently as a means of payment, in sharp contrast to prevailing conditions in the United Kingdom. Although there is wide agreement that gold coins were rarely used by the Swedish public, few quantitative measures of the extent to which they

369

Swedish Experience, 1873-1914

were held exist. According to Davidson (1896), a study by the Ministry of Finance estimated the public's possession of gold coins in 1885 at around 5.5 million kronor. The figure corresponds to about 8 percent of the total note issue of the Riksbank and the private banks, and to about a tenth of a percentage point of the Swedish money stock defined as the sum of the public's holdings of notes and commercial-bank demand and time deposits in that year. These figures suggest that the volume of gold held by the public for monetary use was insignificant relative to the money stock. Consequently, gold coins may safely be excluded from the definition of the Swedish money stock during the pre-1914 gold standard era. Several factors contributed to the public's small holdings of gold coins compared to the volume of notes and deposits. The most important one was the policy of the monetary authorities of supplying the public with notes of denominations lower than those of gold coins. According to the arrangements of the Scandinavian Monetary Union, gold was initially minted in Sweden in two denominations: ten and twenty kronor. The coins were of such high denominations that they were seldom used to settle transactions involving small sums, as there existed more suitable means of payment of lower denominations, specifically bank notes as well as subsidiary coins of silver and copper. The Riksbank issued a one-krona note prior to 1875, but from then on the five-kronor note was the smallest note denomination supplied. The ten- and twenty-kronor gold coins were clearly inferior to the five- and ten-kronor notes for transaction purposes.7 Furthermore, long before the adoption of the gold standard, the Swedish public became well acquainted with the use of notes. Riksbank notes had been circulating in Sweden since the second half of the seventeenth century and the note-issuing commercial banks had been actively promoting their notes since the 1830s. Consequently, notes issued by the Riksbank and the private banks were more familiar to the public as means of payment than gold coins. Rosenberg (1878, p. 44) stated that "the Swedish public reluctantly uses the gold coins finally introduced." Apparently, the reluctance persisted throughout the gold standard period. For several reasons, it was rational for the monetary authorities to pursue a policy of minimizing the public's holdings of gold. First, the management by the Riksbank of its gold stock was facilitated. To the extent that the public did not demand any gold, the possibility of sudden increases in the internal demand for gold remained small. Consequently, the bank's stock of gold could be maintained at a lower level than were gold in greater domestic use. Instead, the bank could concentrate on meeting fluctuations in the external demand for gold. Second, small public holdings of gold allowed the bank to invest in interest-bearing assets rather than in gold, increasing its profit. Most likely such considerations had some influence on the policy of the bank

370

Lars Jonung

during the gold standard era, as the profit motive-adopting the term frequently used in the Swedish monetary debate during World War I-was of considerable importance in guiding the bank's affairs. Third, a policy to increase the domestic supply of gold coins would also have burdened the Swedish balance of payments in the nineteenth century, as Sweden was not a gold producer. Sweden as a rule had a deficit on current account prior to 1914, reflecting foreign-capital inflows to finance the domestic industrialization process. Thus, attempts to encourage the use of gold coins would possibly have encroached on the borrowing opportunities of other sectors of the Swedish economy.8 Commercial banks that issued notes also had two incentives to withdraw gold coins from circulation: first, they increased the circulation of their notes by eliminating a potential competitor, and / second, they obtained an asset that satisfied the reserve requirements stipulated by law. The successful attempts by the note-issuing commercial banks to drive Riksbank notes out of circulation were based on the same motives.9 There is generally a strong complementarity between the use of an asset as a means of payment and as a store of value, that is, if an asset is demanded as a medium of exchange, that increases its employment as a store of value. However, in Sweden gold evidently did not serve either of these two functions, probably because gold was not adopted as a commonly used medium of exchange. The public had great confidence in the solvency of the private banking system, and the political situation remained stable, so there was little reason to use gold as a store of value. In brief, the Swedish public's holdings of gold coins were extremely small during the classical gold standard era. The Swedish gold standard system was thus far from a "pure" gold-specie standard or gold-coin standard where gold coins circulated as currency in significant domestic monetary use. 8.4.3

Use of Gold by Commercial Banks

As a consequence of the replacement of the silver standard by a gold standard in 1873, the legislation pertaining to the reserve requirements of the note-issuing commercial banks, the enskilda banks, was changed accordingly in the banking law of 12 June 1874. The law of 1874 (article 26) regulated the maximum volume of notes that a private bank could issue, the banksedelutgivningsriitt, literally the right to issue notes. The volume was equal to the sum of a number of commercial-bank assets, including holdings of bonds and gold. The difference between the banksedelutgivningsriitt and the actual note issue in circulation was termed obegagnad sedelutgivningsriitt, literally the unused right to issue notes. The unused right to issue notes, documented in the monthly bank reports for each enskilda bank, was substantial throughout the prewar gold standard period, constituting about 25 to 60 percent of the actual

371 Table 8.4

Swedish Experience, 1873-1914 Note Issues and Volume of Unused Right-to-Issue Notes (obegagnad sedelutgivningsriitt) of the Riksbank and Note-Issuing Commercial Banks (enskilda banks): Selected Years, 1875-1910

Notes in Circulation Enskilda Riksbank Banks (million kronor) (1) (2)

1875 1880 1885 1890 1895 1900 1905 1910

37.7 37.6 37.2 42.8 50.6 63.6 163.1 189.3

60.3 47.9 50.1 59.6 60.0 79.4

Share of Riksbank Notes of Total Circulation (percent) (3) 38.5 44.0 a 42.5 41.8 45.8 44.5 100.0b

Unused Right to Issue Notes as Share of Circulation Enskilda Riksbank Banks (percent) (4) (5)

17.2 39.6 20.6 56.5 66.2 105.7

25.1 53.7 49.3 32.7 37.7 36.6

100.0

Source: Series, calculated by the author from the monthly bank reports, are annual averages of end-of-month figures. aThe increase in the market share of Riksbank notes between 1875 and 1880 was due to legislation that prohibited the issue of five-kronor notes by enskilda banks, leaving the Riksbank as the sole supplier of the denomination, the lowest available. bprivate bank notes were taken out of circulation in 1903 when the Riksbank obtained a monopoly over the n.ote issue.

volume of private notes in circulation (see table 8.4). Consequently, banking laws concerning the maximum size of the private note issue did not effectively restrict the private supply of notes. Because the legal reserves behind private notes included bank assets that did not constitute base money, private banks 'could, within wide limits, increase their banksedelutgivningsriitt without obtaining base money, that is, gold and Riksbank notes. lO Private note-issuing banks had thus little need to demand gold for legal reserve requirements against their circulation during the pre-1914 gold standard period. The banking law of 1874 (article 28) made private bank notes redeemable in gold. Such notes, however, were redeemable on demand in legal tender gold coins only at the head office of the note-issuing bank. In a case where payment in gold was refused, the issuing bank was obliged to pay a 6 percent rate of interest to the note holder until gold redemption. In case of prolonged nonredemption, the enskilda bank was subject to loss of its government charter and thus extinction. As Sweden was sparsely populated and lacked a well-developed communications network, the stipulation that notes were only redeemable at the head office and not at branch offices was clearly a method of reducing potential

372

Lars Jonung

strains on enskilda banks in case of a loss of public confidence in the redeemability of their notes. As a consequence of the adoption of the gold standard in 1873, private note-issuing banks started to purchase considerable quantities of gold during the decade, increasing their holdings of gold from virtually a zero level to 8.5 million kronor in 1876. The volume of commercial-bank-held gold remained constant around 7 to 8 million kronor until the mid-1890s, then rose to a peak of 9.4 million kronor in 1900 (see figure 8.1). Private note issues were gradually withdrawn from circulation during the period 1900-1903, and commercial banks coincidentally sold their gold to the Riksbank. As a result, by 1905 they held practically no gold. The note issue of the private banks expanded from 48 million kronor in 1880 to 80 million kronor in 1900 (see table 8.4). Since their gold holdings remained practically constant in the 1880s and 1890s, the ratio of gold to notes of the enskilda banks declined from about 19 percent in 1879 to about 10 percent in 1900 (see figure8.2).

Million kronor

125r--------------------------100 80

60

50

40 30

Riksbank

-

-

-

_" -

-

-

-

-

-

-

_;I' - ' \

\

Enskilda banks

\

,

\

\ \ \ \

1890

Fig. 8.1

1895

1900

1905

1910

Year

Gold reserves of the Riksbank and of the note-issuing private commercial banks, 1870-1914. As commercial banks that did not have the right to issue notes held an insignificant amount of gold, these holdings are not included here. Sources: Riksbank 1870-74: Sveriges Riksbank 1668-1924 1931, 5:58, 60; for 1870, average of 30 June and 31 Dec.; for 1871, average of June-Dec. inclusive; for 1872-74, annual averages of endof-month data. Riksbank and enskilda banks, 1875-1914: monthly bank reports, annual averages of end-of-month data. Before 1875, gold assets of the enskilda banks are not shown separately from silver assets.

373

Swedish Experience, 1873-1914

The absence of reports that private banks could not redeem their notes into gold or that runs on the commercial banking system occurred because the public tried to convert deposits and notes into gold is confirmed by the constant level of the gold holdings of the enskilda banks (see figure 8.1). Bank runs would have caused significant fluctuations in the curve. As argued previously, the major reason for the fairly constant level of gold holdings by the private banks in the face of a rising note issue was the fact that the Swedish public never demanded gold for domestic monetary use-private and Riksbank notes were a superior means of payment. Unlike the enskilda banks, the joint-stock banks-eommercial banks that did not issue notes-were not subject to gold-reserve requirements. Their holdings of gold were consequently negligible. Note-issuing commercial banks were not obliged to hold gold against demand and time deposits. Thus, the commercial banking system held gold solely as cover

Percent 60 55 50 45

Riksbank notes 40 35

30 25· 20

Enskilda bank notes

............

15

_//

.... - - - - - - ............ ........

/

10

-

..................

5

o

1875

Fig. 8.2

_----

....

_- _........

Swedish money stock

1880

1885

1890

- .....

........................................-

1895

1900

1905

1910

Year

Gold-reserve ratios of the Swedish monetary system, 18751914. Riksbank notes include public and commercial-bank holdings. Commercial-bank notes include holdings by the public, the Riksbank, and commercial banks that did not have the right to issue notes. Gold reserves of the Swedish monetary system include Riksbank, enskilda-bank, and joint-stock-bank holdings; the latter are not included here, however, since they were insignificant before 1914. After the elimination of private notes in 1903, commercial banks held practically no gold. Sources: For gold reserves of the Riksbank and the enskilda banks, see figure 8.1; for note issues, see table 8.4; and for money stock, see table 8.2, col. 1.

374

Lars Jonung

for bank-note issues and sold its gold to the Riksbank when private notes were eliminated at the turn of the century. To sum up, legal gold-reserve requirements did not represent an effective restriction on the supply of notes and deposits by the commercial banking system during the gold standard era. Indeed, the holdings of gold relative to notes in circulation by the note-issuing banks declined during the 1880s and 1890s, while throughout the period the enskilda banks had the right to issue a far larger volume of notes than they actually did. The gold reserves of the enskilda banks were stable, reflecting the absence of monetary disturbances in the form of sudden domestic or international gold outflows. 8.4.4 Gold and the Reserve Requirements of the Riksbank The rules regulating the relationship between the note issue of the Riksbank and its gold reserves were modified several times during the prewar gold standard period. The basic reason for the changes was a desire to accommodate the long-run growth of the note supply. Prior to 1872 the note issue of the bank was based on its reserves of silver, gold, foreign assets ll plus the kontingent, i.e., a fiduciary reserve of 30 million kronor. Silver was the dominant metallic asset. The sum of these assets represented the sedelutgivningsriitt, i.e., the maximum amount of notes that the Bank was allowed to supply (see table 8.5). As a consequence of the adoption of the gold standard in 1873, Riksbank notes were made redeemable in gold at the head office in Stockholm. Gold thus took the place previously held by silver (see figure 8.1). The fiduciary reserve was raised to 35 million kronor in 1879 when private five-kronor bank notes were prohibited, leaving the Riksbank as the sole supplier of the denomination. The reserve requirements were changed again in 1887. The fiduciary reserve was increased to 45 million kronor and covered by a secondary reserve requirement stipulating that the fiduciary reserve should correspond to the sum of easily salable bonds and bills. The Bank was also prohibited from holding less than 15 million kronor in gold for any prolonged period of time. When private notes were abolished, the reserve requirements of the Riksbank were changed. In 1899, the fiduciary reserve was increased from 45 to 100 million kronor. The use of silver as a legal reserve asset was discontinued and a minimum level of 25 million kronor was stipulated for gold holdings. Due to a considerable expansion of the note issue, the rules were again altered in 1901. The end result was a complex system based on a combination of a fiduciary reserve and a reserve ratio, the latter strengthening the link between gold holdings and the note issue. In 1913 another revision of the rules expanded still further the right to issue notes.

375

Swedish Experience, 1873-1914

The note issue of the Riksbank was thus determined by three types of reserves: (1) specie, i.e., gold and silver, (2) foreign assets as defined by the reserve rules, and (3) the fiduciary reserve. Table 8.5 shows that the fiduciary reserve was most important and gold second in importance relative to the note issues during the prewar gold standard period. Due to the size of the fiduciary reserve, the right to issue notes was considerably larger than the volume of Riksbank notes in actual circulation. The note supply was thus not effectively restricted by required reserve holdings, including those of gold. Gold and foreign assets were initially regarded as perfect substitutes by the reserve requirements. Consequently, the bank tried to maximize its holdings of interest-bearing foreign assets at the expense of gold. For this reason, rules pertaining to minimum holdings of Table 8.5

Legal Reserves for the Riksbank's Note Issue: Averages for Selected Periods of Unchanged Reserve Regulations, 1860-1912 (million kronor) 1860-68 1873-87 1887-98 1899-1912 (coefficient of variation in parentheses)

1. 2. 3. 4. 5. 6. 7. 8. 9.

Gold Silver Foreign assets Fiduciary reserve (kontingent) Right to issue notes (sedelutgivningsriitt ) Riksbank notes in circulation Unused right to issue notes (sedelreserv) Supplementary coyerage (supplementiir tiickning) Gold and silver as share of note issue

0.3 10.9 6.3 30.0

(.53) (.13) (.34) (.00)

12.3 5.8 7.6 32.7

(.22) (.53) (.42) (.08)

19.7 2.5 9.5 44.4

(.25) (.19) (.41) (.04)

21.6 (.31) 119.8 (.18)

63.3 (.30)

47.4 (.07)

58.3 (.10)

76.2 (.11)

205.4 (.23)

29.6 (.16) 4.5 (.83)

35.9 (.13) 10.3 (.38)

47.4 (.16) 28.8 (.14)

144.7 (.35) 60.7 (.22)

55.3 (.13)

144.3 (.24)

38%

50%

47%

44%

Sources: Lines 1-8, Sveriges Riksbank, 1668-1924, 1931,5: pp. 58-71; 1860-68, end-of-year

data; 1873-87 and 1887-98, annual averages of end-of-month data; 1899-1912, annual averages of weekly data; line 9, Source in figure 8.1 for gold and silver divided by line 6. Notes: The selection of time periods in the table is based on dates of major changes in reserve regulations (see text and Brisman 1931). The years 1869-72, the transition from a silver to a gold standard, have been excluded. Line 3, foreign assets that were not held as legal reserves against notes are excluded. These assets, notably foreign bills and foreign bonds, were considerably larger than foreign assets shown in line 3. Lines 1-4 sum to line 5 (except for rounding differences)-the sedelutgivningsriitt, that is, the maximum volume of notes the Riksbank was allowed to circulate. Line 7, the sedelreserv, the unused right to issue notes, is the difference between line 5 and line 6, except for 1860-68 and 1873-87, when reserve requirements covered other short-term Riksbank liabilities, like demand deposits, and bank postal bills. Line 8, the supplementary coverage, shows the fiduciary reserve, i.e. the volume of notes the Riksbank was allowed to circulate without the backing of any assets held.

376

Lars Jonung

gold were introduced, and a ratio system between gold and the note supply was eventually established. Figure 8.1 shows that the Riksbank's holdings of gold expanded rapidly in the early 1870s reaching a peak of about 20 million kronor in 1874. From then on until 1880 they declined. A slow but steady expansion occurred''in the 1880s and 1890s. A period of very rapid long-run growth started at the turn of the century when private bank notes were taken out of circulation. By then all monetary gold in Sweden was concentrated in the Riksbank, where it was held for the management of Sweden's international transactions. To sum up, the legal rules pertaining to gold and other assets did not restrict the note supply of the bank during the prewar gold standard period. The link between the gold holdings of the Riksbank and its note supply was loose due to the flexibility of the reserve rules. The sizable fiduciary reserves of the bank gave the Riksbank considerable short-run autonomy in reacting to temporary disturbances, and the authorities' propensity to change the reserve rules provided substantial long~run autonomy.

8.5 Gold and the Money-Supply Process This section examines the role of gold in the Swedish money-supply process both in the long and the short run during the pre-1914 gold standard period. In addition, Swedish evidence on the key currency issue is presented. 8.5.1

Long-Run Movements in Gold and Money

The role of gold in the money-supply process is examined here using a framework pioneered and applied to U.S. monetary history by Friedman and Schwartz (1963) and Cagan (1965). The sources of secular and cyclical growth of the Swedish money stock, specifically the contributions of gold and foreign assets, can thereby be quantified. Currency (C) is defined as the nonbank public's holdings of Riksbank notes, and deposits (D) as the sum of commercial bank notes and demand and time deposits. The money stock (M) = C + D. The monetary base (B) is set equal to Riksbank notes held by the public (C) and base money reserves (R) held by the commercial banking system, that is, Riksbank notes and gold held by the note-issuing private banks. The money stock is related to three aggregates: the monetary base, the currency ratio, and the reserve ratio. (1)

M=

1 C+ R -_ C.R -

M

D

M D

B or

377

Swedish Experience, 1873-1914

(2)

M= m -B,

where m is the monetary-base multiplier. By expressing (1) in terms of rates of changes, the sources of secular growth of the Swedish money stock can be calculated. During the period 1871-1913 the Swedish money stock grew at an average annual rate of 6 percent, of which in absolute terms the monetary base contributed 4.5 percent, the currency ratio 0.6 percent, and the reserve ratio 1.1 percent (see Jonung 1975). The relative contribution of the monetary base was thus 75 percent while the base money multiplier accounted for the remaining 25 percent. Secular decline in the currency ratio and the reserve ratio produced the positive contributions of the asset ratios. The currency ratio fell from 14.0 percent in 1871 to 9.9 percent in 1913. The corresponding figures for the reserve ratio are 10.0 percent and 2.3 percent, respectively. The monetary base accounted for the major share of the secular expansion of the Swedish money stock under the prewar gold standard. The effects of gold movements on the monetary base and thus on the money stock can be derived from a breakdown of the balance sheet of the Riksbank. The Riksbank was the major holder of gold in the Swedish economy. Riksbank notes constituted the most important component of the monetary base. Their share was around 75 to 80 percent in the 1880s and 1890s. After the elimination of private notes in 1903 and the sale of commercial-bank gold to the Riksbank, the note issue of the Riksbank was identical to the monetary base. Gold held by the commercial banks, the second major component of the monetary base, was practically constant in the period 1875-1900 according to figure 8.1 and is therefore ignored in the following computations. Changes in the volume of Riksbank notes in circulation are equal to the sum of the changes in the following components of the balance sheet of the Riksbank: the volume of specie (G), primarily gold but also silver, net foreign assets (FA), and net domestic assets (DA), i.e., loans minus deposits at the Riksbank. When the note issue is set equal to the monetary base, expression (3) is obtained: (3)

B=G+FA+DA.

The sum of specie and foreign assets (G) + (FA) is the foreign source component of the monetary base while domestic assets are the domestic source component. Equation (3) is the basis for the calculations in table 8.6 displaying the contributions of specie, net foreign assets, and net domestic assets to the growth of the Riksbank note issue during the period 1871-1913 as well as during a number of subperiods. According to table 8.6, specie, i.e., primarily gold, accounted for 46 percent of the relative contribution, foreign assets for 52 percent, and the domestic source component includ-

4.8 1.4 4.6 9.1

Annual Growth Rate of Riksbank Notes (percent) (1) 2.2 0.6 4.4 3.6

(2)

Specie

2.5 -0.3 4.0 4.1

Foreign Assets (percent) (3)

Absolute Contributions

(6) 52 -21 87 45

(5) 46 43 96 40

(4) 0.1 1.1 3.8 1.4

Foreign Assets

Specie

Relative Contributions Domestic Assets

Contributions to the Growth Rate of Riksbank Note Issue, 1871-1913

2 79 -83 16

(7)

Domestic Assets

Source: Based on Sveriges Riksbank, 1668-1924 1931, 5: pp. 24-31. Notes: Growth rates are continuously compounded. Cols. (2)-(4) growth rates are weighted by the average ratio of the individual components to the Riksbank note issue at the initial and terminal years of each subperiod. Cols.(5)-(7) express the rates in (2)-(4) as relative to 100. Specie includes gold and silver and other items such as coins and private bank notes. No correction was made for the latter since these items were fairly small. Foreign assets are the sum of net deposits at foreign banks, foreign bills, foreign bonds, and foreign notes. Domestic assets are the sum of domestic bonds, bills, and loans, minus deposits of the Riksbank (i.e., giroriikning, upp- och avskrivningsriikning, depositionsriikning, and postremissviixlar) minus capital items of the Riksbank.

1900-1913

189~1900

188~1890

1871-1913

Period

Table 8.6

379

Swedish Experience, 1873-1914

ing capital items for 2 percent. Although both gross domestic assets and gross domestic liabilities increased during the gold standard period, the net volume made an insignificant contribution to the expansion of the note issue. Subperiods of the gold standard show differences in growth patterns, but specie holdings expanded throughout while foreign assets declined in the 1880s and domestic assets in the 1890s. Consolidation of specie and the net foreign assets shows that the foreign source component of the monetary base accounted for practically all the secular growth of the note issue and thus of the monetary base. This statement implies that specie, including gold held by the commercial banks and foreign assets of the Riksbank, accounted, as noted above, for about 75 percent of the growth of the Swedish money stock during 1871-1913, the decline of the currency and reserve ratio accounting for the rest of the expansion. The close association between the long-run growth of Riksbank notes and the secular increase in gold is also seen in figure 8.2 which displays the ratio of gold to the notes of the Riksbank. The ratio has an average value of around 40 percent for the entire gold standard period 1873-1914. The long-run ratio of gold to the Swedish money stock in the same figure exhibits a surprisingly constant level of around 5 percent. This constant level implies that an increase in the money stock was associated with an identical relative expansion of the monetary gold stock within Sweden, ignoring the issue of causality. Sweden experienced substantial deficits on current account throughout the prewar gold standard period. Except in a few years, exports were smaller than imports. Since inflows of gold and foreign assets were the major factor behind the growth of the Swedish money stock during the period, long-term borrowing from abroad is of central importance in explaining the expansion of the money stock. The entries in table 8.6 support this conclusion. Deficits on the trade account would otherwise have caused a continuous outflow of specie and foreign assets from the Riksbank. The accumulation by the Riksbank of growing foreign re-' serves was based on the import of foreign capital. Long-term borrowing was primarily arranged by the Swedish government during the gold standard period by issues of state debt in Germany, France, and the United Kingdom. The largest volume of capital imports occurred in the 1880s and between 1898 and 1911.12 Swedish municipalities and mortgage banks also borrowed abroad. In addition, direct investments by foreigners and foreign firms contributed to the import of longterm capital. By 1908 Sweden's long-term debt roughly corresponded to 40 percent of the national income with the National Debt Office liable for close to 40 percent of the country's total foreign long-term indebtedness (Fleetwood 1947, pp. 39-42). In short, long-term capital imports during the gold standard period were the central factor accounting for the

380

Lars Jonung

secular increase in gold and foreign assets, and thus for the long-run growth of the Swedish money stock. 8.5.2 Short-Run Movements in Gold and the Policy of the Riksbank On a year-to-year basis, the monetary base accounted for a major share of movements in the money stock under the gold standard although the contributions of the asset ratios were larger than in the long run (see Jonung 1975). The result implies that the Riksbank's role was central for short-run changes in gold, foreign assets, notes (which constituted most base money), and the money stock. The Riksbank, while serving as the bank of the Parliament, during the nineteenth century was a mixture of a central bank and a commercial bank. Prior to the 1870s, regulations of Parliament determined the interest rates the bank charged. The rates were fixed at levels below market rates in order to subsidize favored borrowers. Changes in the discount rates were thus not a policy tool. Reserve regulations stipulated a rigid link between Riksbank notes in circulation and foreign assets of the bank. Since Sweden. was on a silver standard prior to 1873, foreign reserves consisted mainly of silver, foreign bills, and deposits at a few foreign banks. When confronting an outflow of metallic reserves and foreign assets, which occurred regularly in the 1850s and 1860s, the bank responded by reducing its loans outstanding. The bank thus reinforced an outside drain of its foreign reserves by decreasing the volume of domestic assets. This was a policy of "direct deflation," according to Brisman (1931, p. 88). Since commercial banks maintained silver and Riksbank notes as reserves, the policy of "direct deflation" caused a contraction throughout the Swedish monetary system. The overriding goal of the Riksbank prior to the 1870s was maintenance of external stability. Parliament's instructions to the bank in the midnineteenth century actually stipulated that domestic lending was to be reduced in cases of decrease in metallic reserves and foreign assets. Domestic lending could be resumed when the currency drain had been arrested. The rule reflected the belief that Riksbank notes in circulation represented a potential threat to its metallic reserves, since the notes were convertible into specie. Brisman (1931, p. 87) notes that this rule was "one of the most remarkable in the history of international banking." The policy of direct deflation was gradually modified and replaced in the 1860s and 1870s, owing in part to the influence of U.K. practice. A growing trade with the United Kingdom paved the way for changes in the instructions to the bank in the early 1860s, permitting it to hold funds in London, although the United Kingdom was not on silver.l3 Prior to this, Hamburg-Altona, a financial center for the silver standard, was the only place outside Sweden where the Riksbank had a banking connection. In

381

Swedish Experience, 1873-1914

1866 the bank was granted the right to engage in financial transactions in foreign currencies with financial centers of its own choosing. The Riksbank steadily increased its holdings of nonmetallic foreign assets. It also started to purchase Swedish government bonds. These had high internationalliquidity and could be sold abroad to obtain foreign reserves. When the gold standard was introduced in 1873, the reserve regulations were changed such that foreign bills and foreign bonds were excluded from the legally required reserves of the bank. The creation of selected foreign-reserve assets legally independent of the note issue conferred a considerable autonomy on the bank in framing its policy under the prewar gold standard. In the 1870s the bank's goals were broadened to include maintenance of domestic as well as external stability. Facing a drain of foreign reserves in the depression of 1874, the bank responded by discounting foreign bills in London and Hamburg, selling bonds abroad, and placing the proceeds with foreign bankers, thus increasing legal reserve assets and the right to issue notes. Encountering an outflow of gold in 1875, the bank obtained loans from Rothschild and the London Joint Stock Bank. In 1877, although the bank borrowed in Hamburg, the following year it felt obliged to call in loans to a limited extent-the last time in the nineteenth century such action was taken. Eventually, the bank began to rely on changes in the discount rate. From the 1880s until the outbreak of war, the Riksbank policy was to maintain constant specie holdings, employing a number of techniques to that end. First, the bank used foreign assets, both those that were counted as satisfying reserve requirements and those that were not, to prevent and offset outflows of gold. Specifically, the bank maintained holdings at foreign banks and wrote checks on them to meet the demand for international means of payment. Second, the bank changed its discount rate to protect its foreign reserves. Figure 8.5 shows that the Swedish discount rate followed the Bank of England rate fairly closely. Third, the bank relied on foreign loans to meet an increased demand for foreign reserves. Fourth, the bank paid little attention to movements' in its unused right to issue notes (sedelreserven). Reductions in this volume were more or less ignored and/or met by alterations in reserve regulations (see section 8.4.4). Fifth, the bank apparently actively discouraged outflows of gold from Sweden. According to Heckscher (1926, p. 4), this policy explains why the pound rate remained above the upper gold point of 18.27 kronor (with the gold-parity rate at 18.16 kronor) on average for several years, without significant net outflows of gold from Sweden. Finally, the Riksbank established agreements with the Bank of Denmark and Bank of Norway in the early 1880s eliminating gold as an international means of payment within the Scandinavian Monetary Union. Using these tech-

382

Lars Jonung

niques of monetary control, the Riksbank managed to stabilize in the short run its holdings of gold. That foreign-exchange operations were important in the pre-1914 policy of the bank may be seen in the annual change of gold and net foreign assets of the Riksbank, 1876-1913 (see figure 8.3). The two assets were of roughly equal average size during this period, 35.4 and 30.3 million kronor respectively, so the two series in figure 8.3 may be compared directly. The figure shows that annual movements in net foreign assets were as a rule larger than those in gold. This was the case for twenty-nine out of thirty-eight observations over the period 1876-1913. The coefficient of variation of the annual first differences is 1.5 for gold reserves but 4.4 for net foreign assets. The corresponding figure for the note issue of the Riksbank and the Swedish money stock are 1.8 and 1.0, respectively. These figures indicate considerable variability in foreign assets compared to gold, the note issue, and the money stock. Nurkse (1944) and Bloomfield (1959) investigated central-bank practices in following the "rules of the gold standard game." The rules are Percent 100

90 80

70 60

'\ ,,,

\

SO 40

, ,

30 20

,

I (

10

o -10

-20 -30

\ \ I I I I

, ''

\

, I

,'Net foreign assets 1

-40

-SO

Fig. 8.3

Annual percentage changes in gold and net foreign assets of the Riksbank, 1876-1913. Foreign assets do not include foreign bonds held by the Riksbank, since they are not shown as a separate item in the source. For end-of-year data for foreign bonds, see Sveriges Riksbank 1668-1924 1931, 5: pp. 25-31. Source: For gold and net foreign assets (sum of foreign bills and Riksbank deposits, net, at foreign banks and bankers), see fig. 8.1. Annual averages were constructed from end-of-month data.

383

Swedish Experience, 1873-1914

generally taken to mean that central banks reinforced gold flows, Le., an increase (decrease) in gold and foreign assets should be met by an increase (decrease) in domestic assets. Bloomfield, comparing year-toyear changes in foreign reserves and in gross domestic assets for eleven pre-1914 central banks, found no support for a rules-of-the-game interpretation. Foreign and domestic assets changed in the opposite more often than in the same direction. Table 8.7, comparing direction of movements in thirty-four changes of net foreign reserves and net domestic assets of the Riksbank for the Table 8.7

Annual Changes in Riksbank Net Foreign Reserves and Net Domestic Assets, 188~1913 Change in Change in Foreign Domestic Reserves Assets (million kronor) (1) (2)

1880 81 82 83 84 1885 86 87 88 89 1890 91 92 93 94 1895 96 97 98 99

2.8 .6 -6.6 1.5 4.9 -.4 .6 -1.1 6.4 -1.4 -1.7 -.9 -1.0 1.8 8.6 4.3 1.7 -0.3 1.3 18.0

4.6 -1.7 5.3 -1.3 -3.9 1.3 1.7 1.5 -4.7 2.1 2.1 -.1 .6 .7 -6.0 -0.2 2.9 3.4 3.1 -14.5

Change in Change in Foreign Domestic Reserves Assets

Sign of Covariation (3)

+ 0

0 0

+

0 0 0

(million kronor) (1) (2) 1900 01 02 03 04 1905 06 07 08 09 1910 11 12 13

-8.5 20.3 3.0 -1.4 2.8 16.9 8.3 -3.3 8.3 5.0 11.9 37.8 37.3 -5.4

5.9 -9.5 26.9 39.3 11.9 -10.8 5.3 13.9 -12.0 -3.9 -7.4 -28.0 -27.5 + 13.8

Sign of Covariation (3)

+ + +

0

+ 0

+

Source: See table 8.5; annual averages of end-of-month data. Notes: In col. (3), + = change in same direction; - = change in opposite direction; 0 = negligible change (less than one million kronor) in either or both. The sum of the changes in net foreign reserves and net domestic assets equals the change in the Riksbank's note issue (see expression (3) in the text). Changes in capital items are thus included. Bloomfield 1959, p. 49 has a similar table comparing year-to-year changes in foreign assets (gold, foreign exchange, and silver) and in domestic income-earning assets (discounts, advances, and securities) covering eleven central banks. His results for Sweden 190~1913 are identical to those in column (3) for seven out of fourteen observations. The difference is due to Bloomfield's use of gross domestic assets, ignoring changes in the domestic liabilities of the Riksbank and his use of end-of-year data.

384

Lars Jonung

period 1880-1913-similar to the exercise in Nurkse and Bloomfieldshows negative covariation (nineteen observations) to be considerably more frequent than a positive one (seven observations), the rest being zero signs. Table 8.7 is thus consistent with the view that the Riksbank sterilized and offset movements in foreign reserves.14 The policy of "direct deflation" in the 1850s and 1860s, however, conformed to the rules of the game. Under that policy, the overriding goal of the bank was to maintain the convertibility of its notes-the objective of a central-bank policy observing the rules of the game. In brief, the Riksbank carried out an "active" discretionary policy in the short run under the pre-1914 Swedish gold standard, relying on a large number of measures to adjust to and to counteract various disturbances. The Swedish gold standard was thus a managed gold standard. The techniques used by the Riksbank bear a great resemblance to Bloomfield's description of measures adopted by central banks. Bloomfield (1959, p. 52) concluded that the classical gold standard was "not so simple and 'automatic' a mechanism as it is often supposed to be." The evidence on the policy of the Riksbank lends support to his conclusion. 8.5.3

The Composition of Foreign Reserves and Transaction Volumes of Gold and Foreign Assets

Under the classical gold standard, international payments were mainly settled-not with gold-but with certain key currencies: the U.K. pound, the French franc and the German Reichsmark. These key currencies were thus part of private and official foreign-asset holdings of peripheral countries like Sweden. Three questions pertaining to Swedish holdings of key-currency reserves are examined here: (1) How large were the foreign assets of the Riksbank in relation to its specie reserves? (2) Which were the key currencies for Sweden under the gold standard? (3) How large were the transaction velocities of the components of the foreign reserves? (see Lindert 1969). The Swedish monetary data for a few years prior to World War I allow a detailed examination of the questions because they give the distribution by country of all foreign reserves of the Riksbank as well as the transaction volumes. In the first half of the nineteenth century, the Riksbank's holdings of foreign reserves were entirely specie. Gradually, the bank acquired deposits at foreign banks and bankers (first shown in the statistics in 1844), foreign bills (first reported in 1845), and foreign bonds (first reported in 1872). The ratio of net foreign assets to total net foreign reserves of the Riksbank fluctuated around 50 percent during the gold standard period. It was 54 percent at the end of 1913, a high level compared to a ratio of 25 percent that Lindert (1969, pp. 10-11) calculated as the average for

385

Swedish Experience, 1873-1914

Europe for official reserves, excluding the Bank of England, Bank of France, and the Reichsbank (see also table 8.8). The composition of the foreign reserves of the Riksbank at the end of 1913 is displayed in table 8.8. This table shows that the gold stock was about 46 percent of total foreign reserves (silver holdings of 5.2 million kronor are excluded here). Most of the gold was minted gold, Scandinavian gold coins representing more than half the gold reserves. The key currencies in Europe under the gold standard, the pound, the Reichsmark, and to a minor extent the franc were also represented in the form of minted gold in the holdings of the Riksbank. Table 8.8 also shows that use of gold for settling international transactions was limited. The sum of debits and credits during 1913 was around 6 million kronor compared to a stock of 102 million kronor. It is safe to conclude that gold was not used by the Riksbank as an international means of payment. Gold instead remained in its vaults. Transactions in foreign bonds, primarily of German origin, were more frequent, one reason being turnovers due to redemptions. Foreign bills and deposits at foreign banks constituted an important share of total reserves. The distribution of the total of the two assets, according to type of currency, shows that claims in Reichsmark were by far the largest item, followed by pound-sterling claims, and claims in Danish and Norwegian kroner (table 8.8). Assets denominated in francs and other currencies were insignificant. The volume of transactions in column (3) of table 8.8 indicates that foreign bills and deposits at foreign banks were most frequently used for international payments. Settlements in Reichsmark amounting to 1107 million kronor represented the bulk of transactions compared to settlements in pounds of 352 million kronor, and in Danish and Norwegian kroner of 326 million kronor. Judging from table 8.8, the German currency was the most important key or reserve currency held by the Riksbank shortly before World War 1.15 The British pound and the Danish and Norwegian currencies were second in importance. The French franc, although a reserve currency, had no role in this context. The composition of Swedish exports and imports partially explains this result, since the largest share of Swedish imports originated in Germany. Thus, the Riksbank had to maintain considerable claims in Reichsmark to meet the demand for German currency from importers. 8.6

The Law of One Price and International-Market Integration

Much of recent discussion about the adjustment of an open economy to international disturbances has dealt with the monetary approach to the balance of payments. According to that theory, for a small open economy

386

Lars Jonung

Table 8.8

Type of Asset (1) 1. Gold

Composition of Riksbank's Foreign Reserves, 1913

Volume of Transactions (million kronor) (3)

Type of Currency (million kronor) (2) Minted gold Scandinavian Sovereigns Reichsmark Francs Othera Unminted gold

Total gold 2. Foreign bonds b 3. Foreign bills 4. Foreign banks (net) (lines 3 + 4) Norwegian and Danish kroner Pound sterling Reichsmark Francs Othere 5. TOTAL foreign reserves (net)

91.4 56.0 20.5 10.9 2.7 1.4 10.7 102.1 27.5 69.8 24.8 10.2 23.2 57.4 3.3 0.2

326.3 352.3 1107.4 52.2

224.2

TOTAL transactions volume 1879.9

Sources: Riksbankens arsbok 1913, 1918. Notes: Col. (3) sums debits and credits to obtain the volume of transactions. Gold parity rates (Riksbankens arsbok 1918, pp. 74--75) were used to calculate the krona value of the composition and the transactions volume of foreign banks and foreign-bill assets (cols. 2 and 3). a"Other" includes 0.8 million Finnish marks and 0.2 million rubles. ~e item "foreign and Swedish government bonds" (in- och utliindska statspapper) was 27,489,511 million kronor end-of-December 1913. The item "foreign bonds" (obligationer: utliindska) was identical in the balance sheet of the Riksbank (Riksbankens arsbok 1913, table 29). Thus, it is safe to assume that the Riksbank held no Swedish government bonds at that time. Consequently, line 2 pertains to foreign bonds. The composition of foreign-bond holdings is not shown in the 1913 report. The 1911 report, however, shows that 52 percent were German bonds, 37 percent were U.K., and 11 percent were French bonds with a total book value of 16.6 million kronor at the end of 1911. The composition of Riksbank bond holdings was probably similar in 1913. 'Transactions in gold included sales (export) of 0.9 million kronor of Scandinavian gold coins, a purchase (import) of 3.9 million kronor of unminted gold, and sales (export) of 1.4 million kronor of minted foreign gold coins. dTransactions in foreign bonds also included Swedish nongovernment bonds. The latter represented a small share (12 percent) of the bond portfolio of the Riksbank at the end of 1913. e"Other" includes assets in dollars, rubles, Austrian crowns, Finnish marks, and Dutch florins. The volume of transactions in these currencies was so small, it is ignored (col. 3).

387

Swedish Experience, 1873-1914

adhering to fixed exchange rates, prices and interest rates are internationally determined on world markets. Arbitrage in traded goods is given a central role in establishing identical behavior of prices across countries with due allowance for costs of transportation, tariffs, and other restrictions on trade. The model, applied to Sweden during the pre-1914 gold standard era, indicates that the policy of the Riksbank had no effect on prices, interest rates, and the growth of the Swedish money stock-only on the composition of the monetary base. In particular, gold flows in or out of Sweden did not influence prices and incomes but established equilibrium between the demand for and supply of money. This interpretation of the workings of the pre-1914 gold standard is forcefully argued by McCloskey and Zecher (1976, and chap. 2, this volume). They suggest, to the extent that a common-price-Ievel behavior across countries adhering to gold is found, that the monetary approach is confirmed. They contrast the monetary approach with the price-specieflow mechanism of Hume, according to which differences in inflation rates between countries cause deficits and surpluses on trade accounts, including gold flows that reestablish a common long-run behavior of prices. This section applies the McCloskey-Zecher analysis to the Swedish gold standard record. Domestic and foreign individual and aggregate price behavior is compared, in particular, lead-lag relationships are examined. This issue is interesting to explore as one suspects that the Swedish economy, on the periphery of the world economy, adjusted with a lag to events in the center of the gold standard world-the United Kingdom. Evidence concerning some individual prices is available for Sweden. Bengtsson and Jorberg (1975) compared the price of rye within Sweden and between Sweden and Amsterdam and Copenhagen during the eighteenth and nineteenth centuries. First, they established that regional price differences declined within Sweden between 1735 and 1914 for a large number of commodities. They concluded that agricultural markets were well integrated as early as the eighteenth century in spite of bad communications and obstructive government regulations. Second, they applied spectral analysis to movements in the price of rye during two periods of preindustrial Sweden: 1732-98 and 1799-1869. For 1732-98, a high correlation was evident between the price of rye in Stockholm, Gothenburg, Amsterdam, and Copenhagen. The spectra for 1799-1869 showed a still higher correlation with no lags in price movements in the four cities. A close correlation existed between larger fluctuations in interregional and international rye prices. The absence of lags in the nineteenth century was interpreted as signifying improved economic integration.

388

Lars Jonung

According to Bengtsson and Jorberg (1975, p. 105), "prices in Sweden were not determined regionally, but were influenced by interregional and international price movements." The period they investigated ended in 1869, that is, before the introduction of the gold standard in Sweden. To them it was "self-evident" that economic integration became more pronounced with the spread of railways and the rise of industrialization. Their results thus suggested that Swedish prices covaried strongly with foreign prices even prior to the gold standard era. In a monumental study of Swedish price history, Jorberg (1972) compared movements in foreign and domestic prices and in market price scales (markegangstaxor). These price scales, determined at annual meetings in every county by representatives of the tax payers and by the authorities, were used to transform tax payments in kind into cash values. Jorberg examined the correlation of the price scales with "ordinary" domestic and foreign prices. A high correlation was interpreted as indicating that the market price scales reflected actual market prices well. The purpose of the tests was to explore the representativeness of the market price scales-not to study the law of one price. Jorberg's results, however, can be used for a study of the international integration of the Swedish economy. Table 8.9 presents some of Jorberg's correlations of prices of wheat and rye during selected periods of the eighteenth and nineteenth centuries. The correlations support the view that prices within Sweden covaried to roughly the same extent as prices between Sweden and foreign countries like England, Germany, Norway, and Holland. It is worth noting that the price of wheat in England was highly correlated with the market price scale of wheat in Sweden as early as in the eighteenth century. The correlations are generally higher for the nineteenth century than for the eighteenth century. The covariation between Prussian grain prices and grain prices quoted on the Stockholm exchange for 1861-1913 is roughly of the same magnitude as the covariation between the latter set of prices and the market price scales of wheat and rye in Stockholm county for 1838-1913. Jorberg concluded that the high correlations between the market price scales and other price data, Swedish as well as foreign, suggested that the market price scales were good measures of actual price fluctuations. A conclusion may also be drawn from table 8.9 concerning the law of one price: Swedish prices covaried closely with foreign prices during the classical gold standard period. Domestic Swedish and foreign markets were about as well integrated as regional markets within Sweden. The behavior of Swedish and U.K. aggregate price indexes may throw light upon the issue of market integration. Figure 8.4 plots the U.K. wholesale price index (Sauerbeck's index), the wholesale price index for Sweden (Amark 1921), and a cost-of-living index for Sweden (Myrdal

389

Swedish Experience, 1873-1914

Table 8.9

Correlations between Foreign Prices, Domestic Prices, and Market Price Scales: Selected Periods, 1732-1914

Correlations Between (1) Interregional Markets Stockholm exchange commodity prices and market price scales, Stockholm county Stockholm exchange commodity prices (Sept.-Febr.) and market price scales, Stockholm county Price quotations from the town of Malmo and market price scales, Malmohus county International Markets English average prices and market price scales, averages for Sweden Prices for Prussian rye (Amsterdam exchange) and market price scales, averages for Sweden Market prices in Norway and market price scales, averages for Sweden Gazette prices in England and prices, Stockholm exchange Prussian grain prices and grain prices, Stockholm exchange Export prices of butter, Finland, and market price scales, averages for Sweden Beef prices, Finland, and market price scales, averages for Sweden

Period (2)

Wheat (3)

Rye (4)

1838-1913

.86

.83

1880-1913

.81

.94

1880-1914

.84

.93

1732-1759

.95

1732-1789 1790-1808 1816-1874 1836-1914

.59 .61 .67

1860-1913

.74

1861-1913

.89

Other Commodities (5)

.72

.84

1882-1913

.94

1878-1913

.95

Source: Jorberg 1972, table 3.4. Note: The Stockholm exchange traded primarily in commodities.

1933) for 1871-1913 with 1880 as the common base year. The covariation among the three series is substantial. The secular pattern is identical: long-run deflation during the 1880s and the early 1890s followed by secular inflation. The cyclical fluctuations seem similar too. Table 8.10 gives correlation coefficients for leading, concurrent, and lagging observations of annual percentage changes in British and Swedish prices during the gold standard period, 1873-1913. The highest correlation is for simultaneous observations, while leads and lags give considerably lower correlations. Thus, Swedish markets display the same pattern

390

Lars Jonung 1871

125 120

=

lOa

R··

atlo sca 1 e I

115 110 105

Swedish cost-of-living

-, ,

/

.... ,-" .... ",'"

",'

,"

-'I

1 OO~-"""""-+----~,---",~,,,~,:------r-------+-----I

95 90

,

I

"

,.'

',,./'';-

85

Swedish wholesale

80 75 70 U.K. wholesale prices

65

60

T~70 Fig. 8.4

1875

1880

1885

1890

1895

1900

1905

1910 Year

Swedish and U.K. prices, 1871-1913. Sources: U.K. wholesale prices (Sauerbeck's general index): Amark 1919, table 345. Swedish wholesale prices: Amark 1921, a mixture of wholesale and consumer prices (markegangstaxor). Swedish cost-ofliving index: Myrda11933, table A, similar to a consumer price index since it is based on market prices, prices in newspapers, and markegangstaxor.

as British markets. Table 8.10 also shows a higher correlation for wholesale prices in the two countries than for Swedish wholesale prices and Swedish consumer prices. The result probably reflects a larger share of traded goods in the wholesale price index than in the cost-of-living index. There is also a tendency for changes in Swedish and U.K. wholesale prices to lead changes in the cost-of-living if one may judge from the fairly high correlation coefficient for a one-year lead. Sweden became a substantial long-term international borrower in the second half of the nineteenth century. For this reason, as long as Sweden adhered to a fixed exchange rate, the Swedish long-term bond rate was simply equal to the long-term rate of the international capital market. The U.K. discount rate was lower than the Swedish discount rate, reflecting the United Kingdom's role as a net lender in the center of world finance and Sweden's position as a net borrower (see figure 8.5). The covariation between Swedish and U.K. rates appears to be fairly close, at least from the 1880s onwards, but not as high as for prices (figure 8.4 and table 8.10). The highest correlation coefficient occurs when the British discount rate leads the Swedish rate by one year. This lead suggests that the Riksbank altered its rate in response to changes in the U.K. rate but with a lag. The rates charged by Swedish commercial banks and the discount rate

391

Swedish Experience, 1873-1914

Table 8.10

Variables

Correlation Coefficients between Annual Percentage Changes in Prices and Interest Rates in England and Sweden under the Pre-1914 Gold Standard Period

Price Indexes 1873-1913 English and Swedish wholesale prices English wholesale prices and Swedish cost-of-living Swedish wholesale and cost-of-living 1880-1913 Interest Rates British discount rate and Swedish discount rate Swedish discount rate and Swedish commercial bank rate

+2

Correlation Coefficients -1 0 +1

-2

-.019

.289

.842

.287

-.013

.194

.381

.579

.039

-.048

.143

.584

.691

.071

-.147

.018

.507

.288

-.288

-.186

.002

.455

.842

.062

-.308

Sources: See figures 8.4 and 8.5. Note: First listed variable leads the second listed variable 2 years; 1 year; both coincident; first listed variable lags the second variable 1 year; 2 years.

of the Riksbank covaried closely according to table 8.10 and figure 8.4. The pattern is most likely due to a common international influence on Swedish rates-and not to a domestic one running from the discount rate to commercial-bank rates or vice versa, at least not in the nineteenth century when the Riksbank also acted as a commercial bank and some big commercial banks served as central banks for smaller banks. To sum up, Swedish individual prices as well as aggregate price indexes covaried closely with international price movements both in the short and in the long run under the pre-1914 gold standard. Differences in short-run inflation rates between Sweden and the United Kingdom are not evident under the gold standard, at least not on any significant scale. Swedish markets were thus well integrated with international markets. These results challenge any theory that rests upon short-run divergences between Swedish and foreign inflation rates and lend support to the monetary-approach interpretation of the Swedish gold standard, according to which prices and interest rates in Sweden were determined internationally. 8.7

Summary

The following conclusions emerge from the present study: 1. The Swedish gold standard was not a domestic gold standard based on large-scale circulation of gold coins within Sweden. The authorities

392

Lars Jonung Percent

7.5

--,

~

7.0 -

6.5 6.0 -

~Swedish

commercial bank mortgage rate

- - - - \ \

5.5

\

,-',

\

5.0 -

\

/ /

4.5 4.0 .

3.5 3.0 _ :

2.5

.. . .

-• •-

e•

.... U. K. discount rate

2.0 -

Year

Fig. 8.5

Interest rates in Sweden and the United Kingdom, 1871-1913. Sources: Swedish discount rate: Riksbankens arsbok 1913, p. 73, apparently an annual average of daily rates. Mortgage rate on real estate (utlaningsriinta mot inteckning i fast egendom) of Swedish commercial banks: monthly bank reports. Annual averages of end-of-month data averaged for three major commercial banks (Stockholms Enskilda Bank, Skandinaviska Kreditaktieb 0 lag , and Aktiebolag Stockholms Handelsbank). U.K. discount rate: Amark 1919, table 345.

minimized gold holdings of the Swedish public by supplying notes of lower denominations than gold-coin denominations. Legal reserve requirements of the commercial banking system and of the Riksbank were framed so that no rigid link between the gold reserves and the note issue was established. Consequently, considerable short-run as well as longrun autonomy was given the banking system's supply of notes. The incentive structure of the monetary system contributed to monetary stability. 2. The long-run growth of the Swedish money stock during the pre1914 gold standard era was, however, strongly related to the growth of the gold and foreign-asset holdings of the Riksbank.That growth was financed by long-term borrowing on international capital markets. 3. In the short run, there was no rigid link between gold flows and the money stock. The Riksbank's policy of constant gold holdings minimized gold flows by foreign-exchange operations, borrowing abroad, and other monetary techniques. The Swedish pre-1914 gold standard was thus a managed gold standard.

393

Swedish Experience, 1873-1914

4. Swedish prices covaried closely with international prices, both in the long run and in the short run, from 1880 to 1913. The covariation indicates that domestic markets were well integrated with international markets. Consequently, there is strong support for the law of one price and thus indirectly for the monetary approach to the balance of payments. 5. Sweden adhered to the gold standard during 1873-1914, a period of extremely rapid industrialization of its economy as well as of unusual domestic monetary and political stability. The free flow of capital into Sweden from financial centers like London, Paris, and Berlin was an engine for the transformation of its economy. Monetary thought and monetary policy in Sweden after 1914 was greatly influenced by the experience under the classical gold standard system. The gold standard was commonly regarded as the normal state of affairs and World War I as a temporary disturbance. The parity rate of the krona became the norm for postwar monetary reconstruction in spite of far-reaching changes due to World War I and the inflation accompanying the war. In 1922 Sweden was the first country in Europe to decide to return to the gold standard at the 1914 parity rate. Partly as a consequence of this step, a sharp deflati0!l and great unemployment ensued. The policy reflected expectations that the gold standard, once reintroduced, would lead to stability, economic growth, and prosperity-as was the case prior to 1914. When these expectations did not materialize, economists and politicians became skeptical about the workings of the gold standard, a critical view that was absent during the pre-1914 gold standard period.

Appendix

Knut Wicksell and Gustav Cassel on Money, Gold, and Prices in the Long Run

At the turn of the century, Knut Wicksell and Gustav Cassel each presented a theory to explain secular price-level movements during the nineteenth century. Using basically the same data set, they reached different conclusions. Their views are examined here in relation to the analysis of the Swedish-money-supply process in section 8.5.1 (see Jonung 1979a for a detailed analysis of their views). Wicksell's theory (1898) was based on a distinction between two interest rates: the market rate charged by the banking system and the natural rate which equates the demand for and supply of capital. A discrepancy between these rates gave rise to cumulative movements in prices. When the natural rate is above the market rate, investors borrow from the banking system, commercial banks reduce their reserves, the

394

Lars Jonung

money stock and prices increase. When the natural rate is below the market rate, borrowing is arrested, reserves increase, and the money stock and prices decline. Wicksell applied his theory to the behavior of prices and interest rates in the United Kingdom in the nineteenth century. According to him, the U.K. inflations of 1790-1815 and 1851-73 occurred because the natural rate was above the market rate, and the secular deflations of 1815-50 and 1873-96 occurred because the market rate was above the natural rate. Movements in the natural rate, due to technological innovations or wars, were thus the driving force behind secular movements in prices. He assumed that the banking system reacted passively to changes in the demand for credit even in the long run. Gold was assigned a secondary role in long-run processes. . Wicksell's theory does not lend itself to a direct test because the natural rate, the key variable in his theory, is not observable-a fact he was well aware of. His theory, however, has implications for the behavior of the monetary system. The cumulative process implies that the reservedeposit ratio of the commercial banking system should be rising during secular deflations and falling during secular inflations. Alternatively, since Wicksell consolidated the balance sheets of the central bank and commercial banks, the ratio of gold to the monetary liabilities of the monetary system should be rising during secular deflations and falling during secular inflations. An inspection of the Swedish evidence reveals that the reserve-deposit ratio of the Swedish commercial banking system was falling, not rising as suggested by the cumulative process, during the secular deflation of 1873-96. As figure 8.2 shows, the ratio of gold reserves to the Swedish money stock remained constant at about 5 percent during the pre-1914 gold standard period. This constancy is inconsistent with the cumulative process as a theory of long-run movements in prices. Cagan (1965) reached similar conclusions concerning Wicksell's theory on the basis of U.S. evidence. These results do not disprove the existence of a cumulative process as presented by Wicksell. Rather, they suggest that developments other than those discussed by Wicksell dominated actual developments. Six years after Wicksell published his theory, Cassel (1904) provided an alternative one, suggesting that secular movements in prices in the nineteenth century were due to different growth rates of the demand for and supply of gold. His conclusion was based on the following four-step analysis. First, Cassel estimated the annual actual stock of gold in the world in the nineteenth century. Second, turning to the demand side, he argued that the increase in the world demand for gold between 1850 and 1900 was due to factors other than price-level changes because the price

395

Swedish Experience, 1873-1914

level was the same in the two years. With the help of this assumption, he calculated the "normal" growth rate of the stock of gold to be 2.65 percent per year between 1850 and 1900. Had the supply of gold increased uniformly at that rate (the normal gold stock), the growth in the demand for and supply of gold would have been equal and the price level would have remained constant. Third, Cassel estimated the ratio of the actual to the normal gold stock (the relative gold stock). Finally, plotting the relative gold stock against the U.K. price level on the base 1850, he concluded that the two curves followed each other closely in the long run. Cassel believed that he had demonstrated that (a) short-run movements in prices were not related to gold movements; (b) long-run differential growth rates in the demand for and supply of gold caused secular changes in the price level. Cassel's views (1904) should properly be regarded as a precursor of global monetarism. The demand for and supply of gold are determined by two different sets of factors. The demand for gold is proportionate to the price level, i.e. ~ it is a demand for real-gold balances, and it exhibits a stable long-run growth rate. The supply of gold is determined by the opening of new mines and changes in gold-processing technology. It is thus apt to undergo larger fluctuations in the long run than the demand for gold. There is considerable evidence in favor of Cassel's view. Secular inflation in the 1850s and 1860s was due to gold discoveries in California and Australia. The inflation of 1896-1914 was associated with the opening of new gold mines in South Africa, Alaska, and Australia (see Rockoff, chap. 14, this volume). Furthermore, Cassel (1918) argued that the ratio of a country's holdings of gold to its money stock should be constant in the long run. Changes in gold should be followed by "corresponding" changes in the money stock, hence the ratio of gold to the money stock should have been constant, as it was during the pre-1914 gold standard period for Sweden (see figure 8.2). In summary, Cassel's theory appears more plausible than Wicksell's. Wicksell, a critic of the gold standard, wanted to replace it with a monetary policy based on price stabilization. Cassel was initially a supporter of the gold standard. Following an intense debate on the merits and demerits of the gold standard during World War I and in the 1920s, Swedish economists became increasingly critical of the workings of the interwar monetary system based on gold. Eventually that exchange of ideas contributed to the adoption of Wicksell's norm of price stabilization as the guide to Riksbank policy when Sweden left the gold standard in September 1931 (for an account of the implementation of Wicksell's norm in the 1930s, see lonung 1979b).

396

Lars Jonung

Notes 1. An important source of information on the policy of the Riksbank during the prewar gold standard is Brisman (1931). His work is part of the five-volume history of the Riksbank, Sveriges Riksbank 1668-1924. However, Brisman did not have access to the archives of the bank, which have not yet been the subject of a thorough economic study. Heckscher (1926) briefly considers the Swedish gold standard prior to 1914. 2. See Bordo (chap. 1, this volume) for a detailed presentation of the issues in debate concerning the gold standard. 3. The interaction between the legal framework and the growth of commercial banking in Sweden is analyzed in lonung 1978. The growth of the Swedish banking system is described in Flux 1910 and Sandberg 1978. 4. A well-functioning monetary system was probably an important factor in Sweden's rapid economic growth prior to World War I. This point is stressed by Sandberg (1978). 5. Riksbankens arsbok 1913 gives statistics on both the exchange of notes and of subsidiary silver coins between Sweden on one hand and Denmark and Norway on the other. In 1913, 10 million kronor in Swedish notes were returned to Sweden by the latter, and 38 million kronor in Danish and Norwegian notes were sent from Sweden. The figures should be compared to a total average volume of 468.3 million kronor of Scandinavian (Danish, Norwegian, and Swedish) notes in circulation in 1913. The exchange of silver coins was much smaller; about one million kronor in each direction. 6. Riksbankens arsbok prior to 1914 showed the exchange rates of the pound and the Reichsmark but made no mention of the rates of the Danish and Norwegian currencies. The Scandinavian central banks apparently eliminated any deviations from the parity rates. Differences in monetary policies during World War I eventually caused the destruction of the union (see Heckscher 1926). 7. After 1881 five-kronor coins were minted. The twenty-kronor coin accounted for 71 million out of 85 million kronor in gold coins minted prior to World War I (see Wallroth 1918). 8. The authorities made no serious attempts to increase the domestic use of gold. The banking legislation of 1874 gave the government the right to withdraw at its discretion the five-kronor note issued by private banks, presumably to promote the circulation of gold. The gold standard was adopted in Sweden during the boom in the early 1870s. Large export surpluses during those years served to increase the circulation of gold (on this point see Montgomery 1934, pp. 5{}-51). The 1874 legislation was the basis for an attempt in the late 1870s to restrict the supply of private notes and increase the circulation of Riksbank notes-not of gold coins. 9. These attempts included bribes to tax authorities to accept private bank notes that were not legal tender and regular requests to borrowers to repay loans in cash, mainly Riksbank notes. The loans were paid out again next day in private bank notes (see lonung 1978). 10. Riksbank notes were not counted as legal reserves although they were legal tender. However, from the viewpoint of the private enskilda bank, the notes represented base money since they could be exchanged for gold at the Riksbank at a fixed price. 11. Only selected kinds of foreign assets were counted as legal reserves. 12. Fleetwood (1947, p. 45) notes that "the Swedish emigration coincides with the period of the heaviest capital import. Together these two movements made possible the rapid increase in the standard of living in Sweden." Emigration from Sweden facilitated economic growth as Sweden otherwise would have had a "surplus" population, given its natural resources. According to Fleetwood (p. 46), a "surplus of labour and deficiency of capital" were eliminated by the simultaneous occurrence of emigration and capital imports. 13. Brisman (1931) notes that in 1865 the Board of Directors of the bank subscribed to

397

Swedish Experience, 1873-1914

the Economist which promoted Bagehot's views concerning the responsibility of a central bank to use the discount rate to maintain domestic and external stability. 14. It is also consistent with the monetary approach to the balance of payments which suggests that the rules of the game were inconsistent with a gold standard system (see McCloskey and Zecher 1976). 15. The composition of the foreign reserves of the Riksbank had a major effect on the policy of the bank following the outbreak of war in August 1914. The bank wanted to avoid losses on its Reichsmark holdings due to a depreciating German currency by allowing the krona to depreciate with the Reichsmark. This policy was strongly criticized by economists, in particular by Cassel.

References Amark, Karl. 1919. Statistisk oversikt av det svenska niiringslivets utveckling aren 1870-1915 (Statistical survey of Swedish commerce and industry 1870-1915). Statistiska meddelanden, Sere A, vol. 3, pt. 1. Stockholm: Kommerskollegium. - - . 1921. En svensk prisindex 1860-1913 (A Swedish price index, 1860-1913). Kommersiella meddelanden (no. 8): 1259-87. Annual reports of the commercial banks of Sweden. 1871-1914. Archives of the Bank Inspection Board, Stockholm. Typescript. Bairoch, Paul. 1976. Europe's gross national product. Journal of European Economic History 5 (Fall): 273-340. Bengtsson, Tommy, and Lennart Jorberg. 1975. Market integration in Sweden during the 18th and 19th centuries: Spectral analysis of grain prices. Economy and History 18: 93-106. Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard, 1880-1914. New York: Federal Reserve Bank of New York. Brisman, Sven. 1931. Den stora reformperioden 1860-1904 (The period of great reforms). In Vol. 4 of Sveriges Riksbank 1668-1924 (The Riksbank of Sweden, 1668-1924), part 3. Stockholm: Sveriges Riksbank. Cagan, Phillip. 1965. The determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. Cassel, Gustav. 1918. Theoretische sozialokonomie. Leipzig: C. F. Winter. - - - . 1904. Om forandringar i den allmanna prisnivan (On changes in the price level). Ekonomisk Tidskrift 6: 311-31. Davidson, David. 1896. Bankreformen och naringslivet (The bank reform and industry). Nationalekonomiska Foreningens Forhandlingar (Feb.): 24-37. Fleetwood, Erin E. 1947. Sweden's capital imports and exports. Geneva: Institut Universitaire des Hautes Etudes Internationales.

398

Lars Jonung

Flodstrom, Isidor. 1912. Sveriges nationalformogenhet (Sweden's national wealth). Stockholm: Kungl. finansdepartementet. Friedman, Milton, and Anna Jacobson Schwartz. 1963. A monetary history of the United States, 1867-1960. Princeton: Princeton University Press. . Flux, Alfred W. 1910. The Swedish banking system. Prepared for the U.S. National Monetary Commission. Washington, D.C.: GPO. Heckscher, Eli. 1926. Penningvasende och penningpolitik (The monetary system and monetary policy). In Bidrag till Sveriges ekonomiska ochsociala historia under och efter viirldskriget, ed. Eli Heckscher, part 2. Stockholm: Norstedt. Historisk statistik for Sverige. De13. Utrikeshandel1732-1970 (Historical statistics of Sweden. Part 3. Foreign trade 1732-1970). 1972. Stockholm: Statistiska Centralbyran (National Central Bureau of Statistics). Johansson, Osten. 1967. The gross domestic product of Sweden and its composition, 1861-1955. Uppsala: Almqvist-Wiksell. Jonung, Lars. 1975. Studies in the monetary history of Sweden. Ph.D. diss., University of California at Los Angeles. - - - . 1976. Money and prices in Sweden, 1732-1972, Scandinavian Journal of Economics 78: 40-58. - - - . 1978. The legal framework and the economics of private bank notes in Sweden, 1831-1902. In Law and economics, ed. Goran Skogh. Lund: Studentlitteratur. - - - . 1979a. Knut Wicksell and Gustav Cassel on secular movements in prices. Journal of Money, Credit, and Banking 11 (May): 165-81. - - - . 1979b. Knut Wicksell's norm of price stabilization and Swedish monetary policy in the 1930's. Journal of Monetary Economics 5 (Oct.): 459-96. - - - . 1983. Monetization and the behavior of velocity in Sweden, 1871-1913. Explorations in Economic History 20:418-39. Jorberg, Lennart. 1970. The industrial revolution in Scandinavia, 18501914. In The Fontana economic history of Europe, ed. Carlo M. Cipolla, vol. 4. London: Collins. --.1972. A history ofprices in Sweden, 1732-1914. Volume 1. Lund: Gleerup. Lindert, Peter H. 1969. Key currencies and gold, 1900-1913. Princeton Studies in International Finance, no. 24. Princeton: Princeton University Press. McCloskey, Donald M., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance ofpayments , ed. J. Frenkel and H. G. Johnson. Toronto: University of Toronto Press. Montgomery, Arthur. 1934. Riksdagen och riksbanken efter 1809 (The Riksdag and the Riksbank after 1809). In Sveriges Riksdag, ed. Nils Eden, vol. 13. Stockholm: Sveriges Riksdag.

399

Swedish Experience, 1873-1914

Myrdal, Gunnar. 1933. The cost ofliving in Sweden, 1830-1930. London: P. S. King. Nurkse, Ragnar. 1944. International currency experience. Geneva: League of Nations. Riksbankens arsbok (The annual report of the Riksbank). 1908-14. Stockholm: Sveriges Riksbank. Rosenberg, Carl M. 1878. Handbok i bankviisendet (Introduction to the Banking System). Stockholm: Bonniers. Sandberg, Lars G. 1978. Banking and economic growth in Sweden before World War I. Journal of Economic History 38 (Sept.): 650-80. Sveriges Riksbank 1668-1924. 1931. 5 vols. Stockholm: Sveriges Riksbank. Wallroth, Karl A. 1918. Sveriges mynt 1449-1917 (Sweden's coins, 14491917). Numismatiska Meddelanden 21. Wicksell, Kunt. 1898. Geldzins und Giiterpreise. lena: Gustav Fischer. - - - . 1896. Kommentar till Davidson (A comment to Davidson). Nationalekonomiska F6reningens F6rhandlingar (Feb.): 37-39. - - - . 1904. Framtidens myntproblem (The currency problems of the future). Ekonomisk Tidskrift 6: 82-106.

Comment

Peter H. Lindert

Lars lonung has written a successful paper about a successful nation. I would like to underline briefly some of the achievements of his paper and then turn to the still-underexploited opportunity to draw conclusions about the link between prewar Swedish success and the gold standard. lonung has skillfully described the prewar evolution of the Swedish monetary system into a smoothly functioning managed paper currency. Gold hardly circulated at all domestically. It gravitated toward the Rik~­ bank both from within the country and from without, providing a growing metallic reserve. Growth in gold and foreign-exchange reserves accounted for three-fourths of the growth in the money supply, easily eclipsing the contributions of the deposit-currency ratio and the depositreserve ratio. Yet this reserve, as lonung stresses, was largely redundant, since the nonbank public always trusted its paper money and its legislators always kept reserve requirements low enough to leave the Riksbank and commercial banks with considerable unrequired reserves. The domestic shunning of gold and the redundancy of gold and foreignexchange reserves held by the banking system should be viewed as testimony to the efficiency of Sweden's monetary system. The nonuse of gold in transactions is what one would expect of a "successful gold Peter H. Lindert is professor of economics at the University of California, Davis.

400

Lars Jonung

standard"; if people are convinced that paper money is as good as gold, then expensive and barren gold dwindles away as a share of either holdings or transactions within Sweden just as it did in the prewar economy as a whole. The redundancy of reserves also bespeaks freedom from crisis in the same way that infrequent use of bank-deposit insurance is a sign that deposit insurance may be doing its job well. These conclusions are well supported by the material Jonung has presented. In separate and somewhat disjointed sections of his paper, Jonung adds to our appreciation of the Swedish case by exploring the purchasingpower-parity hypothesis, the determinants of velocity, and the early gropings toward a theory of the price level by Wicksell and Cassel. The results here are firmly based, plausible, and noncontroversial. Purchasing-power parity gets the same partial and moderate support it receives elsewhere. The secular decline in velocity is mirrored in indicators of growing monetization. The appendix on Wicksell and Cassel is too brief to give major new insights, but it does remind us of Swedish patent rights in the area of monetary theory and the theory of purchasing-power parity. These contributions set the stage for a rethinking of the central issue relating to Sweden's participation in the gold standard system: What difference did the gold standard make to Swedish economic growth and stability? The simple correlation is perfect: Before Sweden joined the gold standard, she was probably not growing any faster than Europe as a whole, yet upon joining she acquired the fastest growth rate in Europe, surging from a per capita income about two-thirds that of all Europe in 1873 to parity with the European average around 1900 to a 27 percent lead on the continent by 1913 (Bairoch 1976, p. 286). The purchasing power of that national product per capita rose even faster, since Sweden's external terms of trade improved greatly across the last quarter of the nineteenth century and were still better in 1913 than at the onset of her gold standard era.! This neat correlation looks like a fine scaffolding for Jonung's conclusion that Sweden adhered to the gold standard, 1873-1914, a period of extremely rapid industrialization of its economy as well as of unusual domestic monetary and political stability. The free flow of capital into Sweden from financial centers like London, Paris, and Berlin was an engine for the transformation of its economy. (P. 393) Some readers may wish to infer that staying on gold raised and stabilized Sweden's national product. While this inference cannot be firmly rejected, I wish to counterpose it with my own null hypothesis: Sweden's participation in the international gold standard made no visible net contribution to the growth or stability of her national product, relative to a hypothetical world of flexible exchange rates for the krona.2 One may

401

Swedish Experience, 1873-1914

even doubt whether it mattered to Sweden that Britain and other major countries themselves adhered to the fixed-exchange-rate orthodoxy. To support this null hypothesis, I shall first note how easy it seems to give proximate explanations for Sweden's rapid prewar growth without referring to any particular monetary institutions. By about the time she entered the gold standard and the redundant Scandinavian Monetary Union, Sweden was in the potentially favorable position that Lars Sandberg has aptly dubbed "the case of the impoverished sophisticate" (Sandberg 1979). Relative to her average income and capital stock, Sweden had remarkably high literacy, schooling, life expectancy, technological performance in some industries, and acceptance of modern financial institutions. Her rate of return on schooling may have been initially low for noneconomic reasons (mainly Lutheran insistence on universal literacy). In this respect, she bore a remarkable resemblance to the rest of Scandinavia, to North America, and to Meiji Japan at the start of their respective growth accelerations. Her development had been held back by high transport costs and by an initially unfavorable natural-resource endowment. Yet she had the human resources to exploit opportunities when they arose. Opportunities arose after midcentury. Swedish oats, softwood, pulp, iron ore, and mechanical skills were given higher value by several developments in the world economy. World demand for softwood lumber and pulp expanded, helped by a decline in transport costs. That demand was shifted toward Sweden by the end of British preference for Canadian wood in the 1850s and the exhaustion of Norwegian timber. The ThomasGilchrist process in 1879 raised the steelmaking demand for basic Swedish iron ore. Each resource-related sector gave rise to machinery and engineering lines that were tapped with alacrity by a well-trained nation. Meanwhile, a stable government was maintained without heavy military investments. This not-so-monetary story of prewar Swedish growth can be extended further to endogenize Sweden's attraction of foreign capital and her accumulation of gold and foreign-exchange reserves. A nation rich in human capital and newly valued natural resources relative to man-made nonhuman capital is likely to attract foreign lending. This attraction follows from the usual complementarity of man-made nonhuman capital with skills and nature. Such a nation, especially with a rising share of foreign trade in national product, is also likely to have a growing demand for international money balances in a way analogous to the minor role played by the human share of total wealth in Milton Friedman's restatement of the quantity theory of money. Perhaps the accumulation of gold and foreign exchange that Jonung rightly gives credit for 75 percent of Sweden's money growth is itself explained primarily by the opening of new trade opportunities for an already-skilled poor nation.

402

Lars Jonung

Thus far I have merely offered an alternative plausible story of Swedish growth without really showing that participation in the international gold standard made no contribution at all to Swedish stability or growth. International economists still lack a consensus model for giving international monetary regimes their proper due in growth accounting. We must be content with indirect clues and theoretical priors about the gold standard's contribution to Swedish stability and growth. If we could find cases in which the Swedish economy was subjected to severe foreign or domestic shocks, we could see if either Swedish monetary policy or induced private financial movements saved her in a way that would not have been available without fixed exchange rates and gold convertibility. Theory would say that the gold standard, relative to flexible exchange rates, should have partly cushioned the domestic money stock and national product against domestic shocks and partly exacerbated their vulnerability to foreign-trade shocks. The difficulty here is in finding any shocks inflicted on prewar Sweden. As Jonung has shown for the years from 1880 on, the Riksbank never lost foreign reserves in serious quantities. Thus, even though Jonung has helpfully shown us that the Riksbank tended to sterilize her external imbalances and smooth out the growth of the money supply, we should also note that she never had serious payments deficits that needed to be sterilized. As for the surpluses she tended to run on average, there has always been an asymmetry in fixed-exchange-rate systems that has made surpluses less disruptive to the domestic economy than deficits of equal magnitude as a percentage of the stock of reserves. The Riksbank, like the postwar Bundesbank, the Bank of Japan, and other central banks in chronic surplus, could go on for decades without being seriously inconvenienced by reserve accumulation. In fact, the absence of signs of crisis pervades all the Swedish macroeconomic series available for this era. The money supply never dropped seriously. Annual GNP figures show occasional drops, but never large or consecutive ones. Modest nominal wage cuts occurred, but real wage cuts were never severe. In such a setting, it is hard to argue that the gold standard imposed any "discipline" whatsoever. Extending the time period allows us to see how the Swedish monetary system responded to true pressure, and the response is instructive. Sweden, of course, like everybody else, suspended convertibility during the World Wars. In the Great Depression, when the going also got tough, Sweden promptly followed Britain's flight from the gold standard in order to cushion herself against world deflation, setting the stage for her reflationary monetary and fiscal policies. The money stock did not drop, and Sweden enjoyed one of the world's shallowest depressions of that decade, as Lars Jonung (1981, pp. 286-315) has ably noted elsewhere. Thus Sweden, like almost all other countries before and after 1914,

403

Swedish Experience, 1873-1914

observed the gold standard orthodoxy when it made no difference to the domestic economy, but abandoned it as soon as it began to bind. My defense of the null hypothesis remains incomplete. In particular, I have not been able to test Jonung's belief that adherence to the gold standard attracted more inflow of capital from the lending countries. In this belief he has the company not only of other scholars but of most prewar monetary officials. One could counter it with the observation that even borrowing nations with flexible exchange rates attracted prewar capital, sometimes with the expedient of shouldering the exchange risk and agreeing to repay in gold-backed currency. Yet a convincing test eludes us. Flexible exchange rates occurred often before 1914, but only in the wake of crises that the fixed-exchange-rate system could not handle adequately, such as the U.S. Civil War or fiscal profligacy. Any possible relationship between the exchange-rate regime and the supply of foreign funds will be clouded by this background of prior crisis. The debate over the effects of the prewar gold standard on a prospering economy like Sweden thus remains open. As I read it, Jonung's capable summary of Swedish experience allows us to argue that Sweden stayed on the gold standard peacefully because she grew rapidly, and not vice versa.

Notes 1. The conventional terms of trade (PxIPm) moved as follows:

Merchandise (Johansson 1967, pp. 138-49) Merchandise and services (Kindleberger 1955, p. 365)

1872 92.1 87.0

1900 105.0 107.0

1913 100 100

2. In appraising the effect of this gold standard, different scholars will give different answers to the key underlying question: "A gold standard relative to what?" Possible counterfactual alternatives for countries on the gold standard include other commodity standards, barter, exchange controls, the Bretton Woods adjustable peg, dirty floats and clean floats. To make a comparison with an alternative widely used both before World War I and (especially) since 1971, I shall compare the gold standard with floats, both dirty and clean.

References Bairoch, Paul. 1976. Europe's gross national product, 1800-1975. Journal of European Economic History 5 (Fall): 273-340. Johansson, Osten. 1967. The gross domestic product of Sweden and its composition, 1861-1955. Stockholm Economic Studies, n.s. 8. Uppsala: Almqvist-Wiksell. Jonung, Lars. 1981. The depression in Sweden and the United States: A comparison of causes and policies. In The Great Depression revisited, ed. Karl Brunner. Boston: Martinus Nijhoff.

404

Lars Jonung

Kindleberger, Charles p~ 1955. The terms of trade: A European case study. New York: Wiley. Sandberg, Lars. G. 1979. The case of the impoverished sophisticate: Human capital and Swedish growth before World War I. Journal of Economic History 39 (Mar.): 225-42.

Italy in the Gold Standard Period, 1861-1914

9

Michele Fratianni and Franco Spinelli

9.1

Introduction

Little is generally known about Italian experience under the gold standard, especially during the gold standard period before World War I, since Italy adhered to the standard only intermittently. The Italianlanguage literature on the subject is mainly qualitative in nature, while the English-language literature is virtually nonexistent. For a long time the inadequacy or outright lack of data impeded progress. But relevant statistics are now available and we intend to exploit them to remedy, at least in part, this void. Our strategy is to study the 1861-1914 period in light of what the literature today considers to be the important issues; these are discussed in the following section. Section 9.3 gives the reader a brief history of the period-a background essential for a deeper appreciation of the quantitative evidence presented in section 9.4. The salient findings of the paper are summarized in section 9.5. Some data not easily accessible are appended to the paper. 9.2 9.2.1

Theoretical Issues Hume versus the Monetary Approach to the Balance of Payments (MAPA or Perfect Arbitrage)

Kreinin and Officer (1978, p. 10), in their survey of the monetary approach to the balance of payments, remark that Michele Fratianni is professor of economics at Indiana University, Bloomington, Indiana. He was serving as senior staff economist at the Council of Economic Advisers when the paper was prepared. Franco Spinelli is an economist with the International Monetary Fund.

405

406

Michele Fratianni and Franco Spinelli

it is often suggested that the new monetary approach is the intellectual grandchild of the price-specie-flow mechanism developed by David Hume in the eighteenth century. Monetary flows are central to both theories, and both regard external imbalances as self-correcting. However, in the price-specie-flow mechanism, monetaryflows rectify external disequilibria through their effect on relative commodity prices. In contrast, the monetary approach views a stable demand for money as the core of the mechanism, and relative commodity prices play no role in the adjustment process. Price elasticities are therefore considered irrelevant. In fact, some monetarists hypothesize that perfect international arbitrage ensures that one price will prevail internationally on all commodity and capital markets, so that no changes in relative commodity prices are even possible-let alone necessary-for international adjustment. This distinction is fundamental and deserves close scrutiny. The world of Hume is traditionally analyzed in a two-country setting. Assume that an exogenous increase in the monetary gold stock takes place in country A, the effect of which is to raise, with a lag, the price level in A relative to country B. The changing terms of trade cause A to run a trade-account deficit matched by a surplus in country B. The deficit is financed by gold moving from A to B. On the assumption that the authorities do not sterilize gold flows, the trade imbalance produces a redistribution of the world monetary gold stock with the subsequent effect, again with a lag, of bringing the price level in A in line with the price level in B. At that point, equilibrium is restored in the external accounts as well as in the money markets. The world price level would be higher if the gold increase in A represented an increase in the world supply of gold. The original formulation places the entire stress of the adjustment mechanism on the trade account. When capital is allowed to move, there is less stress on the trade account: the outflow of capital brought about by the monetary shock reduces the adjustment in the trade account that would have been required in the absence of capital movements. Several testable implications of this theory emerge: (1) money-supply changes affect the price level with a lag; (2) gold flows are a significant, if not dominant, cause of variation of the domestic money supply; (3) the domestic price level or its rate of change is inversely correlated with the foreign price level or its rate of change; (4) there is a real exchange rate that is serially correlated for relatively long periods of time; (5) a real depreciation of the home currency improves the trade account, which in turn reduces the real depreciation. The version of the monetary approach that assumes perfect international arbitrage (MAPA)-of which McCloskey and Zecher (1976) are ardent proponents-departs from the Humean theory in a fundamental way. Gold flows do not serve to realign country A's price level with the

407

Italy in the Gold Standard Period, 1861-1914

price level prevailing in country B but to restore equilibrium in the money market. Prices of (traded) commodities and assets are determined in the world market. Each country is too small to have a lasting influence on its own price level or interest rates. The law of one price in goods and asset markets prevails. Gold flows are only one of the means to enforce the law of one price; other commodities move from one region to another but gold flows may be quantitatively more relevant because transport costs are smaller relative to bulkier and lower priced goods. The testable implications of MAPA are: (1) purchasing-power parity and interest-rate parity hold in the short run as well as in the long run; (2) the trade account does not respond to changes in relative commodity prices, partly because the law of one price prevents the emergence of such changes and more fundamentally because spending decisions are influenced by changes in money demand and supply only; (3) gold flows are a small source of variation of the money stock, implying that changes in the domestic component of the monetary base dominate gold flows. 9.2.2 The Demand for and Supply of Money In both Hume and MAPA a stable demand for money, influenced by a few variables and in a manner independent of the forces determining money supply, plays an important role. In MAPA an increase in the supply of money relative to demand generates an excess of spending over income whichleads in turn to an outflow of money through a deficit in the balance of payments. The end result is that the monetary shock alters not the total stock of money but its composition between domestic and foreign source components. In Hume the same monetary shock instead affects the total money stock, its composition, and the domestic price level. The two approaches diverge in five respects in their treatment of the supply of money. First, Hume assigns a large role to gold in the money stock process, while for MAPA the role of gold is small. Mo~e to the point, a fractional gold-bullion standard gives the monetary authorities the ability to create monetary-base liabilities against the acquisition of domestic assets, be those claims on government or the private sector. Second, for MAPA, foreign exchange is a significant component of international reserves. For Hume this is not a basic consideration. Third, for MAPA, neither authorities nor the so-called rules of the game playa role, for gold flows as well as changes in foreign exchange are automatically offset by opposite changes in the domestic source components. If a presumptive case for sterilization by monetary authorities can be made, can one discriminate changes in domestic source components of base money that cause opposite movements in foreign source components from sterilization behavior? Fourth, the two approaches differ on the link between monetary policy in Italy and monetary policy in the other

408

Michele Fratianni and Franco Spinelli

member-states of the Latin Monetary Union. Finally, the two approaches assign different roles to the public and banks in the money-supply process. These issues are explored in some detail in the next two sections. We first assess the evidence qualitatively as we guide the reader through the relevant historical account, and subsequently more formally. 9.3

Historical Account

For brief periods in Italy, paper money was convertible at the official price in either gold or silver.! For the most part, fractional reserves of gold and silver bullion were held against paper money (i.e., the monetary base) created by banks of issue and the government. Italy did not develop a single monetary authority until 1926. Up to that time several banks, legally permitted to issue notes, held metallic reserves. Despite the fact that Italy adopted the gold standard intermittently and for brief periods of time, her experience on the whole was not different from what it would have been had she adhered to the standard throughout, particularly from 1900 to 1913. The reason was that Italy's decision makers were aware of the limits of operating on a paper standard. Either they operated responsibly without gold or pulled back from the brink when acting irresponsibly. Briefly put, Italy was guided by the norm of the gold standard. 9.3.1

Competition versus Monopoly of Issue

Upon becoming a unified nation in 1861, Italy inherited the financial structures of the constituent states. Some of these states had banks that possessed characteristics of a central bank; others did not. In brief, there were three banks whose currencies (coins and paper money) were legal tender: the Banca Nazionale (BN) operating in Piedmont and Genoa, the Banca Romana operating in Rome, and the Banca Nazionale Toscana, operating in Tuscany; and six banks whose currency was fiduciary in the strict sense of the word and thus not legal tender (i.e., its acceptance depended on the trust of economic agents): the Banco di Napoli and the Banco di Sicilia, both operating in the Kingdom of the two Sicilies, the Banca Toscana di Credito per Ie Industrie ed i Commerci d'Italia operating in Tuscany, the Banca degli Stati Parmensi operating in Parma, the Stabilimento Mercantile di Venezia operating in Venice, and the Banca Pontificia per Ie Quattro Legazioni operating in Bologna. Two intellectual and political groups competed in Italy to give shape to the country's monetary system. One supported more competition in banking and, in particular, complete freedom of currency issue; the other believed that the political and economic integration of Italy would be enhanced by a strong single monetary authority whose liabilities would be

409

Italy in the Gold Standard Period, 1861-1914

legal tender. The intellectual leaders of the liberal camp (in the English tradition) were the economists Francesco Ferrara (1868), who became minister of finance in 1867, and Antonio Garelli (1879); the leader of the monopoly faction was the powerful Prime Minister Camillo Benzo di Cavour.2 In Parliament the procompetition group usually was stronger than the monopoly-of-issue group. The tug of war between the two opposing camps undoubtedly explains some of the haphazard, if not chaotic, development of banking. For example, in the early 1860s, the senate approved a bill that would have merged BN with the Banca Nazionale Toscana in a single bank of issue, but the lower house of Parliament defeated it. Numerous bills introduced in Parliament would have enlarged the number o~ banks of issue. In addition to the nine banks noted above, a number of firms and individuals printed their own money. However, Parliament put a stop to this practice in 1874 for fear that the bankruptcy of one of them would endanger the entire banking system. In 1893 the Banca Romana failed. BN and the two Tuscan banks merged to form the Banca d'Italia (Bank of Italy). A law of 1894 prescribed that only the Bank of Italy, Banco di Napoli, and Banco di Sicilia had the right to issue currency. This situation lasted until 1926 when the Bank of Italy finally emerged as the institution with a monopoly of issue. In short, though there was intellectual and political opposition to the creation of a single central bank, the belief that a competitive industry would overissue and the failure of a large bank finally prevailed over the opposition to a more centralized and legalistic structure. 9.3.2 The Bimetallic Standard and the Latin Monetary Union In 1862, the newly elected Parliament of Italy approved a currency reform closely resembling that of France. Three options were open to the lawmakers: a gold standard, a silver standard, and a bimetallic standard. Policymakers preferred the gold standard; so did BN which had operated on that principle for several years. However, silver was the dominant money in the South, Lombardy, and Venice. The split between North and South was further complicated by the fact that France, the closest and most important trading partner of Italy, was de jure on a bimetallic standard. We stress de jure because de facto silver had disappeared from circulation in 1848 following the gold discoveries in Australia, California, and Russia? At the official rate of exchange between silver and gold set at a ratio of 15.5 to 1, silver was undervalued relative to market conditions. Therefore, as predicted by Gresham's law, gold drove silver out of circulation. Confronted with the three possibilities, Parliament chose a formal bimetallic standard where silver nominally could be exchanged for gold at the ratio of 15.5 to 1, but in fact, by diminishing the silver content of silver

410

Michele Fratianni and Franco Spinelli

coins of denominations smaller than five lire ,4 established the ratio of 14.38 to I-a ratio that was in line with prevailing market conditions. As to silver coins of larger denominations used in international trade, the 1862 law provided for their issue only upon request; since the official parity undervalued silver relative to the market, the provision served to eliminate large-denomination silver coins. In this manner, Italian policymakers found a compromise between form and the realities of the market. The 1862 law also granted legal tender to the coins of Belgium, France, and Switzerland. In 1865 Italy joined these countries to form the Latin Monetary Union; their coins circulated freely in the union.5 In effect, the union decreed one common money without setting up a common monetary policy. Less than a year after the birth of the union, Italy in 1866 suspended convertibility (see Martello 1881). Italian paper money depreciated relative to gold which was exported to the rest of the Latin Monetary Union. The world demand for gold, however, increased. In 1871 Germany adopted a gold standard; the Scandinavian countries, Holland, and the United States subsequently followed the German example. Concomitantly, the world demand for silver fell, while its supply rose due to new discoveries in Nevada (De Cecco 1984). The market-exchange ratio of silver for gold rose to 18 to 1 in 1876. The Latin Monetary Union, which overvalued silver, was flooded with the metal and by 1878 suspended the free coinage of silver. Although the union had manifestly failed in retaining a bimetallic standard, it was formally kept alive for a few more years. 9.3.3

The First Inconvertibility Period, 1866--84

The state of public finances in Italy in the period from 1861 to 1865 quickly deteriorated (see Fratianni and Spinelli 1982, and figure 9.1). Budget deficits, financed by an increase in public debt, rose over the five years to about 6 percent of average national income. The external accounts over the same period .did not fare any better: the cumulative trade account represented about 4.5 percent of average national income. At the start of 1866 the government announced another large budget deficit, while another war against Austria was imminent. The market perceived Italy to be a decided risk. Bond prices fell dramatically, more so abroad than at home. Foreign-capital inflows dwindled, while domestic capital moved massively abroad, in particular to Paris, to exploit the large profit opportunities from arbitraging differences in bond prices (Majorana 1893). The outflow of gold was so large that some banks were on the verge of bankruptcy. Even the dominant BN had to reduce its portfolio of earning assets to a ~inimum to cope with the liquidity drain. On 1 May 1866, the authorities decreed corso forzoso-the inconvertibility (into gold) of paper money.

Italy in the Gold Standard Period, 1861-1914

411 12

10

8

6

4

2

o -2 1860 1865 1870 1 875 1880 1885 1890 1895 1900 1905 1910

Fig. 9.1

Budget-deficit-to-GNP ratio, Italy, 1862-1913. Budget deficit is the difference between government expenditures and revenues as reconstructed in Pedone 1967, table A.l. Source: GNP at current prices: ISTAT 1957, table 35.

The exchange rate between the lira and the French franc, and the annual rate of change of the money stock, shown respectively in figures 9.2 and 9.3, capture the essence of the aforementioned events. The money stock exploded in 1866; so did the exchange rate. The fact that these two developments accompanied inconvertibility should not be a surprise. Inconvertibility was brought about partly by the injudicious fiscal policy of the government and partly by the war against Austria. Inconvertibility benefited the government by enabling it to monetize the deficit through direct borrowing from the banks of issue. The government's action was backed by the export sector which perceived the devaluation of the lira would stimulate the demand for their products. The 1866 law granted the right to issue inconvertible notes not only to BN and the two Tuscan banks, but also to the Bank of Naples, the Bank of Sicily, and eventually (in 1870) the Roman Bank.6 The major beneficiary of the new law was BN, since its paper money became legal tender throughout the nation. As a result of this provision, BN notes sold at a premium with respect to the paper money of the other five banks of issue. To limit the issue, all six banks had to keep a gold and silver cover of one-third against their note liabilities.

412

Michele Fratianni and Franco Spinelli 116 114 112 110 108 106 104 102 100 98 1865 1870 1875 1880 1885 1890 1895 1900 1905 1910

Fig. 9.2

Lira-franc nominal exchange rate, 1866-1913. Source: Number of lire per French franc (ELF, table 9.A.l, col. 11), from Borgatta 1933.

No sooner was inconvertibility in effect than Parliament contemplated a return to convertibility, creating a commission to study its feasibility (Commissione Parlamentare di Inchiesta 1868-69). Parliament was so eager to act on this matter that a law was passed in 1868--before the commission even produced its report-setting the maximum issue of BN notes at 750 million lire-approximately the amount then in circulation. The growth rate of the monetary base came to a halt in 1868 (see figure 9.3). Monetary discipline lasted less than two years. Budget deficits then rose again. The government desired to firiance the deficits by printing money,7 so BN made loans directly to the Treasury. In return, BN was authorized to print additional currency, ignoring the existing gold-cover requirement. The consequence of these events may be gleaned from figures 9.1 through 9.4. The growth rate of the monetary base took an upward leap, as did the money stock. The lira depreciated drastically against the French franc, reaching its highest level of the sample period. The rate of inflation measured by the annual percentage change of the consumer price index which rose in 1871 reached an all-time peak in 1872 (see figure 9.4). The 1866 law, as noted, gave a competitive advantage to BN, to the

413

Italy in the Gold Standard Period, 1861-1914 35 30

25 20 15 10

5

o -5

, ,

"./

MONEY BASE

"

-10 1860 1865 1870 1875 1880 1885 18901895 1900 1905

Fig. 9.3

1Q1n

Money and money base, annual percent change, 1862-1913. The money base is equal to C. Sources: C = total currency outstanding from De Mattia 1969, tables 5,6,7, and 14; D = total bank deposits from De Mattia 1969, tables 2, 2a, 2b, and 23.

distress of the procompetitive contingent in Parliament. Thanks to their efforts, in 1874 a law was passed designed to eliminate that advantage. The legislation included the following provisions. First, the six banks of issue were asked to repay the outstanding government indebtedness to BN. The repayment changed the ownership of claims on the government. The economic meaning of the transaction is that BN sold its claims on the government to the other banks against an equivalent reduction of BN notes in circulation and an equivalent increase of notes issued by the other five banks. Therefore, while the aggregate amount of notes in circulation did not change, the composition changed against BN notes and in favor of the notes of the other five banks. Second, legal-tender status was granted to the notes of all six banks. Third, the ceiling on currency issue was set at three times each bank's net worth as of the end of 1873.8 Finally, the issue of notes by other institutions was made unlawful. Apparently, the procompetition faction had to concede this point in return for a more equitable market-share arrangement among the six banks of issue.9 By 1874 the government managed to balance its budget (see figure 9.1). The return to fiscal discipline was reflected immediately in low growth of the monetary base. The lira appreciated vis-a-vis the franc by

Michele Fratianni and Franco Spinelli

414 15

(1862 - 1913)

10

5

o -5

-10

-15

-20 1860 1865 1870 1875 1880 1885 1890 1895 1900 1905 1910

Fig. 9.4

Consumer price index, annual percent change, 1862-1913. Source: Consumer price index (table 9.A.l, col. 8) from De Mattia 1977, table 7.

5.5 percent in two years. The rate of change in the price level fell from + 6 percent in 1873 to -16 percent in 1875, in response partly to domestic deflation and partly to the precipitous decline in world prices in 1874 (see figure 9.4). The fall in Italian prices paved the way for serious policy discussions about a return to metallic convertibility. A bill was introduced in Parliament in 1880 to allow the government to borrow abroad up to 644 million lire to repay its indebtedness to the six banks of issue. The repayment would enable the banks to keep reserves only in the form of precious metals and would restore convertibility at a fixed price between paper money and gold and/or silver. Disinflationary expectations were thus promoted. Individuals shifted from goods and real assets to money to capitalize on the anticipated increase of purchasing power. lO Domestic prices fell sharply in 1881 (see figure 9.4). The appreciation of the lira was so pronounced that parity with the franc was restored (cf. figure 9.2). Parliament approved the bill on 1 April 1881. The ensuing sale of government bonds was well received by the market. Since the bulk of the sale was to foreigners, Italy enjoyed a sizable capital flow. The proceeds of the bond sale were used to repay the loans extended by the six banks to

415

Italy in the Gold Standard Period, 1861-1914

the government. Currency was taken out of circulation and the monetary base fell by 8 percent (cf. year 1888 in figure 9.3). 9.3.4 Return to Converbitility in 1884 On 1 March 1883, the government announced that convertibility would be restored on 12 April 1884. Banks were given the option to convert paper money into either gold or silver. Since the official exchange rate between silver and gold had remained fixed at a ratio of 14.38 to 1 and the market exchange rate had risen above 18.5 to 1, convertibility took place only in silver. Again, Gresham's law dictated that cheap money would displace dear money. In addition, Gresham's law-in the context of a world that was operating on a gold standard-implied that gold would be exported and silver imported. In fact, from 1883 to 1885 the outflow of gold coins from Italy rose from 9.2 to 101.3 million lire while imports of silver coins increased from 50.7 to 103.7 million lire (De Mattia 1969, table 10). The course of public finances deteriorated again in the middle of the 1880s (see figure 9.1). The yield on Italian government bonds "Rendita Italiana," quoted in Paris, was higher than its yield at home, reflecting country risk as a factor affecting the market's valuation of Italian bonds. From 1883 to 1888 the lira depreciated marginally but continuously with respect to the French franc. Banks of issue raised the cost of exchanging silver for paper money (Supino 1929, p. 107), while overissuing notesthe Roman Bank was the main culprit-in relation to what was permitted by law. In 1891 the government decided to legalize the currency in circulation in contravention of the law. The aggregate ceiling for the entire system was raised from 755 million to 1064 million lire, which implied a reduction of the gold-silver cover from one-third to onequarter. In 1892 the revelation of the illegal practices of the Roman Bank precipitated a run on the bank that led to its failure. Widespread distrust of the banking system ensued, reflected in a sharp depreciation of the lira on the exchange markets. Banca Nazionale (BN) merged with the two Tuscan banks to strengthen its liquidity position and acquired the name of Banca d'Italia(BI). In the wake of the merger, Parliament passed a law in 1893 which entrusted BI with the task of absorbing the liabilities of the defunct Roman Bank, providing for the substitution of BI notes for Roman Bank notes. The law also made it unlawful for BI, Bank of Naples, and Bank of Sicily to issue notes without direct supervision of the state. "Excessive" circulation was to be reduced over several years, while the reserve requirement in bullion was to be raised to 40 percent.ll Finally, the law called for full convertibility. The economic essence of the 1893 law was that monetary discipline was

416

Michele Fratianni and Franco Spinelli

postponed to the future (Majorana 1893). The call for convertibility was assessed by the market to be wishful thinking. The continuous depreciation of the lira in the exchange markets tells the story very eloquently. The government acknowledged what the market already knew in 1894 when Italy was back to corso forzoso (forced legal tender of the paper currency). 9.3.5 The Second Inconvertibility Period, 1894-1913 This period was the most stable and prosperous of the entire sample period. Budget deficits were first reduced and later converted into surpluses (cf. figure 9.1). Monetary policy was consistently tight. The monetary base grew at a rate below the rate of growth of output. Concomitant with a world boom in economic activity, Italy's real output grew dramatically-at an average annual rate of 2.4 percent from 1897 to 1913, compared to 0.86 percent for the period 1861 to 1896 (cf. figure 9.5). Not surprisingly the lira started to appreciate vis-a-vis the franc in the mid1890s. By 1903 its rate of exchange fell within the gold points and remained there until 1911. The period ended with World War I. In sum, the historical account permits us to draw a few conclusions.

200 (1862 - 1913)

180

160

140

120

100

80

-+----+----+---+----+---+-----+----+----I--~-___4-~

1860 1865 1870 1875 1880 1885 1890 1895 1900 1905 1910

Fig. 9.5

Real GNP, 1862-1913 (millions of 1913 lire). Sources: Real GNP = RYI (table 9.A.l, col. 7), computed as GNP at current prices divided by the GNP price deflator, PYI: De Mattia 1977, table 7.

417

Italy in the Gold Standard Period, 1861-1914

First, the government budget appeared to be responsible for the vicissitudes of the lira and Italy's alternate experiences with the gold standard. Budget deficits caused the abandonment of the gold standard in 1866, the monetary explosion of 186&-67 and 1870-72, and the return to inconvertibility in 1894. As the budget returned to equilibrium, Italy regained monetary and exchange-rate stability-aided by exceptional real growth. Second, the banks of issue tended to overissue. The government induced, tolerated, and legitimized excessive circulation, and possibly was responsible for unlawful note creation that ultimately led to the failure of the Banca Romana. That banks preferred inconvertibility to convertibility and more than once tested the government to see how much they could get away with is supported by some evidence (Supino 1925, p. viii). Third, gold played an explicit role in the period up to 1866; afterwards the system can be best described as a gold-silver-bullion standard. Fourth, the historical account suggests that the monetary base is exogenous relative to prices and income, in light of the tight link between budget deficits and money creation. Finally, the gold standard was not a sufficient condition for stability. Politicians had no difficulties in throwing off the straightjacket of the gold standard when it stood in the way of financing large budget deficits. On the other hand, the proper conduct of fiscal and monetary affairs was sufficient to guarantee stability, whether or not the country was on the gold standard. Indeed, the record indicates that the formal adoption of a standard was not very important. Politicians knew the limits of the paper standard and were willing to return to judicious policies after periods of laxity. It is clear, for example, that little would have changed had Italy returned to a formal gold standard in the first thirteen years of the twentieth century.12 This assessment of the record permits us to treat statistically the periods of inconvertibility as qualitatively similar to the periods of convertibility. 9.4

Evidence

Having outlined the principal theoretical issues in section 9.2 and the broad historical facts in section 9.3, we are now ready to submit more formal evidence concerning the process of money creation, the demand for money, the interaction between the exchange rate and the difference between the domestic and external price level, and more generally the relative explanatory power of the two hypotheses of Hume and MAPA about the balance-of-payments adjustment mechanism. 9.4.1

Data and Sources

The foresight of the Instituto Centrale di Statistica (1STAT) and the Banca d'Italia made possible the quantitative aspect of this study. Some time ago, these two institutions began a painstaking reconstruction of

418

Michele Fratianni and Franco Spinelli

historical time series of the real and monetary sectors of the economy. Three volumes are of particular importance: one by 1STA T (1957) deals for the most part with real-sector data and two by De Mattia (1969, 1977) with monetary statistics. The work of De Mattia is of very fine quality, certainly comparable to the reconstruction of monetary data made for approximately the same period for the United States by Milton Friedman and Anna Schwartz (1963). Many of the time series we used-in particular those we think foreign researchers may have difficulty finding-are shown in appendix A, which also lists definitions and sources. A few words are in order about the definition of the monetary base. While currency holdings of the banking system (i.e., banks other than banks of issue) are known, deposits of the system with banks of issue are not. There is no way to assess the size of this measurement omission, but qualitative accounts suggest it is small. It follows that our definition of the monetary base is equal to the outstanding stock of currency, and that the reserve-deposit ratio is measured by currency holdings of the banking system divided by all private-bank deposits. 9.4.2

The Money-Supply Process

We define the money stock M at time t as (1)

where MB denotes the monetary base and m the multiplier which depends inversely on the currency-deposit ratio (c), and the reserve-deposit ratio (r) (cf. Brunner and Meltzer 1961; Friedman and Schwartz 1963; Cagan 1965): (2)

1+C

m t = - - t· Ct + r t

The relative contribution of m and MB to the growth of the money stock is obtained directly from (1):

(3)

1=

m + MB t

. Mt

Mt

t

'

where a dot above the variable indicates the first difference of its natural logarithm. The first difference of the money multiplier, in turn, can be decomposed as follows:

(4)

m= [In m t t

In (1 + Ct-l)] (Ct-l

+ [ 1n(1 + Ct-l) (Ct-l + r t )

_

+ rt )

1n

]

mt...:...l ,

419

Italy in the Gold Standard Period, 1861-1914

where the terms inside the two bracketed parentheses capture, respectively, the contribution to of the currency-deposit ratio and the reserve-deposit ratio.· Next, the growth rate of the monetary base depends on the weighted average of the growth rates of the domestic- (BD) and foreign- (IR) asset components of the consolidated balance sheet of the banks of issue:

m

(5)

~MBt = (BD) ~BDt MB t - 1 MB t-1 BD t - 1

+ (IR)

~IRt

(MB)t-1 IR t - 1 Finally, the growth rate of the foreign component reflects the growth rate of metallic reserves (IRA) and foreign-exchange reserves (IRFE) , weighted by their initial relative shares: (6)

~IRt = (IRA) ~IRAt IR t - 1 IR t-1 IR t - 1

+

(IRFE) ~IRFEt . (IR)t-1 IRFEt - 1

Table 9.1 shows the basic facts of the process of money-supply creation during the sample period. The first three rows relate to equation (3): column (2) gives the yearly sample means of M, and MB; column (3), the long-run relative contributions of and MB to M; and column (4), the short-run relative contributions. In the long run both and MB turn out to be of roughly equal importance with a slight tendency for the monetary base to dominate. In the short run-and this result is hardly surprising-the multiplier rises in importance relative to MB (lines 1-3). Changes in the currency-deposit ratio, in turn, dominate the short- as well as the long-run behavior of the multiplier (lines 4-5). The domestic component of the monetary base is at least as important for the long-run growth of MB as the foreign component. In the short run, however, matters change dramatically:. changes in BD affect MB negatively, while changes in IR influence MB positively (lines 6-8).' A more detailed breakdown of the data is provided in table 9.2 which pairs yearly changes in BD and IR-the sort of evidence Nurkse (League of Nations 1944) and later Bloomfield (1959) considered to determine whether monetary authorities followed the so-called rules of the game. In the words of Nurkse:

m

m

m

Whenever gold flowed in, the central bank was expected to increase the national currency supply not only through the purchase of that gold but also through the acquisition of additional domestic assets; and, similarly, when gold flowed out, the central bank was supposed to contract its domestic assets also. (P. 66).

m

M 1.96

4.50

Mean Yearly Value (2)

In

Ct - l

Ct - l

rt

-

t

1

)

In m t - l

t

(1 + c+ r

t 1 C ) -

(1 + +

-

0.00

1.96

(~BD/BD_l)

~MB/MB_l

2.77

0.00

100.0

43.6 56.4

Relative Long-Run Contributions of Mean Yearly Values to Growth Rate ofM (3)

(BD/MB)_1 1.56 56.3 8. (!:JR/IR_ 1) (IR/MB)_1 1.21 43.7 Relation of gold and foreign-exchange reserves to foreign component of monetary base 9. (~IR/IR_l) 6.57 10. (~IRA/IRA_l) (IRA/IR)_1 6.04 91.93 11. (~IRFE/IRFE_l) (IRFE/IR)_1 0.53 8.07

6. 7.

Relation of domestic and foreign components to monetary base

5. In

4.. In m t

3. MB 2.54 Relation of currency ratio and reserve ratio to the mUltiplier

1. 2.

(1)

Variables

The Money-Supply Process, 1862-1914

6.65 92.88

-813.1 913.1

9.0

91.0

59.1 40.9

Relative Short-Run Contributions of Mean Yearly Values to Growth Rate ofM (4)

= money stock; m = money mUltiplier; MB = monetary base; c = currency ratio «C-CB)/D); r = reserve ratio (CB/D); BD = domestic-asset component of the base (MB-IR); IR = foreign-asset component of the base (IRA + IRFE); IRA = metallic reserves; IRFE = foreign-exchange reserves; dot above a variable = first difference of its natural logarithm.

Source: Table 9.A.1. Notes: Lines 4-5, 7-8,10--11 sum to 100.00 in cols. (3) and (4) except for rounding errors. M

Table 9.1

421

Italy in the Gold Standard Period, 1861-1914

Table 9.2

Yearly Changes in Domestic and Foreign Component of the Monetary Base, 1862-1914 (millions of lire) BD

IR

1862 1863 1864 1865 1866 1867 1868 1869 1870

101.6 -26.7 -31.8 12.5 447.1 198.7 -81.3 -6.0 39.3

12.4 24.9 -10.7 3.9 -26.9 15.5 91.4 -0.2 20.1

1871 1872 1873 1874 1875 1876 1877 1878 1879 1880

260.9 247.5 103.8 27.5 103.8 -1.0 4.6 -15.6 53.1 -19.8

-24.5 -16.8 4.0 -8.0 -58.6 5.6 0.3 1.5 -3.1 29.8

1881 1882 1883 1884 1885 1886 1887 1888 1889 1890

-42.7 -211.6 -222.6 1.3 52.8 37.3 -2.3 -79.1 40.7 40.7

-40.2 20.9 163.6 48.5 -32.5 6.9 33.6 56.3 3.3 -29.1

Comovement

+ + + + + +

+

+

BD

IR

1891 1892 1893 1894 1895 1896 1897 1898 1899 1900

-49.5 15.1 44.8 -118.1 -63.4 -58.0 64.6 7.3 73.0 -52.0

33.0 4.5 -0.6 90.0 -6.3 28.6 1.1 7.0 -5.1 17.6

1901 1902 1903 1904 1905 1906 1907 1908 1909 1910

-26.4 -8.0 -138.7 7.1 -123.1 43.3 30.6 -117.3 17.3 33.0

26.6 37.2 196.1 23.0 190.6 144.8 206.9 51.1 17.2 14.7

1911 1912 1913 1914

119.7 -35.3 22.8 639.0

66.5 35.4 61.8 79.6

Comovement

+ + + +

+ + + + +

+ +

+ + +

+ +

Source: Table 9.1. Notes: BD = domestic-asset component of the monetary base; IR = foreign-asset component of the monetary base.

There are twenty-four instances in which dBD and dIR move in the same direction, and twenty-nine where they move in the opposite direction. Abstracting from lags, this evidence would indicate, according to Nurkse and Bloomfield, that the five banks of issue on the whole contravened the rules of the game through a policy of sterilizing foreign-exchange flows. However, these twenty-nine observations are also consistent with the monetary theory of the balance of payments (both Hume and MAPA) which predicts that autonomous increases (decreases) in BD cause outflows (inflows) of IR. And indeed there are several historical episodes

422

Michele Fratianni and Franco Spinelli

that we can identify in support of the monetary hypothesis. Take, for example, 1866, 1871, and 1872 when the government ran large budget deficits that were monetized by the banks of issue. The increases in BD were clearly autonomous in the sense that they were not induced by balance-of-payments considerations; yet they produced an outflow of foreign reserves. These outflows are consistent with a condition of excess supply in the domestic money market that finds its way abroad. As another instance, consider 1882 and 1883 when the government deliberately financed its deficits by borrowing abroad. The reduction in BD cannot be interpreted as a violation of the rules of the game, but as a way to restore equilibrium in the domestic money market, while at the same time raising the ratio of foreign reserves to total base money. The key point is that the occurrence of a negative correlation between the two source components of the monetary base cannot be interpreted tout court as evidence in favor of the reserve-sterilization hypothesis. The NurkseBloomfield test ignores the compensatory response of IR to changes in BD. On the other hand, we do not pretend to shed light on the issue of how much of the observed negative correlation between aIR and aBD is attributable to "offset" behavior as opposed to sterilization practices. For that we would require the specification and estimation of the reaction function of the banks of issue (as well as of the underlying macro model), a task we are not ready to tackle.13 Returning to table 9.1, we note that gold flows were the main long-run driving force of the growth of international reserves; as a matter of fact, it was not until 1893 that banks were allowed to count foreign exchange as part of international reserves. But in the short run, once again, the findings are quite different, with gold movements accounting for less than 7 percent of changes in IR (lines 9-11). In sum, the long-run growth of the money stock can be explained by the fall in the currency-deposit ratio (with a relative contribution of 45 percent), the expansion of domestic credit (30 percent), and net inflows of gold (22 percent). The ratio c falls from a value of7.8 in 1861 to 0.35 in 1913, a steady decline interrupted by a sharp upswing in 1866-67 when convertibility was suspended and in 1873 when metal coins appreciated relative to paper money and deposits. The secular decline of c reflects the corresponding secular fall in the cost of maintaining deposit balances relative to currency. The contribution of BD to the growth of M is not surprising in light of section 9.3. Table 9.3 presents, as summary, parameter estimates of money-supply functions for the entire period and the subperiod 1895-1914. These results corroborate our earlier findings, namely, that changes in the multiplier are influenced predominantly by changes in c and that the elasticity of M with respect to MB is unitary. The estimates also suggest that the reserve-deposit ratio plays a more signifi-

423

Italy in the Gold Standard Period, 1861-1914

Table 9.3

Money-Supply Equations (I-statistics in parentheses)

Dependent Variable and Sample Constant Period

m 1863-1914 1895-1914

M 1863-1914 1895-1914

.004 (1.015) -.001 ( -1.679) .008 (1.727) -.001 (.598)

C

;-

-.292 ( -11.039) -.644 ( - 55.468) -.270 (- 9.370) -.618 ( -37.604)

-.022 ( -1.556) -.046 (- 5.042) -.018 ( -1.333) -.039 (- 4.600)

MB

pa

D.W.

R2

1.65

.73

1.08

.99

.913 (18.347) .950 (40.208)

.348 (2.650) .06 (.260) .326 (2.460) .226 (1.001)

1.60

.89

1.47

.99

Source: See table 9.1. Notes: m = money multiplier; M = money stock; c = currency ratio; r = reserve ratio; MB = monetary base; dot above a variable = first difference of its natural logarithm. aFirst-order autocorrelation parameter (Cochrane-Orcutt estimation technique).

cant role in the money-supply process from 1895 to 1914 than in the earlier period. 9.4.3

The Demand for Money

One of us (Spinelli) has demonstrated that there was a well-defined demand-for-money function in Italy during the period under consideration. However, in light of an inadvertent error in the money-stock series that was used and the availability of an expanded data set, we reestimated the demand-for-money function. The model, the statistical procedure, and findings are discussed in appendix B. On the whole, the results do not change appreciably. Money was a "luxury" good over the sample period, a finding that is similar to Friedman and Schwartz's (1963) for the United States (see figure 9.6). As Sylla points out in his comments on this paper, the "secular fall in velocity in Italy, as elsewhere, was in part-perhaps in great part-the result of more and more economic units and activities becoming specialized, commercialized, and monetized. "14 9.4.4 Deviations from Purchasing-Power Parity A cursory look at the lira-French franc exchange rate (EN) confirms that for long periods of time this rate was not contained within the gold points (figure 9.2). In fact, EN was above the upper gold point from 1866 to 1881 and from 1891 to 1901.15 The question posed by this observation is whether the evolution of EN reflects differences in the two countries' price levels; namely, does purchasing-power parity (PPP) hold?

424

Michele Fratianni and Franco Spinelli 5.5 5.0

4.5 4.0 3.5·

3.0 2.5

2.0 1.5 1.0 1860 18651870 1875 1880 1885 18901895 1900 1905 1910

Fig. 9.6

Velocity of circulation, 1862-1913. Velocity is calculated as GNP at current prices divided by M.

A short answer to this question is provided by examining the behavior of the real exchange rate: (7)

where ER PF PI

= the real exchange rate; = the wholesale price index in France 1913 = 100; = the wholesale price index in Italy 1913 = 100.

E R is plotted in figure 9.7. PPP holds if E R stays close to the parity line of 100. Upward sizable deviations from 100 imply a real depreciation of the lira; downward deviations, a real appreciation. By assumption, the 1913 exchange rate is at PPP. The figure tells a lucid story. On the whole, PPP did not hold. For a period of about twenty years, 1866-83, the lira was systematically undervalued relative to PPP. Yet the path of the inflation differential between Italy and France was very erratic (figure 9.8). The lira again became undervalued from 1892 to 1894 in the wake of the failure of the Roman Bank and of generalized mistrust of the banking system. Only the 1900s can be characterized as approximating long-run PPP.

425

Italy in the Gold Standard Period, 1861-1914 140

--r----------------------_

130

120

110

100

90 1865 1870 1875 1880 1885 1890 1895 1900 1905 1910

Fig. 9.7

Lira-franc real exchange rate, 1866-J913. Sources: Nominal exchange rate (ELF) times French wholesale price index from Mitchell 1978, divided by Italian wholesale price index: De Mattia 1977, table 7.

This assessment does not change when we regress changes in the log of EN on changes in the log of PFand PI (see table 9.4). Both the explanatory

power of the hypothesis (R 2 ) and the parameter estimates of the price variables (which a priori should be equal to unity) provide grounds for rejection. We performed modifications of the reported regressionsconstrained estimation on the two inflation variables as well as introducing distributed lags-which however yielded no appreciable change in the statistical results. If PPP cannot explain the movement in the exchange rate, what can? We considered the following model (see Meese and Rogoff 1981): EN=bo+b l (MI -MF )+b 2 CYI-YF ) + b3 (liis,I - liis,F) + b4 (liil,I - liil,F)

(8) where EN

is il Y

+ b s CA I + b6 CA F + e

= number of lire per French franc; = short-term interest rate; = =

long-term interest rate (a proxy for the anticipated inflation rate); real GNP;

426

Michele Fratianni and Franco Spinelli 15

10

5

o

-5

-10 186018651870 187518801885 1890189519001905 1910

Fig. 9.8

Italy-France inflation differential, 1863-1913. Percentage change in the Italian wholesale price index (table 9.A.l, col. 9) minus percentage change in the French wholesale price index.

CA = current-account balance;

dot above a variable = first difference of its natural logarithm; subscripts I and F = Italy, France. Equation (8) is a testable implication of a rather eclectic asset view of the exchange rate, which incorporates the pure monetary model (b 4 = bs = b6 = 0) with short-run deviations from PPP (b 5 = b6 = 0) and possibly long-run deviations from PPP. The interested reader may refer to Meese and Rogoff and the literature cited therein for a deeper discussion of the theoretical and empirical issues underlying equation (8). We fitted equation (8) for both the lira-French franc and the lirapound exchange rates for the period 1882-1913. The results were disappointing: the parameter estimates were either inconsistent with theory or statistically insignificant and often both. A possible difficulty with equation (8) is that the sample period under consideration cannot be characterized as a regime of flexible exchange rates. Often changes in international reserves bore the brunt of the adjustment process. In light of this consideration, we replaced EN with the composite variable, measuring exchange market pressure,

427

Italy in the Gold Standard Period, 1861-1914

Table 9.4

Tests of the PPP Hypothesis (t-statistics in parentheses)

ExchangeRate Changes for Italy vis-a-vis

Sample Period

France

1867-1913

United Kingdom

1882-19131

Inflation Constant

Domestic

Foreign

R2

D.W.

-.002 (- .522) .001 (.166)

.173 (2.192) -.063 (- .666)

-.053 (- .539) .026 (.170)

.10

2.04

.01

2.12

Sources: Lira-franc exchange rate: table 9.A.l, col. (11); lira-pound exchange rate: table 9.A.l, col. (12); wholesale price indexes: Italy, table 9.A.l, col. (9); France, Mitchell 1978; GNP deflator: Italy, table 9.A.l, col. (10); United Kingdom, Lewis 1978. 1Sample period was restricted by data limitations.

used by Girton and Roper (1977) in their study of post-World War II Canada. The outcome was that the coefficient of the Italian-money variable (in this case dBDIMB_ 1 ) turned out to be significant and of the correct sign; for the other parameters, however, there was no improvement. In sum, we can explain (marginally) more of the evolution of the composite variable of EN, but the power of the explanation is far from satisfactory. Again, we pose the question: What drives the exchange rate? Are nominal and real-exchange-rate changes largely unpredictable, or is there a common force underlying the large and persistent deviations from PPP? Our answer is that there is a common factor underlying the movements of the real exchange rate and that this factor can be labelled for short "country risk." More specifically, whenever financial markets perceived that the Italian government was not following prudent fiscal and monetary policies, the markets rated Italian debt instruments as less than risk-free assets. Potential owners of Italian debt instruments demanded a premium for the nonzero probability of a complete or partial default. Consider a model of interest-rate parity (IRP) between debt instruments of wide circulation: (9)

D t = In (1 + if,t) - In (1 + iF,t)

+In E~+l-ln E t

428

Michele Fratianni and Franco Spinelli

where

= yield on Italian bonds at time t (table 9.A.1, col. 13); iF,t = yield on French bonds at time t (table 9.A.1, col. 14); E~+ 1 = the rate of exchange for t + 1 expected at t; Et = the rate of exchange at t (table 9.A.1, col. 11); D measures deviations from IRP (when D = 0, IRP holds perfectly). iI,t

What is of interest here is not so much the absolute value of D, which can be influenced by transaction costs and differences in tax arrangements, but how it changes over time. We interpret these changes as changes in risk premia. Equation (9) was computed using the yield on the Rendita Italiana (table 9.A.1, col. 13), a long-term government bond that was traded both at home and abroad, especially in Paris; the yield on a comparable French government bond (table 9.A.1, col. 14); and assuming perfect foresight in the exchange market, i.e., In E~+1 = In E t + 1.16 Four periods are of interest: 1867 to 1884 when the lira was consistently undervalued relative to PPP; 1885 to 1892 when the nominal exchange rate was within the gold points and the real exchange rate oscillated around parity; 1893 to 1901 when deviations from PPP again became pronounced; and finally 1902 to 1912 when the nominal exchange rate remained within the gold points and the real exchange rate oscillated around parity. The yearly average values for D for these four periods were 2.37 percentage points, 0.02, 1.95, and 0.58. Clearly, large and persistent deviations from PPP were associated with Italian financial assets carrying a higher yield, inclusive of exchange-rate appreciation, than French financial assets. These differences in D are too large and their timing too coincident with the underlying real-exchangerate series to be explained by changes in transaction costs, exchange control, or tax treatment. The underlying force-we posit-is the changing perception by domestic and foreign markets of Italian country risk. 17 Additional episodic evidence can be marshalled in favor of the hypothesis. In 1866 the lira depreciated by eight percentage points vis-a-vis the French franc, yet the Italian price level was below the French one. As noted in section 9.3, two important events took place in 1866: a large budget deficit (figure 9.1) and the decision to make the lira inconvertible. In 1881 the nominal exchange rate appreciated suddenly (figure 9.2)-too suddenly to reflect changes in underlying competitiveness. However, that was the year the Italian government decided to return eventually to the gold standard. The anticipation that fiscal and monetary discipline would ·be restored had an immediate impact on the exchange markets. From 1891 to 1894 the lira depreciated dramatically; during this period uncertainties about presumed banking malpractices culminated

429

Italy in the Gold Standard Period, 1861-1914

with the failure of the Roman Bank. All these historical episodes confirm the more general evidence presented above: It.aly was perceived for long periods of time to be a substantial risk. That risk, in turn, was based on investors' assessment of Italian economic policy as imprudent enough to justify a positive probability of default on its debt instruments. An alternative explanation for the persistence of deviations from PPP could be positive transaction costs thai would inhibit arbitrage in both goods and securities markets. A variable risk premium that causes variations in the real exchange rate does not signify foregone opportunities to arbitrage across goods or between assets and goods. The aim of our tests is to uncover a variable risk premium. They are silent concerning the average transaction cost borne by a purchaser of Italian goods who intends to resell them in France for profit. We have no facts about transaction costs, but the downward trend in the real depreciation of the lira from 1866 to 1883 is consistent with falling transaction costs. The building of railroads, roads, and canals, the development of a merchant marine, and the ongoing process of commercialization that took pace in this period favor an explanation of declining transaction costs. In short, our risk-premium hypothesis does not necessarily imply irrationality in the goods market. 9.4.5

Causality Tests

So far we have identified the determinants of the money market and the exchange rate. In particular, we have demonstrated that PPP did not hold over the sample period, a fact that is more consistent with the Humean view of the adjustment mechanism than with MAPA's. We now want to marshal additional evidence to discriminate more effectively between the two hypotheses. Table 9.5 contains summary information derived by applying Granger-Sims causality tests to variables of interest. The symbol x-->z Table 9.5

A.

Causal Relationships

+

MB t

==-

{ Y,+I ~+ I,

(P,+2)

N t +l

===-n

pt+1,pt+2

B.

pt

c.

15*t

-==

+ +

IRA t + 3

===-

Pt

:::::=-

{ Ij!.A'+2 ~+2

Notes: P = wholesale price index used in B; GNP price deflator used in A and C; Y = real GNP; EN = lira-French franc exchange rate; P* = Sauerbeck price index for Great Britain (proxy for world prices); - = indicates that the variable has been appropriately prefiltered. The reported relationships are statistically significant at least at the 5 percent level.

430

Michele Fratianni and Franco Spinelli

should be read as follows: prewhitenedx, denoted x,causes prewhitened z, denoted Z, in the s.ense that past values of x explain current values of z; better than would ignoring past values of x. Causality in this sense is a matter of timing relationships. The sign ~ indicates bidirectional causality; + or - refer to the sign of the significant cross-correlation coefficient. Finally, the time subscripts denote the "causality lag." For example, changes in the monetary base "cause" changes in the price level with a lag of up to two years. Several observations about the table are in order. First, the monetary base is unresponsive to changes in output, prices, and exchange rates while the converse does not hold. This evidence suggests that the supply of money is independent of the forces that influence the demand for money. Second, changes in domestic and foreign prices take up to two years to affect gold and silver flows-the external imbalance (table 9.5, part A). This evidence is consistent with the Humean adjustment process. Finally, there is bidirectional causality between foreign and domestic prices, a finding that is consistent with the Humean world but not with MAPA. In the end the crucial difference between Hume and MAPA is that while the latter emphasizes changes in the money stock as the variable for restoring equilibrium in the balance of payments, the former recognizes also the importance of relative commodity' prices. This difference can be tested by considering the following equation: EDG

(10)

= bo + bi ESM + b2 E R + b3 g + b4 T + E,

where EDG ESM

= excess demand for goods; = excess supply of money;

g = deviations from trend of real government expenditures; T = deviations from trend of tax rate; E = error term. Assume that both Hume and MAPA share the common vision that g and > 0 and b4 < O. The dispute concerns b 2 which is zero for MAPA and positive for Hume. Note that the dispute is independent of the prevailing-exchangerate regime. At issue is the existence of a real as opposed to a nominal exchange rate and its effect on the goods market. Hume posits that deviations from PPP are part of the engine that drives the adjustment mechanism. MAPA, instead, rules out such a mechanism and goes even further by postulating that PPP holds beyond the "momentary" run. The coefficient b i is positive. An increase in the domestic component of the

T have a role to play in the excess demand for goods with b3

431

Italy in the Gold Standard Period, 1861-1914

monetary base creates excess supply in the money market which spills over into the goods market by raising spending relative to income, which in turn causes a deficit in the trade account. For unchanged values of E R , g, and T, equilibrium in the balance of payments is restored when the outflow of reserves restores equilibrium in the money market and in the goods market. We tested equation (10), using annual data for the period 1867-1914. EDG was proxied by the log of real income minus the log of its trend value; ESM by the residuals of the demand money (cf. table 9.A.2); E R by the log of the real exchange rate shown in figure 9.1; g by the log of real government expenditures minus the log of its trend value; and Tby the log of the ratio of total taxes to nominal GNP minus the log of its trend value. Bothg and Twere lagged one period. The parameter estimates (absolute I-values in parentheses) and other statistics of equation (10) are: 60 = - .007,6 1 = .38, 62 = .22,63 = - .002, 6 4 = - .33, (.86) (2.14) (3.7) (.33) (6.13) R2

= .63, D.W. = 1.77,

which suggests that the data are more consistent with the Humean view of the adjustment mechanism than MAPA's. Not only is arbitrage activity insufficient to equalize the lira price of Italian and French commodity indexes, but changes in these indexes play an important role in reequilibrating the external accounts. 9.5

Summary and Findings

Italy was not on the gold standard except for brief periods of time, and for the most part the lira was inconvertible in either gold or silver, yet fiscal and monetary policies in the latter part of the sample period, especially in the 1900s, achieved stability of nominal magnitudes. Indeed, Italian experience did not differ on the whole from what it would have been had the country adhered formally to the standard throughout. The lira-French franc exchange rate, representative of other exchange ratios, floated more often than not and moved above the upper gold point in twenty-five out of forty-eight years, returning within the gold points during the 1900s when the lira was inconvertible. Deviations from PPP were sizable and long lasting. That persistence cannot be explained by either an Italian inflation rate systematically higher than the world inflation rate or by an expanded-asset theory of exchange-rate determination. Change in the valuation of the lira relative to PPP is best reconciled with the existence of changing perceptions of country risk. Investors demanded a risk premium for holding Italian debt instruments as compensation for a nonzero probability of complete or partial default. That risk, in turn, emerged whenever the market perceived that Italy was engaging

432

Michele Fratianni and Franco Spinelli

in imprudent fiscal and monetary policies; it disappeared when policies reverted to normal. Our evidence in favor of a risk-premium hypothesis does not necessarily imply irrationality in the goods market. Transaction costs on Italian goods, while probably falling from 1866 to 1883, could have been larger than implied by risk-premium calculations. Deviations from PPP were found to have a considerable impact on the goods market and hence on the trade account. The fact that changes in the terms of trade could affect trade flows, and for that matter the very existence of these changes, differentiates the Humean view of the adjustment mechanism from the perfect-international-arbitrage version of the monetary theory of the balance of payments. The Italian evidence supports the former more than the latter. In favor of Hume, one may also note the one- and two-year lag of changes in the Italian price level over changes in the Italian monetary base and the two-year lag of changes in gold flows over changes in the domestic and foreign price levels. On the other hand, the strong form of Hume implies that gold flows are a significant, if not dominant, cause of variations in the domestic money supply. In fact, 45 percent of the long-run growth of the Italian money stock from 1862 to 1914 can be traced to a fall in the currency-deposit-ratio, 30 percent to expansion of the domestic component of the monetary base, and only 22 percent to net inflows of gold. Over the short run, the foreign component, being more volatile than the domestic component, exerted a larger impact on the growth of the money stock. Italy did not have a monopoly bank of issue until 1926. Throughout the period under study several banks were authorized by the government to issue notes. Such an arrangement, however, did not prevent the government from monetizing its deficits. Fiscal and monetary policies were closely intertwined. Yearly changes in the domestic component of the monetary base were in the majority of the cases compensated by oposite changes in the foreign component of the base. Nurkse and Bloomfield would consider these observations as evidence that the authorities did not play according to the rules of the game. Our interpretation is that this negative correlation is often more consistent with.causality running from dBD to dIR, as implied by the monetary approach to the balance of payments, rather than as the reaction of the authorities to dIR. The Latin Monetary Union which Italy joined in 1865 was never an effective constraint on Italian economic policy. Absence of policy coordination doomed the union to failure. Money was as much a luxury good in Italy as it was in the United States during the same period. The demand for real-cash balances responded to changes in real permanent income with an elasticity of about 1.5. The impact on desired real-cash holdings of changes in the opportunity cost of holding such balances is statistically less clear.

433

Italy in the Gold Standard Period, 1861-1914

Appendix A Table 9.A.l

Basic Annual Data Money Stock

(M)

Currency

Bank-held Currency

Bank Deposits

(1) (2) - (3) + (4)

(C) (2)

(CB) (3)

(D) (4)

Metallic Reserves (IRA) (5)

(millions of ·lire) 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903

1179.0 1337.9 1371.5 1336.0 1386.8 1859.5 2041.0 2089.7 2131.5 2432.7 2845.2 3317.0 3339.6 3412.7 3536.7 3613.7 3781.3 3855.5 3960.0 4066.5 4055.2 3968.4 4057.7 4292.5 4561.6 4943.4 5071.2 5107.0 5226.1 5178.2 5110.8 5275.7 5270.6 5186.0 5269.8 5173.2 5317.0 5478.1 5871.4 6009.4 6247.8 6431.2 6850.5

1044.4 1159.3 1157.5 1115.0 1131.4 1551.9 1766.1 1776.2 1770.0 1829.4 2065.8 2296.5 2405.0 2423.7 2468.9 2473.5 2478.4 2464.3 2514.3 2524.2 2441.3 2250.6 2191.2 2241.0 2261.3 2305.5 2336.8 2314.0 2358.0 2369.6 2353.1 2372.7 2416.9 2388.8 2319.1 2289.3 2355.0 2369.3 2437.2 2402.8 2403.0 2432.1 2489.5

n.a. n.a. n.a. 7.4 9.0 9.5 15.9 11.3 12.5 31.1 38.6 66.7 61.7 62.4 55.5 57.7 57.6 55.3 67.6 81.8 80.4 61.8 74.1 76.7 85.5 87.4 95.4 106.4 83.5 78.2 80.4 71.0 64.9 67.9 60.8 65.0 73.3 74.5 83.8 78.0 81.6 87.6 97.5

134.6 178.6 214.0 228.4 264.4 317.1 290.8 324.8 374.0 634.4 818.0 1087.2 996.3 1051.4 1123.3 1197.9 1360.5 1446.5 1513.3 1624.1 1694.3 1779.6 1940.6 2128.2 2385.8 2725.3 2829.8 2899.4 2951.6 2886.8 2838.1 2974.0 2918.6 2865.1 3011.5 2948.9 3035.3 3183.3 3518.0 3684.6 3926.4 4086.7 4458.5

118.0 130.4 155.3 144.6 148.5 121.6 137.1 228.5 228.3 248.4 223.9 207.1 211.8 203.0 144.4 150.0 150.3 151.8 148.7 178.5 138.3 159.2 322.8 371.3 338.8 345.7 379.3 435.6 438.9 409.8 442.8 447.3 446.7 514.2 506.2 516.5 465.5 477.8 469.5 468.4 494.7 528.4 721.6

434

Michele Fratianni and Franco Spinelli

Table 9.A.1 (continued) Money Stock

(M)

1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914

(1) (2) - (3) + (4)

Currency (C) (2) (millions

7327.6 7985.9 8179.8 8960.3 9428.2 9984.5 10564.7 11160.7 11329.9 11849.3 12797.7

2520.1 2587.6 2775.7 3013.2 2947.0 2981.5 3029.2 3215.4 3215.5 3300.1 4018.7

ForeignExchange Reserves (IRFE) (millions of lire (6) 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886

Bank-held Currency (CB) (3) of lire) 102.9 129.6 134.8 178.1 189.4 198.9 198.7 204.1 231.9 229.7 276.1

Bank Deposits (D) (4)

Metallic Reserves (IRA) (5)

4910.4 5527.9 5538.9 6125.2 6670.6 7201.9 7734.2 8149.4 8346.3 8778.9 9055.1

728.0 919.9 1068.6 1270.4 1314.7 1331.7 1344.6 1397.0 1433.9 1495.7 1532.3

GNP Deflator (PYI) (1913 = 100) (10)

Real GNP (RYI) (millions of 1913 lire) (7)

Consumer Price Index (CPI) (1913 = 100) (8)

Wholesale Price Index (WPI) (1913 = 100) (9)

84.1358 87.1822 85.3033 86.4842 87.9622 91.8083 97.4362 99.9044 103.0410 103.4800 108.5480 108.3870 113.1790 110.8970 114.4030 113.0780 113.7130 114.8740 115.0260 119.4100 111.2060 117.5900 115.8470 116.4960 117.9630 122.2680

82.0 82.5 80.1 77.9 76.6 77.4 79.3 82.5 83.0 84.2 86.8 98.1 104.0 106.5 91.2 96.5 100.4 96.7 95.5 99.0 92.6 90.4 87.5 85.8 87.7 87.6

97.58 91.19 87.17 87.17 85.80 89.73 90.18 95.75 89.27 88.54 91.28 99.14 105.08 104.89 92.91 90.04 102.11 98.85 92.82 93.30 87.26 89.56 83.91 80.36 84.67 85.15

73.94 74.74 70.22 70.51 70.86 74.71 76.84 83.68 78.93 78.16 80.14 90.02 98.64 95.90 80.68 81.97 91.59 85.99 89.12 87.12 82.99 85.16 80.90 78.99 84.73 85.28

435

Italy in the Gold Standard Period, 1861-1914

Table 9.A.l (continued)

ForeignExchange Reserves (IRFE) (millions of lire (6) 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914

22.5 24.2 42.5 94.6 89.3 92.5 "111.2 111.5 115.0 117.9 134.9 133.6 129.7 134.8 141.6 141.8 143.6 157.7 156.2 156.2 199.2

Real GNP (RYI) (millions of 1913 lire) (7)

Consumer Price Index (CPI) (1913 = 100) (8)

Wholesale Price Index (WPI) (1913 = 100) (9)

GNP Deflator (PYI) (1913 = 100) (10)

121.1680 120.2110 114.5710 123.0170 126.2310 120.3140 124.5200 122.5690 123.3080 124.9400 119.8550 129.8240 131.2540 137.2890 144.9760 140.8630 148.4830 148.2080 151.6760 149.7520 173.6640 167.6950 179.7000 165.6680 179.2430 182.6290 188.8500 184.5260

87.4 88.5 90.0 93.2 92.9 92.1 90.1 89.7 89.2 88.8 88.6 89.2 87.8 88.2 88.3 87.7 90.3 91.4 91.5 93.2 97.6 96.6 93.9 96.5 98.9 99.8 100.0 100.0

79.41 80.75 85.44 87.64 85.25 81.03 75.96 73.75 77.59 78.16 76.63 78.74 80.75 84.50 84.10 81.30 80.60 77.00 80.30 83.30 89.80 87.40 88.10 88.20 95.30 102.60 100.00 95.80

79.27 78.77 83.66 85.33 87.34 82.06 80.67 78.56 80.40 79.07 80.03 83.96 83.03 84.10 81.31 79.46 "83.08 82.29 82.63 88.60 88.32 86.86 88.67 94.17 96.69 99.42 100.00 98.94

No. of Lire per French Franc (ELF) (11) 1861 1862 1863 1864 1865 1866 1867

n.a. n.a. n.a. n.a. n.a. 107.99 107.37

No. of Lire per British Pound (ELUK) (12) n.a. n.a. n.a. n.a. n.a. n.a. n.a.

Yield on Long-term Government Bonds (percent" per year) Italy (RII) (13)

France (RIF) (14)

6.97 7.26 7.09 7.55 7.83 9.23 9.66

4.38 4.28 4.39 4.54 4.42 4.55 4.41

436

Michele Fratianni and Franco Spinelli

Table 9.A.l (continued)

1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910

No. of Lire per French Franc (ELF) (11)

No. of Lire per British Pound (ELUK) (12)

Italy (RII) (13)

France (RIF) (14)

109.82 103.94 104.50 105.35 108.66 114.20 112.25 108.27 108.47 109.63 109.42 111.19 110.53 100.28 101.26 99.15 100.00 100.38 100.19 100.82 100.98 100.67 101.15 101.55 103.55 107.97 111.08 105.57 107.63 105.14 106.97 107.32 106.44 104.30 101.21 99.95 100.12 99.94 99.94 99.97 100.00 100.42 100.51

n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 25.40 25.55 25.03 25.23 25.38 25.31 25.54 25.57 25.42 25.54 25.65 26.06 27.19 27.94 26.58 27.11 26.45 27.05 27.07 26.77 26.25 25.46 25.15 25.20 25.14 25.15 25.18 25.13 25.29 25.35

9.32 8.95 9.00 7.20 5.98 6.25 6.25 5.88 5.84 5.88 5.60 5.15 4.88 4.87 4.97 4.97 4.61 4.56 4.14 4.46 4.52 4.58 4.60 4.71 4.66 4.63 4.59 4.37 4.35 4.18 4.06 4.01 4.02 3.97 3.91 3.91 3.90 3.82 3.87 3.69 3.63 3.61 3.60

4.27 4.17 4.76 5.51 5.47 5.34 4.90 4.66 4.35 4.27 4.06 3.73 3.56 3.54 3.67 3.83 3.88 3.79 3.66 3.78 3.63 3.51 3.26 3.18 3.07 3.10 2.98 2.96 2.94 2.90 2.92 2.98 2.98 2.96 2.99 3.06 3.11 3.03 3.08 3.16 3.13 3.07 3.06

Yield on Long-term Government Bonds (percent per year)

437

Italy in the Gold Standard Period, 1861-1914

Table 9.A.l (continued)

1911 1912 1913 1914

No. of Lire per French Franc (ELF) (11)

No. of Lire per British Pound (ELUK) (12)

Italy (RII) (13)

France (RIF) (14)

100.52 100.93 101.77 n.a.

25.39 25.47 25.68 26.24

3.67 3.60 3.58 3.75

3.14 3.27 3.44 3.78

Yield on Long-term Government Bonds (percent per year)

Sources: Col. (2), Del Mattia 1969, tables 5-7,14; col. (3), ibid., table 22; col. (4), ibid.,

tables 2, 2a, 2b, 23; col. (5), ibid., tables 11-15,19,23,26,28; col. (6), ibid., table 19; cols. (7)-(10), De Mattia 1977, table 7; cols. (11)-(12), Borgatta 1933; col. (13), De Mattia 1969; col. (14), De Mattia 1977, table 81.

Appendix B

New Estimates of the Demand for Money Function

Despite the research already done by one of us (Spinelli 1980) on the demand for money in Italy for the period 1867-1965, three reasons were persuasive for reestimating the function. First, our money-stock series excludes currency held by banks whereas Spinelli included it by oversight. Second, Spinelli used the official discount rate as a measure of the opportunity cost of holding money. We have uncovered a long-term rate of interest that is in principle superior to the discount rate. Third, we wanted to ascertain the sensitivity of the demand-for-money function to a change in the sample period. The model employed is that used by Spinelli (1980). It postulates that the demand for money depends upon permanent income and the opportunity cost of holding money. Actual real-money balances adjust slowly to the difference between actual and desired real balances; permanent income is a weighted average of past actual values of income. Such a partial-adjustment-adaptive-expectations model can be represented by the following equations: (AI)

Mi= a + bYi + cit; a, b > 0,

(A2)

M t - M t-

(A3)

1";*- Yi-l

1

=

C

< 0;

a(Mi - M t - 1 ) + ut; 0 < a

= ~(1";- 1";~1);

0 <

~ ~

~

1;

1,

where M* is the desired stock of real-money balances, M is the actual stock, y* and Y permanent and actual real income and i the proxy for the

438

Michele Fratianni and Franco Spinelli Money Demand (t-statistics in parentheses)

Table 9.A.2 a

b

c

- .925 ( -1.700)

.868 (2.320)

- .138 ( -1.160)

D.W. .016 (5.938)

.968 (5.309)

.278 (2.562)

.99

2.106

Sources: Real income is GNP at constant prices (RYI, table 9.A.l, col. 7). Real permanent income is solved out by the model. The rate of interest is the yield on the long-term government bonds Rendita Italiana (RII, table 9.A.l, col. 13).

opportunity cost of holding money. The model is estimated for the period 1867-1914. Money data were centered at midyear and deflated by the consumer price index. All relevant variables were deflated by population. A time trend was added as a regressor; its coefficient is denoted by t. The functional specification is logarithmic, except for the time variable. We employed a nonlinear OLS estimation procedure which is subject to the well-known single-equation bias. However, we believe this bias is quantitatively small and not worth pursuing here (cf. Laidler 1977a). Calliari, Spinelli, and Verga (1981) show that simultaneous-equations estimates for Italy are about the same as single-equation ones. The parameter estimates are given in table 9.A.2. The permanentincome coefficient, b, is not significantly different from unity. To correctly evaluate the inflationary impact of monetary growth, however, we cannot overlook the role of the time variable which is strongly significant and has an elasticity of about 0.5. The relative interest insensitivity of the demand for money confirms the earlier results of Spinelli (1980). The reader is cautioned not to draw the conclusion that velocity is unresponsive to interest rates. Changes in transitory income may be regarded as a proxy for interest-rate changes. To the extent that the demand for money responds to permanent income, the estimates are compatible with a positive relationship existing between velocity and interest rates (cf. Laidler 1977b). The adjustment between actual and desired money balances is completed within a year; learning takes longer.

Notes 1. The U.S. Commission of Gold and Silver Inquiry (U.S. Congress 1925, pp. 347-52) has a compact but useful history of this period. 2. The economist R. Busacca (1870) also supported the monopoly thesis. 3. In addition to the increase in gold supply, the increase in the market price of silver relative to gold possibly resulted from the shift of the source of European cotton imports from the Civil War-ravaged United States to Far Eastern countries with silver standards (De Mattia 1959).

439

Italy in the Gold Standard Period, 1861-1914

4. Gold lire and gold francs had equivalent gold content. Silver lire of denominations of five and above and silver francs also had equivalent silver content. 5. There was a condition: Within two years after its secession from the union, a member-state pledged to repurchase, with gold or foreign exchange, small-denomination coins circulating outside its territory. 6. The five minor banks of issue received BN notes for an amount equal to their metal holdings. Since this transaction did not alter the aggregate amount of currency in circulation, the law can be explained as a governmental scheme to raise the market share of BN, thus paving the way for BN as the only bank of issue. 7. The then-Minister of Finance Sella stated several times that so long as the prices of bonds were low, it was preferable to monetize the deficits. 8. The net worth of each bank was as follows (in million lire): BN 450, Banca Nazionale Toscana 63, Banca Toscana di Credito 15, Banca Romana 45, Banco di Napoli 146.25, and Banco di Sicilia 36. 9. The intended effects of the legislation were not fully realized. The national branch network of BN and the regional character of the other five banks of issue preserved the premium of BN notes over their notes. 10. The immediate impact of a change in regime on the exchange rate is not limited to the Italian experience. British economic history offers similar evidence. 11. Paper money in circulation was to be reduced from 1097 to 864 million lire. As to the reserve requirement, the law prescribed that gold had to account for 33 percent and silver for 7 percent of note liabilities. There was no reserve requirement against deposit liabilities. 12. This point is often stressed by Triffin (1964). 13. See Herring and Marston 1977 for an approach of this sort applied to Germany of the 1960s. 14. Velocity of circulation fell by 131 percent from 1861 to 1913 (figure 9.6). 15. The observation for 1865, which we could not discover, is presumed to be 100. In the empirical work this observation is omitted. 16. An obvious criticism is that a careful application of equation (9) requires one-year rather than long-term bonds. Unfortunately, we had no alternative to our procedure. 17. Lindert's evidence (1969) of asymmetries between the interest differentials required to attract funds to major money-market centers may be in agreement with our hypothesis; e.g., the interest differential required to attract funds from small countries to London was greater than that required to attract funds to Paris. Our evidence suggests that if Lindert's asymmetries incorporate differences in country risks, they are far from being constant over time. It may well be that France was preferred to Italy at all times-an inference consistent with the data reported above-but this difference was not time independent. In sum, the variability of Lindert's asymmetries is explained by a variable risk premium against the lira.

References Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard, 1880-1914. New York: Federal Reserve Bank of New York. Borgatta, Gino. 1933. Bilancia dei pagamenti. Milano: Giuffre. Brunner, Karl, and Allan H. Meltzer. 1966. A credit-market theory of the money supply and an explanation of two puzzles in U.S. monetary policy. In Essays in honor of Marco Fanno: Investigations in economic theory and methodology, ed. Tullio Bagiotti. Padova: Cedam.

440

Michele Fratianni and Franco Spinelli

Busacca, R. 1870. Studio sui corso forzoso dei biglietti di banca in Italia. Firenze: Tipografia della Gazzetta d'Italia. Cagan, Phillip. 1965. Determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. Calliari, Sergio, Franco Spinelli, and Giovanni Verga. 1984. The demand for money in the Italian economy: A survey of the literature and new estimates. Manchester School of Economics and Social Studies 54 (forthcoming) . Commissione- Parlamentare di Inchiesta. 1868-69. Relazione della Commissione sui corso forzoso dei biglietti di banca. Firenze. De Cecco, Marcello. 1984. International gold standard: Money and the empire. New York: St. Martin's Press. De Mattia, Renato. 1959. L'unificazione monetaria italiana. Torino: I.L.T.E. - - - . 1969. I bilanci degli istituti di emissione italiani 1865-1936. Roma: Banca d'Italia. - - - . 1977. Storia del capitale della Banca d'Italia e degli istituti predecessori. Roma: Banca d'Italia. Ferrara, Francesco. 1868. Del corso forzato e della maniera per abolirlo-. Firenze: Nuova Antologia. Fratianni, Michele, and Franco Spinelli. 1982. The growth of government in Italy: Evidence from 1861 to 1979. Public Choice 39 (no. 2): 221-43. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. - - - . 1982. Monetary trends in the United States and the United Kingdom: Their relation to income, prices, and interest rates, 1867-1975. Chicago: University of Chicago Press. Garelli, Antonio. 1879. Le banche. Biblioteca dell'Economista. Serie 3, vol. 6. Torino: UTET. Girton, Lance, and Don Roper. 1977. A monetary model of exchange market pressure applied to the postwar Canadian experience. American Economic Review 67 (Sept.): 537-48. Haugh, L. D. 1976. Checking the independence of two covariance stationary time series: A univariate residual cross-correlated approach. Journal of the American Statistical Association 71 (June): 378-85. Herring, Richard J., and Richard C. Marston. 1977. National monetary policies and international financial markets. Amsterdam: NorthHolland Publishing Co. 1STAT. 1957. Indagine statistica sullo sviluppo del reddito nazionale dell' Italia dal1861 a11956. Annali di Statistica Serie 8, vol. 9. Roma. - - - . 1976. Sommario di statistiche storiche dell'Italia 1861-1975. Roma: Istituto Poligrafico di Stato. Kreinin, Mordechai E., and Lawrence H. Officer. 1978. The monetary approach to the balance of payments: A survey. Princeton Studies.in International Finance, no. 43. Princeton: Princeton University Press.

441

Italy in the Gold Standard Period, 1861-1914

Laidler, David. 1977a. The demand for money: Theories and evidence. New York: Dun Donnelley. - - - . 1977b. Demand management in Britain from a monetarist viewpoint. Paper presented at the Conference on Demand Management, London. Lewis, W. A. 1978. Growth and fluctuations, 1870-1913. London: Allen and Unwin. Lindert, Peter H. 1969. Key currencies and gold, 1900-1913. Princeton Studies in International Finance, no. 24. Princeton: Princeton University Press. Majorana, Giuseppe. 1893. I dati statistici nella questione bancaria. Roma: Loescher. Martello, Tullio. 1881. L'abolizione del corso forzoso. Venezia: Visentini. McCloskey, Donald N., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance o/payments, ed. J. Frenkel and H. G. Johnson. Toronto: University of Toronto Press. Meese, Richard, and Kenneth Rogoff. 1981. Empirical exchange rate models of the seventies: Are any fit to survive? International Finance Discussion Paper no. 184. Washington, D.C.: Board of Governors of the Federal Reserve System. Mitchell, B. R. 1962'. Abstract of British historical statistics. Cambridge: Cambridge University Press. - - - . 1978. European historical statistics, 1750-1970. New York: Columbia University Press. Nurkse, Ragnar. 1944. International currency experience: Lessons o/the inter-war experience. Princeton: League of Nations. Pedone, Antonio. 1967. II bilancio dello stato e 10 sviluppo economico Italiano: 1861-1963. Rassegna Economica 67 (Mar.-Apr.): 285-341. Pierce, D. A., and L. D. Haugh. 1977. Causality in temporal systems: Characterizations and survey. Journal of Econometrics 5 (Aug.): 265-93. Spinelli, Franco. 1980. The demand for money in the Italian economy, 1867-1965. Journal of Monetary Economics 6 (Jan.): 83-104. Supino, Camillo. 1929. Storia della circolazione cartacea in Italia dal1860 a11928. Milano: Societa Editoriale Libraria. Triffin, Robert. 1964. The evolution ofthe international monetary system: Historical reappraisal and future perspectives. Princeton Studies in International Finance, no. 12. Princeton: Princeton University Press. U.S. Congress. Commission of Gold and Silver Inquiry. 1925. European currency and finance. Report prepared by John Parke Young. 67th Cong., 4th sess. Washington, D.C.: Government Printing Office.

442

Michele Fratianni and Franco Spinelli

Comment

Richard E. Sylla

Fratianni and Spinelli's paper in my view has two objectives. The broader of the two objectives is to sketch the historical facts and quantitative considerations that are pertinent to what might be called "A Monetary History of Italy, 1861-1914." This sketch appears to be modeled on the Monetary History of Friedman and Schwartz. It is filled with fascinating interpretations and hypotheses. One hopes that it is preliminary to a full-length study. The second objective, more specific and in keeping with the theme of this conference, is to ask the question, Does Italian experience, 1861-1914, support the classical analysis of Hume and others, or does it lend more credence to the more recent "monetary approach" developed by a number of writers but applied most prominently to economic history in the work of McCloskey and Zecher? The answer that Fratianni and Spinelli give is almost certain to attract a lot of scrutiny because it runs counter to the results of many other tests of the two theories. Their answer holds that at least for Italy in the period studied, the price-specie-flow analysis is more supported than the monetary approach and purchasing-power parity. My comment covers each of the two objectives. Italian Monetary History, 1861-1914 For most of the period, as Fratianni and Spinelli point out, Italy was not on the gold (or silver, or bimetallic) standard in the strict sense because Italians could not convert their paper bank notes and bank deposits into gold (or silver) at fixed rates. Nonetheless, eve~ during the long periods of inconvertibility (1866-84, 1894-1913) the Italian banks were required by law to maintain a fractional specie cover for their note issues (but not deposits), the fraction being one-third from 1866 to 1891, one-quarter from 1891 to 1893, and two-fifths after 1893. Fratianni and Spinelli describe the periods of inconvertibility as a "gold-silver bullion standard" in which the requirement of a specie cover for bank notes exerted enough discipline (they presume) to allow the periods to be treated as one for econometic analysis with the periods of de facto gold standard (1861-66) and de facto silver standard (1884-94). Indeed, one of the main lessons they draw from their study of the Italian experience is that a standard or lack thereof is not very important: The "gold standard is not a sufficient condition for stability" because politicians had no difficulties in throwing off the straightjacket of the gold standard when it stood in the way of financing large budget deficits. On the other hand, the proper conduct of fiscal and monetary

Richard E. Sylla is professor of economics at North Carolina State University.

443

Italy in the Gold Standard Period, 1861-1914

affairs was sufficient to guarantee stability, whether or not the country was on the gold standard ... little would have changed had Italy returned to a formal gold standard in the first thirteen years of the twentieth century. (P. 417) Most economists and politicians probably would agree in general with these views. Fratianni and Spinelli's description and analysis of the proximate determinants of the Italian money supply is a welcome addition to the growing literature on quantitative monetary history. In the proximatedeterminants framework, the long-term growth of money in Italy was the result in roughly equal measures of a rise in the monetary base and a rise in the monetary-base multiplier. The base, defined as the currency stock, grew in two ways, also of roughly equal importance. It grew first as the banks issued currency to buy domestic assets, with issues to purchase the government securities generated by budget deficits being the main irregular force in this process, according to Fratianni and Spinelli. And it grew, secondly, as currency was issued to buy "foreign" assets, which were mainly metallic reserves in the long run. Changes in the domestic and foreign components of the base were in opposite directions in more than half of the years studied, but Fratianni and Spinelli are not sure whether this finding should be interpreted to mean that Italian banks sterilized specie flows, thereby contravening the so-called rules of the game, or whether the finding is an example of one source of base offsetting the other, as either Humean or monetary-approach analysis might predict. The lack of a central monetary authority and-most of the time-a convertible currency raises questions about how applicable the rules of the game were in Italy. The money multiplier contributed almost as much as a rising base to long-term money growth. The main reason for the rising multiplier was that the Italian public changed from holding most of its money in the form of currency in 1861 to holding most of its money in the form of bank deposits in 1913. The development of banking and a growing appreciation of the conveniences of bank money were as characteristic of Italy as of other modernizing economies during the nineteenth century. Whether money was a luxury in Italy, as Fratianni and Spinelli say is implied by their analysis of money demand, may be doubted. The secular fall of velocity in Italy, as elsewhere, was in part-perhaps in great part-the result of more and more economic units and activities becoming specialized, commercialized, and monetized. Economic historians can teach economists that there are better ways to describe, for example, the declining share of nontraded agricultural products in GNP in a developing economy than as "a decline in velocity" or money as "luxury." To teach these lessons, however, we have to do some hard work on the relative importance of production in the money-using and nonmoney-using sectors of developing economies.

444

Michele Fratianni and Franco Spinelli

Purchasing-Power Parity (PPP) Since the lira was not convertible at a fixed rate into gold for most of the 1861-1914 period, it is hardly surprising that "for long periods of time [the lira-French exchange] rate was not contained within the gold points." The lira depreciated with respect to the franc when Italy abandoned convertibility because the politicians found that convertibility impeded the ability of Italian banks to create new money to finance the public sector's deficit. When suspension of convertibility allowed the new money to be created, advocates of PPP would expect Italian prices to rise relative to French prices and the exchange rate to rise in order to equalize exchange-adjusted prices in France an~ Italy. Fratianni and Spinelli say that it didn't happen that way. There was a substantial real, not merely nominal, depreciation of the lira from 1866 to 1883 and again from 1892 to 1894. As an aside one might note that this story, if true, confirms the alleged irrational attachment of the French to gold: they kept their gold and paid high French prices when they could have used the gold to buy lira and then lira to buy lower-priced Italian products. But is the story true? Fratianni and Spinelli argue that it is true, and they attempt, after trying some tests that are inconclusive, to explain the deviations from PPP by introducing the hypothesis of "country risk": Whenever financial markets perceived that the Italian government was not following prudent fiscal and monetary policies, the markets rated Italian debt instruments as less than risk-free assets. Potential owners of Italian debt instruments demanded a premium for the nonzero probability of a complete or partial default. (P. 427) In support of their hypothesis they introduce evidence that "large and persistent deviations from PPP are associated with Italian financial assets carrying a higher yield, inclusive of exchange-rate appreciation, than French financial assets." This financial theory of deviations from PPP merits closer examination~ As the lira depreciated relative to the franc, allegedly because of the Italian government's untrustworthy fiscal and monetary policies, one would expect the franc price of Italian debt to fall accordingly to a new equilibrium reflecting the new exchange rate. But Fratianni and Spinelli appear to argue that the process of arbitrage didn't stop at this point. Rather, the (all powerful?) French investors marked down the Italian debt instruments still further be~ause they expected the Italian policies to become still worse. With interest arbitrage one would predict that Italiangovernment debt would decline in price (and yields would rise) in Italy. Or, without instant arbitrage, the French would sell their Italian bonds back to Italians until they fell enough in price to equalize the exchangeadjusted prices in the two countries. In either case the French investors supposedly enforce a real depreciation of the lira and create all those

445

Italy in the Gold Standard Period, 1861-1914

opportunities for French goods buyers to increase their incomes by buying Italian goods. This analysis is curious because it appears to assume that the rationality that characterizes French investors is not shared by French goods buyers. If we assume, on the other hand, that rationality is evenly distributed across French persons, then the case for deviations from PPP collapses. In such circumstances, the suspicions of French investors would lead only to a rise in yields on Italian debt relative to yields on French debt, and this rise is precisely what the evidence brought forward by Fratianni and Spinelli indicates. "Country risk" remains intact as an explanation of relative interest-rate movements between the two countries, but it does not seem to be able to account for deviations from PPP. For country risk to perform the latter feat, Fratianni and Spinelli need to demonstrate that French goods buyers were irrational and they do not do this. Indeed, they say very little about the trade account or even capital flows despite the predictions that a more fully developed analysis of their country-risk hypothesis generates with respect to these variables. If PPP is not refuted by the country-risk hypothesis, how might one account for the behavior of the real exchange rate calculated by Fratianni and Spinelli? It is, of course, calculated by multiplying the nominal lira-franc exchange rate by the ratio of French to Italian wholesale price indexes, with each index set at 100 in 1913. If one plots the two price indexes, it becomes apparent that they are very close together from 1885 to 1913, but that before 1885 the French index is well above the Italian one with a noticeable tendency for the two to converge in the 1870s and early 1880s. The behavior of the two indexes is consistent with a hypothesis that international transactions costs (the costs of transportation, information, and so on) declined from the 1860s to the 1880s~ And, as a matter of economic history, the completion of the Italian railroad network in the 1880s, the decline of ocean freight rates, and the spread of telegraphic and other information networks are often cited by historians as key developments in this period. So the data are not inconsistent with the view that Italian prices were world prices, adjusted to take account of transactions costs. Nonetheless, pending further study, an agnostic position on instantaneous PPP versus price-specie flow seems· warranted. The GrangerSims causality tests reported by Fratianni and Spinelli support the Humean approach, but now that such tests have been used to demonstrate that fluctuations in U.S. GNP have caused variations in sunspots, one might be a little skeptical (Sheehan and Grieves 1982). (Moreover, Fratianni and Spinelli provide no information on the statistical significance of the "causal relationships" in their table 9.5.) The results of Fratianni and Spinelli's estimation of equation (10) are more troublesome for advocates of the monetary approach because the statistically

446

Michele Fratianni and Franco Spinelli

significant coefficient on the real exchange rate indicates that (1) rises in French wholesale prices, ceteris paribus, increased demand for Italian goods; (2) rises in Italian wholesale prices, ceteris paribus, reduced demand in Italy; and (3) rises in the nominal lira-franc exchange rate, ceteris paribus, raised demand in Italy, presumably by making French imports more costly. One may not be persuaded by Fratianni and Spinelli that Italian goods were persistently undervalued for almost two decades, but the authors at least cast doubt on the notion that instantaneous arbitrage worked to bring about purchasing-power parity at all times in Italy during the 1861-1914 period. Reference Sheehan, R. G. and R. Grieves. 1982. Sunspots and cycles: A test of causation. Southern Economic Journal 48 (Jan.): 775-77.

General Discussion of Jonung and Fratianni-Spinelli Papers commented on an important paradox raised by the conference-that the rules of the game are frequently violated by most of the participants but nevertheless the system functioned remarkably well and that there was no major financial crisis within the system over thirty or forty years. He submitted the following thoughts on this issue. During the gold standard era we observe a huge variety and array of adjustment mechanisms involving allocation of resources between production and trade. Also we observe changes in relative prices, changes in the allocation of real capital, changes in the shares of traded and nontraded goods, changes involving long-term and short-term capital, etc. All these operations and transactions involve a wide spectrum of information and transaction costs, and they all occur in the context of a variety and in response to a variety of shocks-nominal shocks, real shocks, and shocks with variable durations, i.e., more or less transitory shocks and more or less permanent shocks. Shocks also occur in the context-and that is really the basic theme for our purposes-in the context of well-established expectations that the system will be maintained, that the central bank will honor the gold standard and will honor the buying and selling of gold at the stipulated prices. This context suggests that we would not always expect to see the price-specie-flow mechanism involving relative price changes operate. In the Dornbusch-Frenkel case, which is a classic case of real transitory shocks, everybody would understand that a bad harvest is a transitory

BRUNNER

447

Italy in the Gold Standard Period, 1861-1914

event. Under such circumstances we would not expect the Humean mechanism to operate. We would see adjustment mechanisms that operate in the range of information and transactions costs that are very small-exactly what we find in the Dornbusch-Frenkel piece. On the other hand, in the case of more permanent phenomena of the kind we saw in the gold standard era, in countries such as Sweden, Germany, Switzerland, and the United States, where we observe a massive transformation from a predominantly rural society to a highly industrialized one, expressed by a similar pattern of real goals and supplemented by a similar pattern of monetary goals and no fiscal policy, the price-specie-flow mechanism probably helped. Thus it depends very much on the mixture of shocks that occur, to what extent the various mechanisms operate, and whether the adjustment takes place primarily in traded goods, financial markets, or substantial changes in relative prices. LINDERT, following his comment on Jonung's paper, made some remarks on the self-destruction of a successful gold standard. His argument was that if a gold standard were successful, the main visible symptom of that success without any particular causal modeling would produce a key-currency system. This outcome occurs because people recognize that some currencies are as good as gold; once they are as good as gold, why ship or hold as a large share o( reserve backing a barren, unproductive, and non-interest-earning metal? Thus a successful gold standard will lead to a key-currency system. But this phase is only the first, and if it were the only one, little more would need to be said. The distinction between a key-currency system and a gold system is merely semantic, which may be sufficient in a world of contract enforcement and deposit insurance where the person, agency, or nation issuing an obligation to the rest of the world must back it up in the metallic way if a crunch comes. However, a second phase would almost surely have to come eventually in any kind of twentieth century that we can imagine. Sooner or later the key-currency country will be subjected to a foreign-sector shock that will require deflation to hold the domestic money supply down to match the external demand for her money. Now if the shock is more severe and long lasting than say the 1847 harvest failure, would the key-currency-country officials be expected to deflate enough to meet the shock? Lindert answered in the negative-sooner or later the authorities will jettison the system. With respect to the Bretton Woods system, with or without Vietnam, America seemed headed on a path that would eventually require her to waive the rules, e.g., as soon as there was a serious deflationary bid imposed by the gold-rattling French and others. Interwar Britain came to a similar end. Such an outcome would have happened to pre-WW1 Britain. Even if World War I had not disrupted the entire

448

Michele Fratianni and Franco Spinelli

world, there is reason to doubt how long Britain would have held out in any case. As he (Lindert) has shown elsewhere, by 1913 Britain had enormously high external liabilities relative to any measure of her gold reserve, and that ratio looks high even by the post-World War II dollar standard. Lindert argued that even without war in August 1914, there were plenty of gathering clouds that would have made anyone question convertibility. The growing public awareness of the relevance of a decline in the money supply on the real economy, Lloyd George and the people's budget, and a considerable shift in political power-all of these factors would have made somebody question whether simple convertibility would have dominated British policy forever. Moreover, a rising foreign share of world output and foreign competition, in addition to domestic pressure, would have led to the abandonment of full convertibility. fRENKEL made several comments. He referred first to Jonung's correlations of prices of wheat and rye within Sweden and across countries. High correlations, he suggested, do not necessarily imply unified markets. Similar price movements might simply reflect a similar response to the same climatic conditions in otherwise unconnected markets. Frenkel's second remark concerned Fratianni's paper. He argued again that the Humean approach, which places great weight on relative price changes, and the purchasing-power-parity approach, which permits no such relative price changes, are both consistent with the monetary approach to the balance of payments. Frenkel's last point had to do with interest parity. Fratianni attempts to relate deviations from purchasing-power parity to deviations from interest-rate parity. His results indicate the existence of large deviations from interest-rate parity. However, because he has no data on the forward market for foreign exchange, he is forced to use the future spot rate as proxy for the forward rate. Yet, on the basis of our experience with forward markets in the 1970s, we now know that the current forward rate can be a poor predictor of the future spot rate. As a matter of fact, since the spot rate follows approximately a random walk, the current forward rate and the current spot rate are almost identical. Therefore, using the future spot rate as the measure of the current forward rate may indicate large deviations from interest parity that are not really there. BORDO commented that the similarity between the experience of Italy and Argentina is striking. The parallel illustrates that the gold standard only seemed to work well for relatively stable economies. In response to a conjecture made by Peter Lindert that the gold standard would have been abandoned because of the inevitable conflict between deflation and convertibility, Bordo described a simple counterfactual experiment he had conducted. He asked what would have been the behavior of the U.S. price level and real output had she followed classical gold standard rules throughout the post-World War II period. Assuming a fixed ratio of the

449

Italy in the Gold Standard Period, 1861-1914

monetary base to the monetary gold stock, a fixed ratio of the U.S. monetary gold stock to the world monetary gold stock, a fixed ratio of the world monetary gold stock to the world total gold stock, but allowing the money multiplier to vary as it actually did, Bordo constructed a hypothetical gold-based U.S. money supply. He then used St. Louis-type price and nominal-income equations to simulate the behavior of the price level and real output since 1960 under a gold standard regime. He found exactly what one would expect-the classical gold standard pattern of a stable trend in prices, surrounded by alternating shorter periods of inflation and deflation, and alternating short-run movements in real activity. He conjectured that faced with such a pattern, the United States would have inevitably left gold. BARRO asked the authors whether they had attempted to include in their money-demand functions measures of the extent of monetization? JONUNG described his attempts to incorporate variables measuring monetization, financial sophistication, and the growth of the welfare state in money-demand functions. These attempts employ, as a proxy for monetization, the number of inhabitants per bank office, which fell from 30,000 people per bank office in Sweden in 1871 to a low in 1922 with 4,500 people per bank office, and is now 5,500 per bank office. A number of other proxies for financial sophistication and monetization was used. FRENKEL questioned whether inhabitants-per-bank-office is an adequate measure of financial sophistication. FRIEDMAN described his and Anna Schwartz's attempts to introduce a measure of financial sophistication in the determination of the velocity of circulation in the United States before World War I. None of their proxies proved successful. Friedman also made a remark concerning the proper way to test purchasing-power parity. He argued that it was not advisable to rely on the correlation of individual prices, but that it was necessary to construct frequency distribution of prices. One might take the price of wheat in Sweden and the price of wheat in Great Britain and divide one price by the other to generate a wheat exchange. Similarly, the ratio of rye prices could be used to generate a rye exchange. Taking as many identical commodities as possible, one can construct a frequency distribution of relative prices and study their behavior over time. This method would be a way of constructing a measure of changes in the degree of market integration over time. MCCLOSKEY responded to Friedman's suggestions about purchasingpower parity, pointing out that the origin of modern notions of purchasing-power parity was the use by Cassel, Keynes, and others of purchasing-power parity as a guide to the exchange rate governments should adopt after exchange rates had been floating for a considerable period of time. Price indexes were commonly used in these early calculations.

450

Michele Fratianni and Franco Spinelli

McCloskey suggested that there can be drift in purchasing-power parity for a number of different reasons. Even when arbitrage in traded goods is highly efficient, nontraded goods cannot be directly arbitraged. Productivity changes between traded and nontraded sectors can lead to changes in equilibrium exchange rates and in the purchasing-power parity index. By implication, under certain conditions purchasing-power parity cannot be used as an accurate predictor of equilibrium exchange rates. But this implication does not call the underlying concept into question. Just as we hold constant income and tastes in any analysis of demand, so too when we talk about purchasing-power parity we should hold constant other things that are themselves unaffected by the particular experiment we are trying to perform. For example, with regard to Fratianni's paper, McCloskey suggested that a problem with figure 9.7 is that the apparent deviation from purchasing-power parity might well be explained on productivity grounds and not be a deviation from purchasing-power parity at all. KOCHIN emphasized the importance of the distinction between periods of irresponsible fiscal policy and wartime exigencies. Borrowing during wars will cause capital inflows which would raise the purchasing power of the Italian lira against the French franc. This circumstance could explain some of the deviations from purchasing-power parity observed in Fratianni's paper. JONUNG responded to Peter Lindert's comments by suggesting that the gold standard was beneficial for Sweden in a number of respects. In particular, the standard made economic planning easier. In comparing Sweden with Italy, Jonung emphasized the importance of recognizing that Sweden was a smaller country than Italy. And Sweden was closely connected to Great Britain. In such an environment, the relationship of Sweden to Britain is not unlike that of the State of N ew York to the rest of the United States. FRATIANNI raised a number of issues in his reply. One dealt with deviations from purchasing-power parity and the real exchange rate. He suggested that some of the disagreement among discussants may be terminological. What some label the real exchange rate, others label deviations from purchasing-power parity. As Richard Sylla points out, a number of factors can cause the equilibrium real exchange rate to change over time: differences in productivity in Italy and France, changes in production processes, and changes in tastes. In response to Frenkel's comments, Fratianni suggested that he had not attempted to exclude either the Humean mechanism or the perfectarbitrage version of the monetary approach to the balance of payments. He agreed that the monetary approach does not rule out by definition changes in relative prices. The two views differ only in emphasis: one

451

Italy in the Gold Standard Period, 1861-1914

emphasizes relative prices; the other, changes in the level of expenditure. The paper takes extreme versions of the two theories in order to delimit the relevant issues as clearly as possible. Fratianni also cautioned that the 1970s are a very different period than the 1880s and 1890s. National economic policies and inflation rates differ more now than then. Therefore it may not be advisable to generalize concerning current policy on the basis of nineteenth-century experience.

PART

IV.

International Linkages under the Gold Standard

10

The Gold Standard and the Transmission of Business Cycles, 1833-1932 Wallace E. Huffman and James R. Lothian

Descriptions of the gold standard have stressed two very different aspects of that monetary system. Modern observers, concerned with high and rising rates of inflation, have written enthusiastically and often nostalgically of the longer-term price stability that existed during the gold standard era. Many other economists during the past century and a half, however, have rendered a less kindly judgment, emphasizing instead the frequent and sometimes severe business contractions that characterized the period as well as the substantial shorter- and intermediate-term swings in the price level. Irving Fisher (1920, p. 65), for instance, phrased his criticism thus: The chief indictment, then of our present [gold] dollar is that it is uncertain. As long as it is used as measuring stick, every contract is necessarily a lottery; and every contracting party is compelled to be a gambler in gold without his own consent.... One of the results of such uncertainty is that price fluctuations cause alternative fluctuations in business; that is, booms and crises, followed by contractions and depressions. The objective of this paper is to investigate the incidence of cyclical fluctuations within countries adhering to the gold standard and the transmission of these fluctuations among countries. In investigating these The authors' affiliations are Iowa State University and Citicorp Investment Bank, respectively. Moses Abramovitz, Michael Bordo, Michael Connolly, Barry Falk, Arthur E. Gandolfi, Thomas Huertas, Robert Lewis, Anna J. Schwartz, and Harvey Segal provided useful comments on various aspects of this paper. Richard M. Timberlake, Jr. read and commented in detail on an earlier draft of the historical section. The authors would like to thank all of these individuals without implicating them in the end result. In addition, the authors wish to thank Donna Bettini, Connie McCarthy, and Mark McNulty for extremely able assistance.

455

456

Wallace E. HutTman and James R. Lothian

topics we first review each of the important cyclical contractions in the United Kingdom and United States during the century 1833-1932. We then present the results of more formal tests of hypotheses about the causes of such contractions and their dissemination across countries. The basis of these tests is a vector autoregressive model estimated for both countries for the combined subperiods 1837-59 and 1882-1914. The main objective of the historical narrative is to see whether a monetary explanation of the business cycle is at least broadly consistent with the data for the two countries. To do so we analyze the movements in the U.K. and U.S. money and gold stocks, the apparent causes of those movements, and their relationships to one another and to output over the cycle. In the course of this analysis, we track over territory touched upon to varying degrees by a number of other authors. Insofar as possible, we have tried to integrate their accounts with ours. Our analysis, however, differs from most of these earlier analyses both in its breadth of coverage, spanning both the United Kingdom and the United States and a century of data, and in its emphasis, being concerned almost exclusively with cyclical fluctuations and with monetary, as opposed to credit or interestrate, data. The vector autoregressive model and associated hypothesis tests are direct complements of the historical narrative. They enable us to evaluate in a more rigorous fashion the apparent relationships uncovered by the simpler historical approach. Again our chief concerns are the association of monetary shocks and cyclical declines in output within each of the two countries and the strength of possible alternative channels of transmission between the two countries. The latter include specie flows, price and interest-rate arbitrage, asset-market adjustments, and direct absorption effects. Since the historical and econometric sections contain separate summaries of results and the last section of the paper an overall summary, we skip a detailed synopsis at this juncture. Instead, we merely mention the two principal findings: monetary shocks were the main source of cyclical fluctuations during this period, and the monetary system itself-the gold standard-was the main mechanism through which the shocks and associated fluctuations in output were disseminated. 10.1

Historical Overview

At the start of our sample period, the United Kingdom was a large country, London the main financial center of the world, and the Bank of England a central figure in international monetary activity. The United States, in contrast, started the period as a significantly smaller economy. During most of the nineteenth and early twentieth centuries, however,

457

The Transmission of Business Cycles, 1833-1932

the United States grew rapidly. Immigration rates were high, except for during the Civil War and major economic depression years, and the frontier moved steadily westward. As a result, by 1914 the U.S. net national product was about three times that of the United Kingdom versus roughly three-quarters that of the United Kingdom in 1834. In the eighteenth century, the United Kingdom and most other countries had been on a bimetallic standard, primarily gold and silver. The United Kingdom restored specie payments in 1821 after the Napoleonic Wars and remained on the gold standard continuously through 1914. Then in 1915, with the economic and financial disruptions of World War I, the United Kingdom left the gold standard and in its stead adopted a managed fiduciary standard that lasted until the middle of the next decade. The United Kingdom returned briefly to gold in 1925, this time a gold-exchange standard, but that system was short lived. In 1931, faced with the massive balance-of-payments deficits engendered by the deflation then underway in the United States, the United Kingdom left gold for good. The United States came to the gold standard later than the United Kingdom (1834), but stayed on it two years longer. Like the United Kingdom, the United States too had a temporary break with gold, the episode beginning in 1862 after the start of the Civil War and lasting de facto until 1879, de jure until 1900. Gold during those years remained an official currency along with the greenbacks issued to finance the war. The United States was in effect on a dual monetary standard with the price of one currency, greenbacks, in terms of another, gold, determined by the market. And since gold remained the international currency, flexible exchange rates prevailed between the United States and the rest of the world. Only after the United States deflated its price level did convertibility of the dollar with gold at the pre-Civi"War parity become possible. The international gold standard that the United States and United Kingdom participated in during the period 1834-1914 was a mixed rather than a pure gold standard. Under the latter, the only money in use is gold coins or notes backed by 100 percent gold reserves, and gold is transferred between countries to meet balance-of-payments obligations. The modified gold standard of 1834-1914, however, had many of the features of a fiat currency system: domestic central-bank operations, international reserve currencies, and domestic fiduciary monies that functioned as substitutes for gold coins. Nonetheless, the monetary systems were operational gold standards whether pure or not. Under the modified gold standard, central banks engaged in openmarket operations of buying and selling domestic securities. Some, like the Bank of England, reputedly "played by the rules of the game," permitting the domestic money supply to adjust-in the direction required

458

Wallace E. Huffman and James R. Lothian

for long-run international economic equilibrium.1 Other central banks, though, frequently followed temporary policies of sterilizing gold flows, buying or selling domestic securities, and hence changing the domesticcredit component of the money supply to offset the monetary effects of such flows in the short run. Over the longer run, however, the ability to intervene was necessarily limited unless, of course, as often happened in time of war, a country left gold and thereby let its exchange rate float. Under this system, the Bank of England maintained its reserves in gold, but most other countries held their reserves in gold and sterling assets. Thus, balance-of-payments adjustments could be made by transferring currencies and titles to securities and gold in financial centers rather than by shipping gold per se. Given that London was the world financial center and that sterling was a reserve asset, the Bank of England could have a significant effect on money supplies abroad via its openmarket operations and manipulations of Bank rate. 10.2 Theoretical Considerations As an empirical proposition, the link between money and business fluctuations has long been known to exist. Well before our own era, monetary economists such as David Hume, Henry Thornton, and Irving Fisher took this association as a datum, second in importance perhaps only to that between money and the price level. These writers, moreover, seem to have been well aware of the apparent contradiction between the two relationships. One of the questions they, like modern economists, sought to answer was how changes in the stock of money, a nominal variable, could in the long run affect only the price level, another nominal variable, but in the short run affect output and employment, real variables. The distinction made by Fisher, for one, to rationalize these seemingly anomalous effects, was between the expected and the unexpected effects of monetary changes: unexpected changes giving rise to "money illusion" and thereby impinging upon output and employment. In the past two decades, Milton Friedman (e.g., 1968) has used a similar line of reasoning. Output in this view will fall below its permanent level or unemployment rise above the natural rate as a consequence of some economic agents' inability to see through monetarily induced expenditure and price changes to their ultimate source. In the empirical implementation of this model, a sudden change in the nominal stock of money or in the price level (or in their rates. of change) is, therefore, the causative variable.2 Over the past decade, this approach has been extended and otherwise recast by proponents of the rational-expectations hypothesis. In these models, economic agents as a general proposition are posited to take

459

The Transmission of Business Cycles, 1833-1932

account of more than simply the past behavior of money or the price level in forming their expectations. They are assumed instead to know the structure of the relevant economic relationships and to make unbiased forecasts of the relevant economic variables. In empirical applications of this rational-expectations approach, output or unemployment depends upon deviations in money (or other variables) from the values individuals predict on the basis of that knowledge. Until very recently, models of. this sort, with their emphasis upon expectations and dynamic adjustment, were almost exclusively applied to closed economies. The standard models of open economies and international adjustment that dealt with behavior of output were all in the Meade-Mundell tradition-static rather than dynamic and devoid of any distinction between actual and anticipated values? In the past several years, however, the two strains of the literature have begun to merge. Michael Darby and Alan Stockman (1983) have estimated a simultaneous model for the United States and seven other industrialized countries during the Bretton Woods era that is consistent with a natural-rate-rational-expectations approach. And Nasser Saidi, in two separate theoretical papers (1980, 1982), has applied a rationalexpectations model to questions of international transmission under both floating and fixed-exchange-rate regimes. Underlying our empirical analyses of U.S. and U.K. business cycles under the gold standard is a set of maintained hypotheses of a similar sort. For each country the proximate determinant of output fluctuations was sudden, unanticipated changes in domestic monetary variables. Transmission between countries occurred mainly via specie flows and the monetary reactions they induced, either on the part of the monetary authorities or on the part of the banking system. An unanticipated decrease in monetary growth in the United Kingdom, for example, initially reduced output growth in the United Kingdom, raised (real) interest rates, produced downward pressure on the rate of rise of prices, and induced a balance-of-payments surplus and hence inflows of specie and capital from the United States. Monetary growth in the United States decreased as a result of the specie outflow, the real rate of interest rose, and output growth and the rate of rise of prices fell. After the shocks worked their way through both economies, output in each returned to a level consistent with its permanent rate of growth, real interest rates to their initial levels, and the nominal stocks of money to levels consistent with worldwide monetary equilibrium. Part of the adjustment to the initial monetary deceleration could also have occurred via price and interest-rate arbitrage. Whether the former in turn had a depressing influence on output would depend, however, upon the underlying model. If price shocks rather than monetary shocks

460

Wallace E. Huffman and James R. Lothian

affected aggregate supply, then price arbitrage would be a channel through which monetary disturbances in one country could have real effects in another.4 The alternative view is that cyclical fluctuations in the two countries resulted from some common real shock. According to this explanation, contractions in the money stock were an effect rather than the cause of the declines in income. Declines in American and British real output due, say, to decreased demands for their exports on the part of other nations led to deficits in the balance of payments, gold outflows, and declines in the nominal stocks of money in· the two countries. The role of financial panics-an integral part of the history of the period-also differs according to the two sets of hypotheses. Under the first, it was purely monetary. Panics were shocks largely if not completely unrelated to prior income movements. They affected output only via their impact on the nominal stock of money. Under the second, the reverse held. Panics resulted from prior declines in income or one of its components and were a method by which the requisite reduction in the nominal stock of money was produced.5 In pure form, the two sets of hypotheses are, therefore, competing. In actuality, one can easily envision a more complex situation, feedback from income to money, or vice versa, also being of some importance in the one case or the other. 10.3 Historical Evidence on the Cyclical Behavior of Money and Output The National Bureau of Economic Research's chronology of reference cycles serves as a convenient point of departure for discussion of the cyclical contractions in the two countries. For the United States, this chronology begins in 1834, the start of our sample period; for the United Kingdom-actually Great Britain-it begins forty-three years earlier. Table 10.1 lists the calendar-year reference-cycle dates for the two countries, starting with 1836, the peak in both countries for the first full contraction encompassed by our data, and ending with the Great Depression of the 1930s. In the United Kingdom over this period there were nineteen reference-cycle contractions. In the United States there were either twenty-five or twenty-three depending upon the treatment of the contractions of 1847-48 and 1892-94. If viewed as distinct entities, as the official NBER classification does, there were twenty-five. If, however, we combine the first with the earlier contraction of 1845-46 and the second with that of 1890-91, which is done in the table and which may make more sense from the standpoint of intercountry comparisons, the total for the United States reduces to twenty-three. One aspect of these data that has attracted attention is the tendency for

461

The Transmission of Business Cycles, 1833-1932

Table 10.1

U.S. and U.K. Calendar-Year Reference-Cycle Dates

---------------------- Peak ----------------------

--------------------Trough --------------------

U.K.

Both

U.K.

U.S.

1836 1839

1837 1842

1838 1843 (1846)

U.S.

1845 1854

1848

(1847) 1853

1855 1856 1860

1866

1864 1869

1883

1882 1887

1873

1857 1862 1868 1870 1879 1886

1890 1900

(1892) 1895 1899

1901

1858 1861 1867 1878 1885 1888 (1891)

1904 1908

1910

1911 1913

1918

1924 1927

1923 1926

1894

1896 1900

1903 1907 1917

Both

1914 1919 1921

1920 1926 1928

1924 1927

1929

1932

Source: Burns and Mitchell 1946. Note: Parentheses indicate NBER reference-cycle contractions in the United States, sub-

sumed in our analysis into a longer corresponding cycle for the two countries.

the U.K. reference cycles to lag slightly those in the United States. Judged in terms of the yearly dates, the lag for peaks and troughs combined is approximately four-tenths of a year. The popular interpretation of this lag views it as indicative of a systematic causal relationship running from the United States to the United Kingdom. We present evidence later on, however, that contradicts this interpretation, particularly as it applies to the cyclical contractions prior to the Civil War. Before we turn to that evidence, however, it may be useful to examine the output and monetary data themselves. To that end we present tables 10.2 and 10.3 in which we detail the movements in the neighborhood of reference-cycle peaks in the United States and United Kingdom, respectively, of business activity and of two monetary variables, the monetary gold (or total specie) stock, and either the M2 definition of the overall

Precontraction

1834-1836 1838-1839 1843-1845 1846-1847 1851-1853 1855-1856 1870-1873 1879-1882 1885-1887 1888-1890 1891-1892 1894-1895 1897-1899 1900-1903 1904-1907 1908-1910 1911-1913 1915-1918 1919-1920 1921-1923 1924-1926 1927-1929

Contraction (peak-trough)

1836-1838 1839-1843 1845-1846 1847-1848 1853-1855 1856-1858 1873-1878 1882-1885 1887-1888 1890-1891 1892-1894 1895-1896 1899-1900 1903-1904 1907-1908 1910-1911 1913-1914 1918-1919 1920-1921 1923-1924 1926-1927 1929-1932

Periods Precontraction 24.2 2.4 10.0 13.6 14.1 12.9 6.1 15.3 7.1 7.1 8.2 3.4 13.6 9.1 7.6 7.7 5.4 13.9 11.5 5.4 6.2 2.0

Contraction -0.1 -6.8 4.3 -2.6 5.1 0.1 -0.5 2.9 2.7 4.0 -1.7 -1.8 8.0 6.3 -1.4 5.7 4.1 14.9 -5.8 5.3 2.4 -11.9

-9.8 -5.4 1.0 -6.9 1.4 -7.1 3.7 0.0 -3.4 4.0 -5.1 -2.6 3.2 -2.2 -11.4 1.7 -10.4 4.0 -4.3 2.8 1.0 -14.5

Money

Contraction

Real Income

9.2 1.8 -1.0 -2.1 13.0 6.2 10.3 5.2 8.0 -8.7 -4.5 -8.6 6.4 6.6 9.9 6.9 1.1 -1.3 13.5 10.1 3.2 -7.1

Contraction

Gold

18.1 3.3 -0.7 15.0 15.1 11.8 -14.5 26.4 5.3 -0.9 1.9 0.0 18.7 6.8 6.7 0.6 3.3 15.5 -8.2 10.5 -0.4 -3.1

Precontraction

Rates ~f Change of Real Income, Money, and the Monetary Gold Stock before and during Reference-Cycle Contractions in the United States, 1834-1932

the difference between the natural logarithms of the terminal and initial values by the time interval between them.

Source: See appendix A. Notes: Real income pre-1870 is an index of total trade, thereafter NNP. Money is M2. Rate of change is from initial to terminal dates; computed by dividing

Table 10.2

Precontraction

1835-1836 1837-1839 1842-1845 1851-1854 1855-1857 1858-1860 1864-1866 1871-1873 1881-1883 1888-1890 1898-1900 1901-1903 1906-1907 1911-1913 1915-1917 1919-1920 1922-1924 1926-1927 1927-1929

Contraction (peak-trough)

1836-1837 1839-1842 1845-1848 1854-1855 1857-1858 1860-1862 1866-1868 1873-1874 1883-1886 1890-1894 1900-1901 1903-1904 1907-1908 1913-1914 1917-1919 1920-1921 1924-1926 1927-1928 1929-1932

Periods Contraction 1.2 0.8 -3.5 7.2 4.1 1.8 1.7 3.7 1.0 1.9 0.9 -1.2 0.8 8.5 15.6 -2.3 -0.2 2.1 0.6

-1.5 -2.4 2.7 2.7 --2.3 1.5 2.3 2.8 0.8 0.9 2.4 1.3 -1.0 0.7 -6.9 -4.9 0.6 1.6 -1.9

Money

Contraction

Real Income

0.9 -4.3 5.5 8.3 -2.5 2.0 5.0 7.5 2.1 3.7 3.0 0.4 2.1 3.7 14.3 10.4 -3.0 1.5 1.4

Precontraction 3.7 3.9 -2.8 6.3 8.7 3.4 2.6 0.8 -0.3 4.9 2.5 0.6 1.7 8.8 22.4 9.2 7.8 3.5 -4.0

Contraction

Gold

-3.8 -5.4 9.3 8.7 -2.6 -0.3 4.2 2.8 -2.9 0.2 -0.2 0.5 3.4 0.1 n.a. 30.3 0.1 1.1 -1.3

Precontraction

Rates of Change of Real Income, Money, and the Monetary Gold Stock before and during Reference-Cycle Contractions in the United Kingdom, 1834-1932

Source: See appendix A. Note: Real income is real GNP. Money is high-powered money pre-1871 and M2 thereafter. A proxy series for money for the pre-1871 period-net public liabilities of joint-stock banks-showed the following peak-to-trough average annual rates of change (in percent): 1845-48, - 5.8; 1854-55, 14.1; 1857-58, -12.5; 1860-62,8.1; 1866-68, - 6.8. Rate of change is from initial to terminal dates; computed by dividing the difference between the natural logarithms of the terminal and initial values by the time interval between them.

Table 10.3

464

Wallace E. Huffman and James R. Lothian

money stock or, in the case of the United Kingdom prior to 1871, high-powered money. Judged on the basis of these data, severe business contractions were a common occurrence in both countries, in the United States even more so than in the United Kingdom. During such episodes, output generally contracted sharply in absolute terms. (Appendix B lists severe contractions.) In many of the milder NBER reference cycles, however, the movements are virtually imperceptible: real output actually increased and at an average rate close to its secular rate of growth. The contraction that occurred in those episodes was in the rate of growth relative to the rate in the previous expansion phase rather than in the level of output or in the rate of growth relative to its secular average. The most striking feature of the data is the clearcut association between decreases in the rate of growth of money (or high-powered money) and cyclical fluctuations in output. In the great majority of cycles in both countries, the monetary stringency preceded or was coincident with the downturn in output. The degree of stringency, moreover, in general conformed to the severity of the cycle. The gold stock often exhibits the same general patterns as M2. The movements in gold, however, sometimes failed to account for anything close to the full movement in M2. Furthermore, in several instances there was little or no correspondence between the two. In many of these episodes, as the narrative below indicates, the cause of the monetary decline was a financial panic that reduced the ratios of M2 and highpowered money to gold. To investigate these relationships further we turn to the analysis of severe individual cyclical contractions in the two countries, neglecting mild cyclical contractions. We divided the seven episodes and the accompanying narrative into four parts based on their chronological ordering. As it turned out, these groups are also of some economic significance, with the direction of transmission between the two countries differing considerably among the groups. In the antebellum period, the United Kingdom appears to have exerted the predominant influence. By the early twentieth century the situation was reversed-the United States becoming the senior partner in the process, the United Kingdom the junior. 10.3.1

Antebellum Cycles

The four major antebellum business contractions with which we deal are those of 1836,1839, 1845, and 1857. All four were relatively severe in at least one of the two countries. Most of these severe contractions, moreover, were accompanied by substantial monetary decelerations. And all provide evidence of a causal relationship running primarily from the United Kingdom to the United States.

465

The Transmission of Business Cycles, 1833-1932

1836 According to Burns and Mitchell (1946), the first of the two cyclical contractions that marked the second half of the 1830s began in both the United States and the United Kingdom in 1836, ended in the United Kingdom in 1837 and in the United States a year later. In both countries, monetary factors appeared to have played an important role, with declines in either the stock of specie or its rate of growth taking place at the onset of the business declines and a further panic-induced decrease in the U.S. money stock accompanying the more protracted and more severe drop in output there (see appendix B for further discussion). This latter monetary contraction, moreover, appears attributable in large part to the restrictive policies of the Bank of England, themselves in turn the result of the Bank's reaction to the drain of specie. By all the measures we examined, the cyclical contraction in the United States was relatively severe: Smith and Cole's (1935) separate domesticand foreign-trade indexes fell by average annual rates of 5.5 percent and 16.5 percent, respectively, between 1836 and 1838; Ayres's (19"39) index of business conditions at an annual rate over the same period of 8 percent; and Gallman's (1966) real-capital-formation series by 13.8 percent between 1837 and 1838. In the United Kingdom, the contraction was not only of shorter duration but also apparently much milder, real GNP falling by 1.5 percent from 1836 to 1837 and then rebounding by 5.6 percent the next year. Growth in the U.S. money stock began to decline prior to the cyclical peak and then turned negative: from an increase of 31.5 percent in 1835-36, to 16.9 percent in 1836-37, to - 3.4 percent in 1837-38. Specie accounted for all the change in the rate of change of money between 1834 and 1836. The next year, as a result of the banking panic, a decrease in the ratio of M2 to specie became of primary importance. For the United Kingdom, we have data only for specie and for highpowered money. The total monetary specie stock exhibits a substantial decline in each of the years from 1834 through 1836, as do the specie holdings of the Bank of England, the more accurately measured component of that total. High-powered money, after declining in 1834 and 1835, increased by just under 1 percent in 1836 and just over 1 percent in 1837. According to John Francis's (1862) account, the loss of specie by the Bank prior to 1836 was a reflection of overseas investments gone sour. The Bank's specie stock, which in 1833 had reached a high of £10.9 million, fell from an average of £8.2 million in mid-1834 to an average of £6.2 million in mid-1835. Then in the first quarter of 1836 the Bank's holdings temporarily rose, only to resume their decline a quarter later as pressure from the United States developed. The Bank's reaction-belatedly, in the eyes of some contemporary

466

Wallace E. HutTman and James R. Lothian

observers-was to increase its discount rate from the 4.0 percent that had prevailed for close to a decade to 4.5 percent in July 1836 and then to 5 percent in September. At the same time, the Bank imposed quantitative controls, refusing to discount the bills of joint-stock banks or to handle acceptances of Anglo-American discount houses (Matthews 1954, p. 58). The first signs of a financial crisis in the United Kingdom came with the suspension of payments of the Agricultural and Commercial Bank of Ireland in November 1836 and the near demise of the Northern and Central Bank, a recently formed Lancashire joint-stock bank. The panic in the United States began in the spring of 1837 with the failure of a New Orleans bank. A run soon developed on New York banks and payment was suspended in May of that year. What heightened the monetary effects of these actions was the legislation enacted by most states that prohibited banks that had suspended payments from expanding their note and deposit liabilities. "Under these conditions," Clark Warburton (1962) has claimed, "suspension of specie payments provided relief from an immediate banking panic but led to a curtailment of bank loans and discounts and contraction of bank supplied circulating medium." Considerable debate has centered around the exact events that triggered the U.S. crisis; the specie circular, the actions of the Bank of England, and the sharp decline in the price of cotton all figure prominently in the various explanations offered. Those who have emphasized the first, moreover, ascribe crucial importance to it as a cause of the business contraction itself. The monetary data belie that explanation. As we have shown, the first year of the cycle in the United States was accompanied by a decline in monetary growth that was wholly due to a decline in the rate of growth of the monetary specie stock.6 That decline in turn was the result of the Bank of England's restrictive posture and one of the causes of the ensuing banking crisis. What added to the pressures on the banking system and indeed may have been the key exacerbating element was the disbursement of the Treasury surplus to state treasuries and hence drain of specie from the banking system (Timberlake 1978). As the U.S. money stock fell, the economy deteriorated further. That of the United Kingdom, which has escaped the contractionary monetary effects of the panic, recovered.

1839 The depression of 1839 was one of the most severe on record in both countries. Real GNP in the United Kingdom fell for three years running for a total decline of 7.2 percent, making the depression comparable in both magnitude and duration to that of 1929-32. In the United States the contraction lasted a year longer, and, as near as one can tell, was equally sharp. Smith and Cole's total trade index fell by 21.4 percent from its

467

The Transmission of Business Cycles, 1833-1932

peak in 1839 to its trough in 1843; their domestic index fell by 10 percent over the same period (12 percent from 1839 to 1842); Gallman's capitalformation series fell by 26.3 percent; and Ayres's index of business activity declined by 22.0 percent. The only difference, other than duration, between the U.K. and u.s. contractions was that the latter appears to have been made up of two separate episodes: All four real series for the United States show a substantial drop from 1839 to 1840, a slight pickup over the next year (next two in the case of capital formation), and then in two of the remaining three instances a further decline of roughly the same magnitude as that of 1839-40. As in 1836, monetary fluctuations appear to have played important causative roles in the two countries. In the United Kingdom, both gold and high-powered money reached peaks in 1838, gold declining by 11.9 percent per annum over the next two years and high-powered money by 6.6 percent per annum. Then between 1840 and 1841 the U.K. gold stock reversed direction, increasing by 4.4 percent, while high-powered money remained roughly constant. In the United States, the monetary contraction began a year later than in the United Kingdom. Gold fell by 6.9 percent and M2 by 11.1 percent between 1839 and 1840 and continued to decline the following year, though at slower rates. In 1841-42, the decline in M2 accelerated and the gold stock fell somewhat further. By the time the trough in both monetary series had been reached,the gold stock had decreased by a cumulative total of 12 percent and M2 by a cumulative total of 32 percent. The only comparable period of monetary contraction in the hundred years that our data span is the Great Depression of 1929-33. The lag between monetary changes in the two countries at the beginning of the cyclical declines suggests a chain of causation that ran from the United Kingdom to the United States. Historical accounts buttress this conclusion. In early 1839, the Bank of England began to experience another specie drain. The cause, according to Matthews (1954), was an increase in expenditures on imports of grain, due in turn to a crop failure the year before. A contributing factor, according to some commentators, was a lack of trust on the Continent in the Bank's ability to maintain specie payments. The Bank reacted to the outflow by raising its discount rate in May of 1839 from 4 percent to 5 percent. By that time its specie reserve had been almost halved, from £9.0 million in January to £5.0 million in May. In late June it raised Bank rate further to 5.5 percent and finally in the beginning of August to 6 percent. As a result of these actions, out-and-out panic never really took place in the United Kingdom. More harmful repercussions of the Bank's actions were, however, felt in the United States. Interest rates rose markedly, Bigelow's commercialpaper-rate series showing an increase from 6 percent in January to 15

468

Wallace E. Huffman and James R. Lothian

percent in August. At the same time, banks in the United States were losing specie. In July, the Bank of the United States, by then a Pennsylvania-chartered bank, began experiencing trouble. By early October it failed and a run on Philadelphia banks began. They suspended payments in response and banks in the South and West soon followed suit. The New York banks held out, but according to the state bank commissioners cited by Sumner (1896), they experienced a $20 million decrease in their liabilities in the space of three months ending in late January 1840. 1845 According to the National Bureau's chronology, the United Kingdom in the second half of the 1840s experienced a three-year contraction, lasting from 1845 to 1848, and the United States two one-year contractions spaced one year apart. Output data, however, tell a different story. In the United Kingdom, real GNP increased by 6.4 percent from 1845 to 1846 before slowing to an average annual growth rate of slightly less than 1 percent the next two years. In the United States, Smith and Cole's trade indexes show peaks in 1844, slight declines between 1844 and 1845, and then offsetting increases the next year. The level of the total index (the combination of domestic and foreign) was the same in 1847 as 1844; the·domestic index alone, the same in 1846 as 1844. Between 1847 and 1848 both indexes then decreased substantially, the total by 6.9 percent and the domestic alone by 5.7 percent. Gallman's capital formation series, after rising by 25.7 percent from 1845 to 1846, shows a 1.1 percent increase during the next year and then a 7.4 percent average annual rate of decline the following two. In both countries, therefore, the pattern is similar even though the reference-cycle chronology differs. Whatever contraction took place in 1845-47 was relatively mild. Over the next year, the situation worsenedin the United States apparently by a considerable degree. The monetary data are in rough agreement with the movements in output. High-powered money in the United Kingdom rose at an average annual rate of 4.2 percent from 1844 to 1846 and in the next three years fell at an average annual rate of 6.1 percent. In the United States, M2, after rising by 10.4 percent per year from 1842 to 1844, increased by only 4.1 percent per year over the next two, accelerated the following year, and then declined by 2.6 percent between 1847 and 1848. The only surprise in the data is that the U.K. recession does not appear to have been worse, given the amplitude and duration of the monetary contraction. Movements in the gold stock of the two countries in general conform to those of the other monetary aggregates. The U.K. gold stock decreased slightly between 1844 and 1846, after rising by 14.0 percent per year the

469

The Transmission of Business Cycles, 1833-1932

prior two years, and then fell by 2.6 percent per year from 1846 to 1849, with the largest annual decrease, 9.8 percent, coming in 1847-48. The U.S. gold stock behaved in like fashion, increasing by 11.3 percent per year between 1842 and 1844, falling by 2.9 percent per year over the next two years, then increasing by 15.0 percent from 1846 to 1847, and finally between 1847 and 1848 dropping by 2.6 percent. In both countries, therefore, the decreases in gold in the earlier part of the period were at least partially offset while those at the end of the period led to actual decreases in broader monetary aggregates. As in the two earlier contractions, the sequence of events seems to have been a specie drain in the United Kingdom, in this instance, particularly due to a trade deficit brought about by the Irish potato famine, subsequent increases in Bank rate (in 1847) to check the drain, and as a result a sizable gold outflow from the United States. In the United Kingdom, an exacerbating factor, at least as far as the monetary situation was concerned, was the widespread financial panic that began in the summer of 1847 and continued through the fall (see Dornbusch and Frenkel, this volume). The cause, contemporary observers claimed, was the gold outflow and the Bank's failure to contract its note issue gradually when the outflow began. Sir Robert Peel phrased his criticism thus: "If the bank had possessed the resolution to meet the coming danger by a contraction of its issues, by raising the rate of discount, by refusing much of the accommodation which they granted between the years 1844 and 1846 . . . the necessity for extrinsic interference might have been prevented; it might not have been necessary for the Government to authorize a violation of the Act of 1844" (MacLeod 1896, p. 148). The United States also experienced a panic, though not nearly so severe as the one in the United Kingdom: "embarrassments were slight and brief," according to Juglar (1916). The reason, as Warburton (1962) has pointed out, quite likely was the U.S. Treasury's purchase of government securities under a resale agreement that offset the initial declines in the money multiplier.

1856 The business contractions in the late 1850s-1856-58 in the United States, 1857-58 in the United Kingdom-took on familiar dimensions: pressure on domestic gold stocks, a reaction by the Bank of England, panic, and then a monetary contraction in both countries. The only difference between this and past cycles was in the accidentals. The Bank's defensive actions, for example, in this episode came in two stages rather than the heretofore usual one. Similarly, the major focus of investment in the period preceding the panic was different from those of

470

Wallace E. HutTman and James R. Lothian

the 1830s and 1840s. Hence, so also were the areas-both geographic and economic-in which the most notable bankruptcies and failures occurred. The behavior of output requires only slight elaboration. A relatively severe contraction took place in both countries. In the United Kingdom it was brief, but, as reference-cycle dates suggest, in the United States it was somewhat more protracted. The pattern of movements in the various monetary totals was similar to that described for earlier cycles. In the United Kingdom, a net gold inflow, which had produced a 6.3 per year increase in the monetary gold stock from 1854 to 1855, ceased the year after, and the gold stock remained nearly constant on a yearly basis. Then in 1857 the drain began, and gold declined by 5.5 percent. High-powered money behaved in a virtually identical manner with annual rates of change of 7.2 percent, 1.3 percent, and -6.3 percent in the three years, respectively. In the United States gold never decreased absolutely, but between 1856 and 1857 it rose by only 2.2 percent after having increased at an average annual rate of 12.6 percent in the preceding three years. The data for M2 show movements similar to those of gold: a 7.7 percent average annual rate of growth from 1853 to 1856, a 0.9 percent drop the next year, and then a slight 1.0 percent rise the year after. These yearly data, therefore, suggest that the slowdown in gold inflows in 1856 was the initiating factor in the cy'clical declines. As its gold reserves decreased, the Bank.of England raised its discount rate by 250 basis points in the space of a week in October of that year. That, in turn, intensified the pressure on the United States where banks in New York and on the rest of the East Coast were already trying to cope with an internal drain. They reacted by building up their reserves (Temin 1975), thus adding to the contraction in money. Insolvencies and suspension of payments followed in the late summer and early fall. The panic and run on the banks then spread to the United Kingdom. In November, even after having raised its discount rate from 5.5 percent to 10 percent in the short span of five weeks, the Bank of England asked for a suspension of the Banking Act of 1844. Suspension allowed it to expand its note issue, and by December the panic was over. The number of failures, however, rose considerably. A recessi9n that initially had a mild impact in both countries, intensified and spread, mainly in the United States. Given the linkages between the two countries, it is doubtful that the end result could have been much different in any event. Had the Bank of England not reacted to the pressure on its reserves in 1856, a contraction in money, and presumably the recession, would have taken place sooner than that year in the United Kingdom. The Bank's actions merely staved off both for a while. That, however, added to the problem in the United

471

The Transmission of Business Cycles, 1833-1932

States. As the U.S. recession became more and more severe, the feedback to the United Kingdom became greater and greater. A panic in the United Kingdom resulted and recession began there in earnest as well. 10.3.2 The Greenback Period During the seventeen years the United States was off the gold standard, the close economic linkages with the United Kingdom that existed prior to the Civil War broke down. Cyclical fluctuations took place at one time in one country and not in the other. And even in the instances in which there was a temporal coincidence, the channels through which these fluctuations might have spread were less than obvious. As illustrations of the two types of episodes respectively, we discuss the U.K. contraction of 1866 and the coincident contractions of 1873.

1866 The contraction of 1866 and associated panic in the United Kingdom produced no reaction in the United States. The contrast between this episode and the four just described thus provides one bit of evidence on the role the gold standard played in the transmission of fluctuations among countries. This evidence, however, is not totally unambiguous. The contraction in the United Kingdom was not severe. One could argue-though 1836 seems to run counter to this hypothesis-that the nonmonetary linkages between the two countries were more important than the monetary and that their operation, in turn, hinged on the severity of the initial contraction~ In terms of yearly GNP, the contraction of 1866 to 1868 manifested itself as a decline in the rate of growth, not an absolute decrease. Commensurate declines occurred in the rates of growth of gold and highpowered money and in the level of joint-stock-bank liabilities. The decline in gold, however, came in 1867-68, the second year of the recession. The decrease in the ratio of high-powered money to gold and, judging from Collins's (1981) series for liabilities of joint-stock banks, probably the ratio of M2 to gold as well, was due to the banking difficulties that began in early 1866. The cause of the decrease, both Clapham (1945) and MacLeod (1896) claim, was a drain on the Bank's specie reserves that began in late 1865 and induced the by-then-usual sharp increase in Bank rate. In February 1866 the first failure occurred, that of the Joint Stock Discount Company. In March Barned's Bank in Liverpool stopped payment. The highlight of that decade's panic was, however, the failure of Overend, Gurney and Company on 10 May with liabilities of over £10 million. The next day, the Banking Act of 1844 was suspended and the panic subsided. .

472

Wallace E. HutTman and James R. Lothian

1873 The contraction of 1873 in the United States by Burns and Mitchell's reckoning was the longest on record, not ending until 1878. The period of actual decline or sluggish growth in real income, however, was much shorter-1873 through 1875. From then on, real NNP rose rapidly, though prices continued to fall. The panic that took place in September of 1873 in the United States seems to have been largely domestic in origin. Friedman and Schwartz (1963b) cite the financial difficulties' of certain U.S. railroads and the resulting default on their debt as the precipitating factor. What seems to have set the stage for the panic was the substantial reduction in U.S. greenbacks and hence in bank-reserve ratios that occurred in the first half of 1873. The United Kingdom escaped the worst part of the U.S. panic. Equity prices were affected which led to a crisis on the London Stock Exchange, but there were no further monetary repercussions. The Bank of England, as it had throughout 1873, altered its discount rate promptly, increasing it to a high of 9 percent on 7 November 1873, and then in the space of four weeks lowering it back to 5 percent. Peel's Act, contrary to the fears of the time, was not suspended and a full-fledged panic was averted. "After 1873," Clapham (1945) states, "neither 9, nor 8, nor even 7 percent was announced again for a whole generation. An occasional 5 and a very occasional 6 was all that proved necessary." The rate of growth of M2 slowed appreciably in the United Kingdom in 1873, to 5.6 percent versus 9.3 percent the year before, while the rate of growth of the monetary gold stock declined by less than a percentage point during the same period. Real GNP grew at an average annual rate of 2.7 percent in 1873-74, about equal to that of 1872. Not until 1875 did real growth slow to any great extent; but from then until the referencecycle trough, its average rate of increase was only 0.4 percent per year. Movements in M2 in the United Kingdom from 1874 ran roughly parallel to those in real GNP: a further fall in the rate of increase of M2 between 1874 and 1875, near constancy in 1875-76, and then absolute declines in the stock during the last three years of the contraction. The cause of the restrictive movements in U.K. money was to a large extent, particularly in the years 1873-75, a series of declines in the rate of growth of high-powered money. These declines in turn were only partially the result of gold flows. In the later part of the period, a decline in the ratio of M2 to high-powered money became important? That in turn appears to have been the result of the failure of the City of Glasgow Bank in early October 1878 and the substantial increase in Bank rate in the middle of that month. The cyclical contraction of the 1870s, therefore, had two elements in common with the U.K. contraction of 1866: the United States was off the

473

The Transmission of Business Cycles, 1833-1932

gold standard and the channels of transmission of the type that were important prior to the Civil War appear not to have operated. Unlike 1866, however, these contractions were severe--one small bit of evidence in favor of our interpretation of 1866. If the sole reason that the earlier episode was confined to the United Kingdom was its lack of severity, there ought to have been some discernible linkages between the cycles in the two countries in 1873. The fact that there were none,or almost none, suggests that the monetary system rather than moderation of the episode was the key reason there was no transmission to the United States in 1866. 10.3.3

The Heyday of the Gold Standard

The United States returned to gold in 1879. During the next three-anda-half decades the United Kingdom and the United States underwent three common business cycles of more than average severity. None of the three, however, was an exact replica of the antebellum episodes. In the first, which began in 1882 in the United States and a year later in the United Kingdom, developments in the United States affected the United Kingdom at the start of the cycle; not until later did feedback occur. In the second, direct links between the two countries seem to have been minimal. Only in the third, the short-lived but nonetheless substantial contraction of 1907, was a strong influence running from the United Kingdom to the United States apparent at the onset of the cycle. 1882 The contractions of the early 1880s were moderately severe in both countries. In the U.K. contraction, dated 1883-86, real income grew at an average annual rate of less than 1 percent; in the U.S. contraction, dated 1882-85, real income was virtually constant for three years as a whole. The decline in the rate of growth of the U.S. money stock was particularly dramatic. The rate fell from an average of 19.3 percent per year in 1879-81, to an average of 6.9 percent in 1881-83, to virtually zero in 1883-84-reflecting a similar series of declines in the rate of growth of the monetary gold stock. In contrast, only a mild decrease in rates of growth occurred in the U.K. money stock during the contraction-in average terms, they were about a percentage point ·lower in 1883-86 than in 1881-83. Highpowered money, however, declined in absolute terms in each year of the contraction; and the gold stock declined in two of those three years. The drain of gold from the United Kingdom was the culmination of a movement that had begun in 1879 and that by 1882 had resulted in a cumulative decrease of close to 10 percent. The direction of movement was from the United Kingdom and other European countries to the

474

Wallace E. HutTman and James R. Lothian

United States; its cause was poor harvests in most of the world and exceptionally good ones in the United States. The response of the Bank of England to these drains was to raise its discount rate from 2.5 percent in April 1881 to 6 percent in January 1882. The end result was a cessation of inflows to the United States and a diminished rate of outflow, followed by an actual inflow of gold to the United Kingdom. According to Friedman and Schwartz (1963b), the reversal of the gold flow was one of the factors, along with foreigners' decreased confidence in investment in the United States and in the country's ability to remain on the gold standard, that precipitated a short-lived financial panic in New York in May 1884. The antebellum problems, therefore, reemerged in the postbellum period. The major differences were the milder fluctuations in output in the 1880s episode than in earlier ones and the reversed direction of causation at the start of the contraction-from the United States to the United Kingdom rather than the other way around. 1890

The U.K. cyclical decline began in 1890 and ended in 1894, making it one of the longest in that country's history. During the same period, the United States experienced two contractions: an exceedingly mild decline between 1890 and 1891 followed by a sharp rise in real growth the next year, and then a much more severe decline between 1892 and 1894. The U.S. contraction of 1890-91 manifests itself in the yearly data as a one-percentage-point decline in the growth of real NNP and a four-percentage-point decline in the growth of industrial production from their respective averages during the preceding two years. The money stock never fell but its growth rate declined. The cause was a gold outflow brought about by a shift of British investment to Argentina in mid-1890 at the same time that New York banks were experiencing the usual seasonal drain of specie reserves to agricultural areas of the country. As a result, a number of bank failures in the United States occurred during early November, and then on 15 November Baring Brothers, a major British merchant bank, suspended payment and the panic intensified. A month later, the panic in the United States was over. In the United Kingdom, it threatened to become severe but never did. The Bank of England immediately prior to the demise of Barings, as it became cognizant of what was likely to happen, raised the discount rate from 5 percent to 6 percent. Early the next week, it borrowed £0.3 million in gold from the Banque de France and bought another £1.5 million from Russia, thus further bolstering its reserves. In the initial year of the U.K. cycle, the growth of both real GNP and industrial production slackened. During the next two, industrial produc-

475

The Transmission of Business Cycles, 1833-1932

tion fell by 6.7 percent and real GNP remained virtually flat. By 1895, the rebound was underway. Growth in the U.K. money supply for the cycle as a whole declined, but the major part of that came after 1892. Hence, even though the reference-cycle dates differ between the two countries, the time pattern of output movements did not. In the United States, gold movements figured prominently in the explanation of movements in the money stock as a whole. The monetary gold stock, after falling by 8.7 percent in 1890-91, increased slightly the next year and then contracted sharply in 1893. The external drains reflected distrust of the Treasury's ability to maintain silver at parity with gold, as well as price deflation abroad. At the same time, an internal drain took place caused by distrust of the solvency of banks. This distrust, in turn, had its roots in the deflation that declines in capital and gold inflows had brought about earlier. In the United Kingdom, a reduced gold inflow was associated with the initial declines in monetary growth between 1891-92 and very likely 1892-93. Thereafter, the nongold component of high-powered money arithmetically accounted for the low rate of monetary growth. The U.S. contraction, therefore, quite clearly had international roots, but not as in many earlier cycles ones that extended directly back to Threadneedle Street. In the United Kingdom, the links with other countries were less obvious.s 1907

The contractions of 1907-8 had many of the earmarks of earlier episodes. From the spring of 1906 on, by Sayers's (1976) account, it became more and more evident that financial difficulties were liable to break out in the United States. In May and again in September of that year the Bank of England took defensive actions, in both instances increasing Bank rate from 3.5 percent to 4.0 percent, a decrease having been effected in June. At the same time, it imposed quantitative restrictions, refusing to discount paper used to finance American speculation. On 5 October it increased Bank rate further to 5 percent, and then on 19 October to 6 percent, the highest level since the Baring Crisis in 1890. "These measures," Friedman and Schwartz (1963b, p. 156) state, "served first to reduce, then to reverse, the flow of gold to the United States, and in this and other ways contributed to a change in the economic situation in the United States." The changes in gold flows, however, only show up to a minor extent in the annual data. The monetary gold stock in the United States, after increasing by 4.4 percent per year on average in 1904 and 1905, rose dramatically in 1906, a 9.0 percent increase relative to the preceding year. In 1907 the increase was only slightly less-8.8 percent. The monthly high-powered-money series, which is apt to be more

476

Wallace E. Huffman and James R. Lothian

dominated by gold than the money stock itself, registered a 2.8 percent annualized decrease from May to September 1907 versus an 8.4 percent annualized decrease in the stock of money. From then until February 1908, the money stock continued to fall (at a 12.3 percent annual rate) while high-powered money rose continuously. Not until July had money regained its May 1907 level. The major factor accounting arithmetically for the decrease in money, therefore, was the panic that broke out on 21 October 1907. The panic in turn was at least to some extent the result of the previous gold outflow and its impact on the reserve position of banks. The contraction in output in the United States, though lasting only a year, was sizable. Real NNP fell by 11.4 percent from 1907 to 1908 and industrial production by 17.0 percent. In the United Kingdom, the movements in both money and income were considerably more moderate. Real GNP decreased by 1.0 percent, industrial production by 8.4 percent, and the rate of monetary growth by the same amount as that of GNP. The cause of the monetary deceleration was a decrease in the rate of growth of the monetary gold stock, from 3.4 percent in 1906-7 to 1.7 percent in 1907-8. As in the case of the United States, though, these movements may well have been somewhat more severe when viewed intrayearly. Bank of England gold holdings, one of the few such series available, in March 1907 stood at £36 million. After rising by £2 million between then and September, holdings dropped to a low of less than £30 million on 4 November. The Bank's response to this outflow, as in the past, was to increase its discount rate. It did so by successive fifty- and then a hundred-basis-point amounts from 4.5 percent in September 1907 to 7 percent on 4 November. These increases, though probably necessary from the U.K. standpoint, worsened the problem in the United States. 10.3.4 The Interwar Period Taken together, the severe interwar contractions beginning in 1920 and 1929 provide almost a controlled experiment, the outcome of which demonstrates the important roles played during business contractions by monetary fluctuations within countries and by the gold standard in disseminating these fluctuations among countries.9 In both periods, the United Kingdom and the United States experienced sharp decreases in monetary growth beginning before the onset of recession. In 1921, the U.S. money supply rose while the U.K. money stock declined further. The rebound in the U.S. economy was both rapid and strong; the rebound in the U.K. economy was weaker and came later. In 1931 the United Kingdom broke with. the gold standard, thereby severing the monetary link with the United States. As a consequence, the United Kingdom was able to increase its money supply over the next two years, even as the U.S. money supply continued to decline. The depression in

477

The Transmission of Business Cycles, 1833-1932

the United Kingdom was thus cut short while that of the United States intensified. 1920

The 1920-21 contractions in the United Kingdom and the United States were two of the most severe one-year contractions of record. Both were accompanied by equally severe monetary contractions. In the United Kingdom, the annual data show a change in monetary growth of fourteen percentage points: from 12.1 percent per year in 1919-20 to - 2.3 percent in 1920-21. In the United States, the monetary deceleration was equally dramatic: from 11.5 percent in the one year to -0.58 percent in the next. The U.K. monetary contraction, like that of 1873 in the United States, was prompted by the desire to return to gold at the pre-World War I exchange rate. Given the inflation that had taken place in the interim-an inflation appreciably greater than in the United States-a substantial decrease in the U.K. money stock was necessary. Monthly data compiled by Lothian (1976) show monetary deceleration beginning in June 1919, nine months before the cycle peak. The peaks in the annual (1920) and monthly (October 1920) money series were followed by absolute declines that continued through 1925. In the United States, the money supply began to grow again in 1922. The real sides of the two economies reacted accordingly. In the United States, both real income and industrial production picked up rapidly, thereby cancelling out their initial declines a year sooner than in the United Kingdom. There the process dragged on, and not until 1924 did both U.K. series return to levels consistent with a modest 2 percent per year rate of growth. A year later, when the actual return to gold took place, a new recession began. The problems of the 1920s in the United Kingdom, therefore, appear to have been largely monetary in nature. Underlying the monetary fluctuations in turn were international considerations, in particular the return to gold at a price consistent with a $4.86/£ exchange rate. Keynes's assessment in the Treatise ofMoney (1930, 2: p. 181) seems to have been essentially correct: Looking back, we can see that the extreme prolongation of the slump was due to the Profit Deflation which occurred in the first half of 1921. This was doubtless inspired by the object of cancelling some part of the Income Inflation of the war and post-war periods ... but from the standpoint of national prosperity it was a mistake. We might have avoided most of the troubles of the last ten years . . . if we had endeavoured to stabilise our monetary position on the basis of the degree of Income Inflation existing at the end of 1920.

478

Wallace E. HutTman and James R. Lothian

1929 Data for the U.K. money stock at the start of the contraction show a mild deceleration. Yearly figures indicate a rise of 0.6 percent in 1928-29 versus an average rise during the two preceding years of 1.8 percent. The monthly data show a somewhat sharper falloff, from 3.2 percent growth over the twenty-four months ending January 1929-five months prior to the cycle peak-to a 3.6 percent decline in the money stock between then and January 1930. From 1929 until the end of the cycle, the yearly data show a sluggish 0.6 percent annual rate of increase-an average of 0.8 percent increase in 1929-30,1.2 percent decrease the next year, and 2.3 percent increase the year after that. Movements in gold were largely responsible for the monetary stringency. With the exception of 1929-30 when it rose by 5.4 percent, the monetary gold stock declined in three of the four years from 1928 to 1932. It fell by 6.1 percent in 1928-29, by 10.2 percent in 193{}-31, and then finally by 7.1 percent in 1931-32. Real GNP in the United Kingdom over the whole period fell by 5.7 percent and industrial production by 11.4 percent. In the United States, the money supply declined by a much greater amount during the period of the U.K. contraction-8.7 percent per year from 1929 to 1932. Moreover, it continued to decline at a 2.0 percent average annual rate from 1932 to 1934. Both real income and industrial production fell precipitously as a result: real NNP by a total of 34.5 percent from 1929 to 1932 and industrial production by a total of 62.7 percent. The U.S. declines continued into 1933. And, contrary to the experience of the United Kingdom, neither reached its 1929 level until almost the end of the decade. The 1929 contraction thus was marked by a reversal of the U.S. and U.K. roles in 1920. In the 1920 cycle, the United States became expansive earlier and thus escaped the problems that plagued the U.K. economy in the 1920s. In the second cycle, the U.K., abandoning gold in 1931, was able to avoid the further monetary contraction that took place in the United States. As a result, the U.K. economy rebounded more quickly in the 1930s than the U.S. economy did. lO During both interwar cycles, gold was in one way or another a key. The commitment to the return to the gold standard .provided the impetus for British deflation in the first instance; the abandonment of gold was the sine qua non for avoidance of further deflation in the second. 10.3.5

Conclusions from the Historical Analysis

Our analysis of individual reference-cycle contractions, to our minds, strdngly suggests that money was an, and most likely the, important causative factor in the major cyclical contractions in both countries. In almost all of the episodes a clearcut association is evident between

479

The Transmission of Business Cycles, 1833-1932

monetary decelerations and movements in output. That association, moreover, does not appear to be simply a reflection of reverse causation. For one thing the monetary shocks, as we have measured them, in almost all instances preceded or were coincident with the cyclical contractions. In relatively few instances did the monetary deceleration come after the fact. Nor do we find it plausible to believe that the association between money and output is largely the result of some common third factor that affects both variables. For one thing, the proximate causes of the monetary declines differed considerably across cycles, suggesting the absence of any simple mechanism to account for either feedback or the operation of such a third variable. Similarly, additional comparisons (described in appendix B) allow us to rule out one potential and often-suggested candidate-financial panics. A final bit of evidence is the difference in the incidence and duration of cyclical fluctuations between gold and non-gold standard periods. Direct monetary linkages were weaker in the latter; so also was the association between the cycles in the two countries. These results also provide evidence on how the transmission mechanism worked. Gold flows clearly were of direct importance in a considerable number of episodes. They also appear to have had an indirect effect in a number of others, acting as the proximate cause of financial crises that in turn led to substantial reductions in the ratios of commercial-banknote and deposit liabilities relative to gold. The analysis of the individual cycles, however, is rather moot with respect to other possible channels of influence-price and interest-rate arbitrage and direct-absorption-type effects on output. It also provides only limited information on the extent of feedback in the system. In addition it is almost solely concerned with severe cycles, which according to Cagan (1965) differ qualitatively from the less severe. At the same time it raises a number of questions about the stability of the relationships between the two countries over time. To try to resolve some of these issues, we now turn to the more formal statistical investigation. 10.4 Econometric Evidence

We estimate vector autoregressive models for the two countries combined and then use these models as the basis for conducting a series of tests of Granger causality.11 The advantage of these models is that they allow for simultaneous dynamic interaction among the variables while at the same time requiring relatively few identifying restrictions. We view these traits as particularly desirable in a study such as ours, which is concerned with short-run adjustment within and between economies of somewhat uncertain degrees of openness. The models require neither answers before the fact to the series of largely unsettled issues surround-

480

Wallace E. HutTman and James R. Lothian

ing cross-country channels of transmission and feedback mechanisms linking within-country variables, nor explicit modeling of a host of possible alternative dynamic relationships of both sorts. Given the attention they have received, however, the objections to these models and the associated Granger tests also require mention. Foremost among these objections are specification errors of the types outlined by Zellner (1979) and by Sims (1982). Having some notion of the possible temporal orderings of variables under different hypotheses and initially choosing those variables on the basis of theoretical considerations can reduce the likelihood of such errors and thus limit the effects on the statistical inferences being made.12 10.4.1

Model and Method

We consider a two-country macroeconomic model that emphasizes monetary variables. Each variable in the model is represented as a multivariate vector stochastic process (Sime 1980). In particular, the general, unrestricted autoregressive-reduced form is:

X t = 7T'(L) X t - 1 + Ut, t = 1, ... , T, (m x 1) (m x m) (m x 1) (m x 1)

(1)

where X t [Xt -

1]

= a vector containing current (one-period lagged) values of m

different economic variables (or their rates of change); 7T'(L) = an m x m matrix that contains polynomials in the lag operator that are one-sided on the past; Ut = a vector containing a random disturbance for each of the m equations; Ut is multivariate normal with E Ut = 0, and E UtU;

= I.

In this specification, all variables contained in X t are considered (potentially) endogenous, and in simultaneous-equation terminology, the set of predetermined variables that are regressors contains only lagged values of endogenous variables. The set of current exogenous variables is empty. This model is used to conduct the Granger tests. To illustrate these tests in a single-equation context, consider the first equation of the m-equation system (1): m

(2)

Xlt

= 7T'1 (L)

X t - 1 + U lt = .I 7T'1j(L) ~t-1 + U lt J=1

7T'11(L)X1t - 1 + 7T'12(L)X2t - 1 + ... + 7T'1m(L)Xmt - 1+ U lt , t = 1, ... , T, =

where 7T'ij(L) is the (n + 1)th order polynomial in the lag operator applicable to the jth variable in the ith equation. The null hypothesis that

481

The Transmission of Business Cycles, 1833-1932

X m does not Granger cause Xl is the restriction that all coefficients of the polynomial lag operator 1T 1m (L) [1T12( L) to 1T 1m(L)] are zero, i. e., that all lagged values of X m are excluded from the equation. We perform the test that X m does not cause Xl by comparing the error sum of squares of a model with q linear restrictions imposed on the coefficients of 1Tl(L) [ESS(w)] with the error sum of squares of an unrestricted model [ESS(n)]. We use the statistic (3)

F= [ESS(w) - ESS(O)]/q ESS(n)/[T - k] ,

which has an F-distribution with q and T - k [= T - men + 1)] degrees of freedom. 13 If we fail to reject the null hypothesis, then X m does not Granger-cause Xl. If we reject the null hypothesis, then X m is said to Granger-cause Xl, and we would like to think there is a behavioral structure underlying the reduced-form specification of the equation system (Sims 1980). We perform these Granger tests in two different contexts: (a) single equations (univariate models) independent of the other eleven equations in the macroeconomic model system, and (b) two or more equations jointly within the twelve-equation system (multivariate models). Our single-equation tests of the null hypothesis that X m does not Grangercause Xl are themselves of two types, weak and strong. The weak tests are essentially pairwise comparisons in which the "unrestricted" version of the test equation contains values of only two variables and is of the form: (4)

Xlt

1Til (L) X lt - l + 1Tim (L) X mt - l + Ut t = 1, ... , T =

where an asterisk indicates that the relevant terms are part of a twovariable rather than the more general m-variable system. On this we impose the restriction that the coefficients of 1Tim(L) are all zero. The disadvantage of this test is that one might falsely reject the null hypothesis because of omitted variable bias in the estimates of 1Ti(L) , resulting from exclusion of lagged values of X 2 through X m --...1. One or more of these may be truly Granger-causing Xl, but we could erroneously reach the opposite conclusion if the variable being analyzed were correlated with one or more of the other variables. Accordingly, we also employ a singleequation strong test that X m does not Granger-cause Xl by imposing the restrictions on equation (2) that the coefficients of 1Tlm(L) are all zero. The test tells us whether X m contributes significantly to explaining the variance in Xl, holding variables X 2 through X m constant. Tables 10.4 and 10.5 contain the results of these two sets of tests, respectively; table 10.7 contains an overall summary of these and of subsequent test results. The final tests that we perform are tests of multiple causes. These also

.30 .56 .03 2.84 .31 .45

3.13 b 1.22 1.60 6.67 .25 .04

YS

2.92 4.58 2.15 3.29 .29 .66

1.61 2.60 .42 5.26 2.13 .86

PS

IS DCS

NS YK

PK

2.69 1.69 .56 1.38 1.23 .86

1.64 2.92 1.54 2.44 .52 .03 4.42 .50 .92 3.42 .65 1.73

7.15 3.95 1.84 9.39 5.08 1.05 .43 .76 2.26 .43 .16 .91

1.35 3.29 1.29 4.34 .89 .42

.09

.06 1.75 .17 .18 2.44

.07 .79 2.84 1.43 .02 2.68 1.43 2.56 7.02 7.22 2.15 4.64

.82 .04 .07 3.41 .56 1.19

.22

.66

.56 1.26 .04 .44

.54 .43 .28 3.53 .34 1.73

(regressor with coefficients restricted to zero in null hypothesis)1

FRS

6.51 4.08 2.44 1.54 .42 1.68

6.17 2.10 .96 .96 3.25 1.22

FRK

1.34 .47 .31 2.62 .18 .40

1.86 .07 .40 1.33 1.00 .08

IK

.20 7.76 1.31 9.59 6.94 .08

.99 .57 2.15 4.75 2.29 .22

DCK

Weak Tests of Granger Causality in a Two-Country Macroeconomic Model: U.S.-U.K. Gold Standard Period, 1837-59 and 1882-1914a (F -statistics)

.97 .08 .80 .84 1.60 71.26

.32 .32 1.13 .78 .47 .36

NK

aEach "unrestricted" model contains four regressors: the one- and two-period lagged values of the dependent variable, the one- and two-period lagged values of one other variable (except when a variable is regressed on the lagged values of itself), a dummy variable (= 1 for pre-Civil War years), and an intercept. All variables are expressed as percentage rates of exchange except for interest rates that are differences of levels; the FR and DC terms were weighted by their respective shares in high-powered money. bFor off-diagonal tests, the critical F-values for 2 and 51 degrees of freedom at 10,5, and 1 percent significance levels are 2.41,3.18, and 5.06 respectively.

YK PK FRK IK DCK NK

United Kingdom

YS PS FRS IS DCS NS

United States

Equation

Table 10.4

.36 3.54 1.31 .12 .27 .94

PS

FRS

IS

DCS

NS

YK

PK FRK

IK

1.96 3.57 .64 1.98 .31 1.49

1.22 .46 1.44 7.94 1.58 1.68

2.64 1.91 .94 3.87 .03 5.68

2.20 5.27 1.10 6.56 .72 1.08 .04 .52 .53 2.69 1.49 2.61

1.40 2.78 .92 6.25 .99 .69 1.19 1.44 .55 2.18 .04 .02

.97 3.09 5.98 1.15 .51 .45 .46 .54 .30 .59 1.63 1.17

.05 .29 4.04 5.58 .25 4.40 .06 1.06 6.27 2.51 4.06 4.47

1.40 .44 2.16 .84 .99 .46 .14 3.27 .40 .02 1.24 .09

1.65 1.55 2.79 .90 .08 1.36 1.77 4.51 3.49 .06 .96 3.28

1.72 1.21 2.57 .46 1.00 1.91

.04 2.67 1.75 1.94 1.10 .64

.30 .76 .45 1.41 2.35 .33

(regressor with one- and two-year lagged values having coefficients restricted to zero in null hypotheses)a

2.28b 2.58 .71 .61 .21 1.12

YS

.60

1.43 6.37 .88 5.26 4.33

1.36 2.30 2.31 .62 4.47 .32

DCK

Strong Tests of Granger Causality in a Two-Country Macroeconomic Model: U.S.-U.K. Gold Standard Period, 1837-59 and 1882-1914a (F -statistics)

1.79 1.40 2.25 1.39 1.05 55.20

1.48 1.31 2.64 1.97 1.07 1.66

NK

aEach "unrestricted" model contains one- and two-year lagged values of the twelve different variables (twenty-four regressors), a dummy variable (= 1 for pre-Civil War years), and an intercept; the two high-powered-money variables are excluded. All variables are expressed as percentage rates of change, except for interest rates that are differences of level. The FR and DC variables were weighted by their respective shares of high-powered money. bCritical F-values at 10, 5, and 1 percent significance levels for 2 and 30 degrees of freedom are 2.49, 3.32, and 5.39 respectively.

YK PK FRK IK DCK NK

United Kingdom

YS PS FRS IS DCS NS

United States

Equation

Table 10.5

484

Wallace E. HutTman and James R. Lothian

are of two types. The first multiple-cause tests are on each of the single equations taken independently of the other eleven equations of the system. In these tests, the unrestricted regression is of the form of equation (2). The restriction is that all coefficients of all of the polynomial lag operators applying to either all foreign variables or all domestic variables other than the regressand are zero. The other multiple-cause tests are tests on two or more equations jointly. In conducting these tests, we take account of the contemporaneous correlations across all twelve equations of our macroeconomic system. These tests are the direct analogues in a multiequation context of the single-equation multiple-cause tests just described. Under the various null hypotheses we impose restrictions on entire blocks of the coefficient matrix 71"(L) rather than on portions of one particular row. The F-statistics for multiple-cause tests on single equations are reported in table 10.6, part A, and chi-squared statistics for joint tests across two or more equations in part B. For the latter test, we base our conclusions on Sims's (1980) version of the chi-squared statistic, which is reported in columns (la) and (2a) of table 10.6, but we also report the other frequently used chi-squared statistic in columns (lb) and (2b ).14 To make the model operational for the study of macroeconomic interrelationships between the United Kingdom and the United States during the gold standard period, we initially assigned the following twelve variables to the X matrix in equation (1): YS

= U. S. real NNP, or prior to the Civil War a proxy

PS

=

U.S. NNP deflator

FRS = U.S. specie reserves IS

= U.S. short-term interest rate

DCS

=

U.S. domestic-credit component of high-powered money

NS

=

U.S. population

= U.K. real GNP PK = U.K. GNP deflator YK

FRK

=

U.K. specie reserves

IK

=

U.K. short-term interest rate

DCK = U.K. domestic-credit component of high-pow- .

ered money NK

= U.K. population

485

The Transmission of Business Cycles, 1833-1932

We used annual data to estimate the model over the combined subperiods 1837-59 and 1882-1914.15 We omitted the Civil War and greenback periods since the United States was off the gold standard during those years. Additional observations at the start of each subperiod were lost in differencing and in the process of taking lags. With the exception of interest rates and monetary variables, we entered all variables in the model as percentage rates of change. For interest rates we used first differences of levels and for the monetary variables first differences of levels scaled by the level of high-powered money. The latter is equivalent to weighting the percentage rates of change of the DC and FR variables by their shares in high-powered money. In each instance the equations included an intercept term, a dummy variable for the second subperiod, and two lagged values of each of the independent variables.16 We estimated all equations using ordinary least squares. In our multivariate, multiple-cause tests we do, however, take account of contemporaneous cross-equation correlation of error terms. These crossequation correlations may capture sources of business cycle transmission omitted from the model (table 10.7). 10.4.2 Tests Based upon the Full Model Since our principal interest is in the real-income tests, we turn to these first and find the results are rather mixed. In the single-equation weakform tests we find some direct influence of monetary variables on real income in the two countries: FRK is a significant predictor of both YK and YS, and FRS (as well as PS) approach significance in the YK relationship. In addition IS, which in turn is influenced by FRS, DCS, and DCK, significantly affects both YK and YS. In the strong-form tests, however, most of these relationships break down: FRS is significant at the 10 percent level in predicting YK, at somewhat less than the 10 percent level in predicting YS. Nothing else apparently matters. The single-equation multicause tests reported in the top half of table 10.6 are even less informative. For both YK and YS we are unable to reject either the null hypothesis of no-domestic-cause or of no-other-country-cause. A number of possible reasons can be found for our failure to discover much in the way of a relationship here. One is that a strict version of the rational-expectations-natural-rate hypothesis holds (Sargent 1976; Leiderman 1980). Another is that some subset of the variables-say foreign reserves and domestic credit-is jointly significant but that the effects are being masked by the inclusion of a large number of truly insignificant variables. A third, related to the second, and to which we return below, is that we have misspecified the monetary variables. A further possible reason for little or no influence of other variables on real

1.24 3.73 3.36 2.39 .93 .85 1.46 3.80 2.62 2.52 1.60 2.41

No Domestic Cause 2 (1)

Null Hypothesis

1.51 1.52 1.18 1.73 1.68 .99 1.27 1.69 .99 1.78 .97 1.72

No Other-Country Cause 3 (2)

Tests of Granger Causality in a Two-Country Macroeconomic Model: U.S.-U.K. Gold Standard Period, 1837-59 and 1882-19141

A. Univariate Model 4 1. YS 2. PS 3. IS 4. FRS 5. DCS 6. NS 7. YK 8. PK 9. FRK 10.IK 11. DCK 12. NK

Model

Table 10.6

87.9 81.3 161.5 123.2 125.3

(50.9)6 (50.9) (84.4) (84.4) (84.4) 164.2 151.7 301.4 230.0 233.8

(50.9) (50.9) (84.4) (84.4) (84.4)

T· In I!w/!n I (lb)

57.7 (58.6) 68.2 (58.6) 120.0 (102.8) 110.2 (102.8) 116.3 (102.8)

(T-k) . Inl!w/!n (2a)

(58.6) (58.6) (102.8) (102.8) (102.8)

(2b)

Inl!uJ!nl

107.7 127.3 224.0 205.8 217.0

T·

Notes: 1. Each unrestricted equation contains one- and two-year lagged values of the twelve different variables (twenty-four regressors), a dummy variable (= 1 for pre-Civil War years), and an intercept; the high-powered-money variables are excluded. All variables are expressed as percentage rates of change, except for interest rates that are differences of levels. The FR and DC variables were weighted by their respective shares of high-powered money. 2. Each equation in the restricted model has ten coefficients set equal to zero, all lagged values of the domestic country's variables except lagged values of the dependent variable. 3. Each equation in the restricted model has twelve coefficients set equal to zero, all lagged values of the other country's variables. 4. The tests ignore potential contemporaneous cross-equation correlation-of-error terms. The critical F-values of 5 and 1 percent significance levels for 10 and 30 degrees of freedom are 2.16 and 2.98, respectively, and for 12 and 36 degrees of freedom are 2.07 and 2.80, respectively. 5. All tests are performed within a twelve-equation system where cross-equation contemporaneous correlation-of-error terms are taken into account. In column (1), rows 13 and 14, a total of30 zero restrictions are imposed; for rows 15-17, 60 zero restrictions are imposed. In column (2), rows 13 and 14, a total of 36 zero restrictions are imposed; for rows 15-17, 72 zero restrictions are imposed. For the distinction between la and Ib and 2a and 2b, see note 14. 6. Critical X 2 values at the 1 percent significance level are reported in parentheses.

B. Multivariate Model5 13. YS, PS, IS 14. YK, PK, IK 15. = 13 & 14 16. All 6 U.S. dependent var. 17. All 6 U.K. dependent var.

(T-k)·lnl!uJ!n (la)

x2 -statistic

weak strong multicause

·strong multicause

NS: weak

strong multicause

DCS: weak

IS:

strong multicause

FRS: weak

strong multicause

PS: weak

strong multicause

YS: weak

Dependent Variable

Table 10.7

N IN

N IN

Y IN

N ,N

N

I®

1-

YS

N N

N N

Y Y

N N

-

N N

PS

N N

N N

Y

®

-

Y

®

N N

FRS

I

N

I

N

I

Y

I

Y

I

Y

I

N

N N

Y N

N N

Y

®

Y N

IS

N N

Y N

N Y

Y Y

N N

DSC

I

1

-I

N N

N Y,

Y,

®

N N

N N,

NS

N IN

N IN

Y IN

IN

N

N IN

N IN

YK

Granger-causing Variables

N N

N N

Y N

N N

Y N

N N

N

®

N

N N

Y N

FRK

®

N N

N N

PK

,

I

N

N

I

I

N

I

N

I

N

N

N N

N N

N N

N

N

N N

N N

IK

Three Sets of Causality Tests in a Two-Country Macroeconomic Model: U.S.- U.K., 1837-59 and 1882-1914

N N

N Y

Y N

N

N

N N

N N

DCK

1

N N,

N N

N N,

®,

N

N N,

N N,

NK

®

N IN

N IN

IN

N N

N N

Y N

N N

Y Y

N

® N

N Y

N N

N Y

N N

®

N

1

N

I

I

N

N

1

I

N

N

® ® ,

®

N

N N

®

Y

N N

N N

Y N

N N

N N

N N

N N

N N

N N

1

1

1

N N 1

N

®

N N

~

N

N N

N N I

I

Y Y

IY

N

Y

I®

Y IY

N N

N N

N

N

N

N Y

N N

N

N

-

-

Y

Y

Y N

® N

N N

IN

1-

I

N

N

I

I

Y

Y

I

I

Y

I

N

N N

N N

-

-

N

N

N

®'

N N

N N

-

Y

Y

N

N

Y

Y

N N

reject the null hypothesis of no cause at the 10 percent significance level.

Notes: Y = yes = reject the null hypothesis of no cause at the 5 percent significance level; N = no = fail to reject the null hypothesis of no cause;

NK: weak strong multicause

strong multicause

DCK: weak

weak strong multicause

N IN

FRK: weak strong multicause

IK:

N IY

N IN

PK: weak strong multicause

strong multicause

YK: weak

1

®

=

-I

N N1

N,

N

N

N

N1

N

N N,

490

Wallace E. HutTman and James R. Lothian

income in the two countries is that there are common shocks-financial panics are an obvious example-that we have failed to take into account. In the multivariate multicause tests reported in the bottom of table 10.6, we allow for such shocks by taking account of contemporaneous cross-equation correlation of the errors. In these tests, when either of the real-income variables is examined in conjunction with the domestic price and interest variables or with all five other domestic variables, we almost always reject the null hypotheses of no-domestic and no-other-country causes. These last results, therefore, suggest that there was a set of mechanisms by which disturbances were transmitted internationally. The results say nothing, however, about either the relative importance of the different variables in the different equations or the specific channels of transmission. We can get some notion of both by examining some of the other single-equation test results. The price equations are particularly interesting in both regards. For the United Kingdom as well as the United States in both the weak and strong forms of the tests, own-country foreign reserves and domestic credit as well as own-country rate of interest are significant predictors of own-country price level. For the U.K. price level, Granger-causation results from the U.S. price level. A similar arbitrage relationship appears to exist between interest rates in the two countries.17 The U.S. rate Granger-causes the U.K. rate in both the weak- and strong-form tests. The results are consistent with the existence of a specie-flow channel linking the two countries and, to a lesser extent, direct price and interest-rate-arbitrage channels. They are, however, inconsistent with the simplest model of the monetary approach to the balance of payments. The model assumes that arbitrage is complete within the period, suggesting, therefore, that the domestic price level either Granger-causes or is contemporaneously correlated with money. Correspondingly, the model views domestic credit as affecting only the stock of foreign reserves and not the nominal money stock or the price level. More general models of the types estimated by Darby and Stockman (1983) appear to be required to describe the historical data. The foreign-reserve and domestic-credit tests for the two countries contain additional information bearing on these subjects. In the strong tests (but not the weak) we find Granger-causation of FRS by FRK. The reverse relationship, however, does not hold. At the same time, we find Granger-causation of FRS by DeS. There is, therefore, a further suggestion of a specie-flow channel as well as of a relation between domestic credit and foreign reserves of the kind posited in a broad class of monetary models. No consistency in these relationships between countries is, however, shown. Moreover the direction of influence uncovered for

491

The Transmission of Business Cycles, 1833-1932

foreign reserves, United Kingdom to United States, is the reverse of that suggested by the price-level and interest-rate tests. The remaining set of relationships that are of some interest are those for domestic credit. U.K. real income appears to Granger-cause domestic credit, being significant at the 5 percent level in the strong-form tests and at close to the 10 percent level in the weak-form ones. The U.S. weakform tests give evidence of Granger-causation of DCS by FRK and IS and perhaps also by DCK and PS. In the strong-form tests DCK and perhaps IK Granger-cause DCS. One possible interpretation of the U. K. results-is in terms of a reaction function of the Bank of England. In the case of the United States, which over this period had no central bank, what we may be capturing are the effect of U.K. monetary pressures on the fiduciary component of commercial-bank-note issues. 10.4.3 Further Tests of the Real-Income Relationship As mentioned, a potential source of bias in the real-income tests stems from the way we entered the monetary variables. For both countries we disaggregated high-powered money into domestic and foreign components. By using high-powered money alone, we ignore any contribution the money multiplier might have made. And to the extent that domestic credit and foreign reserves are perfect substitutes in their effects on real income, treating them separately may bias the case against finding Granger causation. Testing the two monetary variables jointly would solve the second problem but not the first. Accordingly, we reran the real-income tests using U.S. M2 and U.K. high-powered money (the only measure available) in place of the other monetary variables. We report the test results based upon this model in table 10.8. The results paint quite a different picture than the previous ones. Unlike the earlier results, these show a clearcut association betwee~ own-country money and real income. In all four instances-YK vs. HK and YS vs. MS, in both forms of the test-we find Granger causation from the monetary variable to income. The relationships, however, are not simple. Other-country money also has significant effects both on owncountry money and on own-country real income in all the comparisons. Similarly, for the United Kingdom there is evidence of reverse causation, YK having a significant effect on HK. We thus find what we failed to establish in the earlier set of results. At the same time, additional evidence emerges of a complex system of interaction between the two countries operating through monetary channels. The relations uncovered between other-country money and owncountry real income are particularly intriguing. One possible explanation is that we are capturing with other-country money the effect of monetary shocks abroad on the money multipliers and, hence, on real income in the

492

Wallace E. Huffman and James R. Lothian

Table 10.8

Additional Granger Tests: U.S.- U.K. Monetary Models, 1834-19141 Strong Tests

Weak Tests

Multiple Cause

Dependent Variable

Causing Variable F

Causing Variable F

Own Variables

YS

MS HK YS YK

5.01 6.80 3.49 0.02

1.97

1.94

1.03

2.24

HK MS YK YS

4.32 1.69 1.62 0.10

MS HK YS YK HK HK MS YK YS MS

1.91

1.71

2.51

1.46

MS YK HK Degrees of freedom for column Critical F at 5% level 10% level

3.63 3.59 0.35 1.47 4.27 3.30 4.02 4.52 1.23 2.92

Foreign Variables

2, 50

2, 34

8,34

10,34

3.19 2.42

3.29 2.47

2.23 1.86

2.12 1.80

1. All equations included both an intercept and (1, 0) dummy variable for pre- and post-Civil War years.

two countries. For the United Kingdom we have been forced to use high-powered money alone, so this explanation is particularly plausible. For the United States the deposit data for the antebellum period are likely to be subject to substantial error. Movements in U.K. highpowered money, therefore, may be a proxy for movements in the true U.S. money stock that are not reflected in movements in the measured money stock. The alternative explanation is that the result is a reflection of some underlying behavioral relationship. One possibility is that the two monies were close substitutes from the standpoint of domestic money holders in the two countries. In that case, the true money stock in each country would be some weighted average of measured U.K. money and of measured U.S. money, with the weights most likely varying from the one country to the other. A further possibility is that we are capturing some aspect of the adjustment mechanism linking the two countries, rather than some aspect of a steady-state equilibium relationship such as currency substitution. Asset-market adjustment across a wider spectrum than the short-term financial assets whose yields we include in the model is a potential candidate .18 10.4.4 Conclusions from the Tests Some of the explanation of results has been conjectural. In addition, certain relationships appear implausible a priori; certain others appear

493

The Transmission of Business Cycles, 1833-1932

inconsistent with findings that can be rationalized. Viewed as a whole, however, the results do tell a story of simultaneous dynamic interaction between the United Kingdom and the United States that in broad outline jibes with the inferences already drawn from the analysis of the data for individual reference cycles. The monetary system appears to be of considerable importance in the transmission mechanism. Monetary decelerations appear to be a significant determinant of cyclical contractions in real income. The evidence for monetary causes of transmission of business cycles is strongest when money itself is included directly in the econometric analysis. The analysis that allowed for cross-equation correlation of the residuals indicates that there probably were additional sources of transmission omitted from the model. 10.5 General Summary and Conclusions

In investigating the causes and transmission of cyclical fluctuations under the gold standard, we pursued two different research strategies. We began with an examination of each important cyclical episode on an individual basis and focused that analysis on the cyclical behavior of the monetary data, the cross-country interrelationships between movements in the specie and money stocks of the two countries, and the within-country interrelationships of those series and output. We then proceeded to estimate vector-autoregressive macroeconomic models for the United Kingdom and United States combined using variables that alternative hypotheses about cyclical fluctuations and transmission during this period suggest are important. We used the models in conducting a series of Granger tests, appropriate to both sets of hypotheses. Because the historical and the econometric exercises are largely complementary, we have greater confidence in those findings that are common to the two approaches. Two items in particular deserve comment. One item is the role of money in cyclical contractions. Taken together the two types of analysis indicate that monetary shocks were important independent factors leading to or worsening the severity and duration of the contractions in the two countries. During severe contractions, moreover, the shocks appear to have been the most important causative factor. 19 The other item is the part played by the gold standard in the process. Both the historical and the econometric analyses point to it as a key element in the transmission mechanism. Reestablishment of the equilibrium conditions of the system after a monetary shock in one country, typically produced both gold flows and also price and capital-market adjustments. These in turn induced cyclical fluctuations in output in the other. Gold outflows, moreover, were particularly important in transmission, having two avenues of influence not only affecting high-powered

494

Wallace E. Huffman and James R. Lothian

money but also, in a considerable number of episodes, leading to financial crises and subsequent declines in the money multiplier. The two types of analysis, separately and combined, also suggest a number of other conclusions of less importance, which we merely list: 1. During the course of the sample period the United States and the United Kingdom appear to have reversed their roles: the United Kingdom seems to have been the senior partner prior to 1860, the United States in the first three decades of the twentieth century; neither was clearly predominant during the intervening years. 2. Within those subperiods, however, causation was not geographically unidirectionaL Shocks initiated in one country, that spilled over to the other appear to have reverberated back to the originating country to greater or lesser degree depending upon the particular episode. 3. Within countries there is evidence of a similarly complex transmission mechanism. Income had feedback effects on money of at least occasional and probably of general importance in both the United Kingdom and the United States. 4. During the relatively short periods when either the United Kingdom or United States was off gold, transmission of cyclical fluctuations is clearly less evident. Flexible exchange rates appear, therefore, to offer some and perhaps a considerable degree of insulation against cyclical contractions. 5. Short-term independence of monetary policy was possible even under the gold standard. The Bank of England often undertook defensive actions that halted and then reversed specie outflows. Those actions, in turn, appear to have had subsequent effects on income in both countries, moderating the decline in the United Kingdom and aggravating the decline in the United States. From these findings, we draw several conclusions relevant to monetary policy. Given the attention the gold standard has received in the United States in recent years, these findings deserve explicit mention. The benefits of a gold standard, as usually enumerated, are that it is both automatic and impersonal and that it e~fectively constrains governments from using money creation as a taxation device. Our analysis suggests that the automaticity and impersonality were less than complete. The Bank of England's intervention alluded to above was a prime example. More important, because cyclical fluctuations were transmitted internationally with apparent ease under the gold standard, one has to weigh'the costs of a greater incidence of such fluctuations against the benefits of a greater degree of secular price stability.

495

The Transmission of Business Cycles, 1833-1932

Appendix A The Data United States: Individual Series and Sources High-Powered Money. High-powered money is defined as the sum of notes and specie held by the banks and the nonbank public. Data for 1833-59 are from Rutner 1974, table 28, col. 15 plus col. 19; for 18791933 from-Friedman and Schwartz 1982, table 4.8, col. 9. Since Rutner's data were reported for varying monthly bank-statement dates, we took appropriate weighted averages of the original data to arrive at figures approximately centered on the end of June. Money. Money is defined as the sum of currency (notes and specie) and commercial-bank demand and time deposits held by the nonbank public. Data for 1833-59 are from Rutner 1974, table 57, col. entitled "Calendar Year"; for 1870-1933 from Friedman and Schwartz 1982, table 4.8, col. 1. Specie. Specie is defined as that held by banks and the nonbank public plus specie held by the Treasury and, from 1914 on, the Federal Reserve. Data for 1833-59 are from Rutner 1974, table 28, col. 1, adjusted by us to a yearly (June-centered) average; for 1879-1914 from Friedman and Schwartz 1963a, tables 5 and 8; for the remaining years from worksheets underlying Friedman and Schwartz 1963a. Real Income. Data for 1833-59 are from the Smith and Cole index described below; for 1870-1933, real net national product from Friedman and Schwartz 1982, table 4.8, col. 3. Price Index. Data for 1833-59 are from a yearly GNP deflator derived from Gallman's benchmark estimates; for 1870-1933 from an NNP deflator from Friedman and Schwartz 1982, table 4.8, col. 4. Interest Rate. Data are for commercial paper rates; for 1833-59 from annual averages of Bigelow's monthly series in Macaulay 1938, appendix table 25; for 1870-1933 from Friedman and Schwartz 1982, table 4.8, col. 6. Population. Data for 1833-59 are linear interpolations of census-year decennial estimates from Rutner 1974, table 37; for 1870-1933 from Friedman and Schwartz 1982, table 4.8, col. 5.

496

Wallace E. Huffman and James R. Lothian

United Kingdom: Individual Series and Sources High-Powered Money. High-powered money is defined as the sum of notes and coin held by the public plus bankers deposits and other private deposits at the Bank of England, 1833-70 from Huffman and Lothian 1980; for 1871-1933 from Friedman and Schwartz 1982, table 4.9, col. 9. Money.' Money is defined as the sum of currency held by the public and total deposit (current accounts and deposit accounts) at commercial banks. Data for 1871-1933 are from Friedman and Schwartz 1982, table 4.9, col. 1. Specie. Specie is defined as the sum of specie held by the public and that held by the Bank of England. Data for specie held by the public for 1833-70 come from Huffman and Lothian 1980; thereafter from unpublished worksheets underlying the data reported in Friedman and Schwartz 1982, table 4.9. Specie held by the Bank for 1833-1879 comes from an unpublished appendix, "Bank of England Liabilities and Assets: 1696 to 1966," col. entitled "Assets. Coin and Bullion," to the article with that title in the Bank of England Quarterly Bulletin, June 1967, pp. 159-63; thereafter from Sheppard 1971, table A 1.12, col. 15. Real Income. Real GNP is from Deane 1968 for 1833-1912; thereafter, the'series is derived by us from Feinstein's (1972, table 7, col. 7) index of real GNP at constant factor cost. Price Index. Data are from the real-GNP series described above divided by nominal GNP from Deane 1968 for 1833-1912; thereafter, from Feinstein 1972, table 2, col. 10. Interest Rates. Data are for first-class three-month bills; for 1833-56 from Mitchell and Deane 1962; for 1857-67, ibid.; for 1868-1933 from Friedman and Schwartz 1982, table 4.9, col. 6. Population. Data for 1833-67 are from Mitchell and Deane 1968, p. 8; for 1868-1933 from Friedman and Schwartz 1982, table 4.9, col. 5. Problems with the U.S. Output Data As a measure of real output in the U.S. during the antebellum period, we used an index derived from Smith and Cole's (1935) separate production indexes for the years 1831-45 and 1843-62. Both indexes are made up of two components-domestic trade (twothirds weight) and foreign trade (one-third weight). The domestic index for 1831-45 was derived from eleven component series, eight of which were expressed in physical units; the domestic index for 1843-62 from ten

497

The Transmission of Business Cycles, 1833-1932

component series, six of which were in physical units. The foreign trade indexes were both nominal measures based on the total of exports and imports in current prices in both periods. None of our conclusions about cyclical movements, however, would have been grossly different had we relied solely on the domestic indexes. We linked the two production indexes on the basis of Ayres's (1939) index of cyclical fluctuations. The Smith and Cole indexes as published are in the form of deviations from trend. Logarithmic first differences of these indexes, therefore, overstate the decline in the non-trend-adjusted series. The overstatement in the case of a series that follows a constant semilogarithmic trend is the intercept term in that trend equation. In spite of their deficiencies, these indexes appeared far preferable to the alternative measure of output we examined, an annual real-GNP series derived from Robert Gallman's (1966) benchmark estimates. Inspection of this series revealed almost no correspondence with the NBER reference cycles-even during the 1839-43 contraction which, by all accounts both contemporary and subsequent, was unusually severe. Most of the physical-volume series for individual industries we examined, in contrast, did exhibit cyclical movements corresponding to the NBER pattern as also did the Smith and Cole indexes. One reason for the lack of cyclical movement in the Gallman series may be its omission of inventories, usually one of the most cyclically sensitive components of GNP. Proxy Series for the U.K. Money Supply Prior to 1870 the U.K. deposit data are incomplete. For a subset of these years, though, we have a proxy series-total liabilities of private and joint-stock bank in England and Wales to the nonbank public-that Michael Collins (1981) has constructed. Movements in these data are summarized in a note at the bottom of table 10.3. We view these data as indicators of the direction but not the magnitude of movement of the overall money stock relative to that of the monetary gold stock. We regard these data as suspect from the latter standpoint because Collins was unable to obtain complete bank-balance-sheet data for the whole period. As an interpolator, he used the number of bank offices. In periods of banking panic when there were substantial bank failures, his series may therefore be more volatile than the true series.

Appendix B Panics and Cyclical Contractions Discussions of financial panics abound in the literature devoted to particular periods in the economic history of each country. More general

498

Wallace E. HutTman and James R. Lothian

treatments of financial panics, either from a primarily theoretical and primarily historical standpoint, however, are few. One group of modern studies that has dealt with the phenomenon of financial panics in some depth are those of the U.S. monetary system at the National Bureau: Milton Friedman and Anna Schwartz's A Monetary History of the United States (1963a) and their related article "Money and Business Cycles" (1963b), and especially Phillip Cagan's Determinants and Effects of Changes in the Stock of Money, 1875-1960 (1965). Charles Kindleberger's Manias, Panics, and Crashes (1978) is a more recent work devoted to the study of such episodes in an explicitly international context. One of Cagan's specific concerns was the interrelation of cycles in monetary growth and business. In a subsection of that title in the summary chapter of his study he concludes: This evidence points to an important independent role of monetary factors in severe business contractions. The six largest declines in money were associated with severe depressions, and severe depressions have never occurred otherwise.... Panics cannot be held solely responsible for the deep declines in both money and business. Two severe contractions had no panic; in addition, some panics did not produce a large drop in monetary growth, and the accompanying declines in business did not become severe. (P. 296) Kindleberger, though he does not refer to Cagan's study, apparently would disagree with his assessment. In Kindleberger's framework, panics are the natural culmination of the previous boom in which speculation and overtrading are rife. An increase in the money supply may alleviate the effects of the panic, but a decrease during the panic is not a necessary condition for a severe cyclical downturn. Cagan's conclusions about the effects of panics stem in large part from the comparisons he makes between cycles that were severe and had no panics and cycles that were not severe but had panics. Of four relevant episodes-two in each category-he excludes two from consideration, ending up with one in each category. Our sample extends farther back in time than Cagan's and covers the United Kingdom as well as the United States. Hence, it offers additional degrees of freedom with which we can assess the relativ~ importance of panics and monetary contractions as proximate causes of business contractions. To that end table 10.A.1 classifies cyclical contractions in both countries according to both degree of severity (severe versus mild) and existence of a banking panic. We exclude cycles that occupy the cell mild, no panic. We further classify each of the cycles that we include according to the degree of monetary contraction. As a starting point in dividing the cycles according to degree of severity, we adopted Burns and Mitchell's classification of 1857, 1873, 1893,

1887 1913

1836 1839 1847 1856 1873* 1882* 1892 1907 1920 1929

Date of Peak!

P P

P P P P P P P P NP P

Panic/ No Panic

S NS

S S S S S S S S S S Not Severe

Severe

Severity Monetary Contraction2 of Cycle

1836 1845 1866

1839 1857 1873* 1883* 1890* 1907 1920 1929

Date of Peak!

P P P

NP P NP NP NP NP NP NP

Panic/ No Panic

NS S NS

S S S NS NS NS S S

Monetary Contraction2

Sources: See tables 10.2 and 10.3. 1. An asterisk denotes positive growth in real output from cycle peak to cycle trough. With the exception of 1873 in the U.S. and the U.K., these growth rates on average were all less than 1.0 percent; the average for 1873-79 in the U.S. was 3.7 percent; for the subperiod 1873-75, it was -1.1 percent; for 1873-79 in the U.K. it was 2.8 percent; for the subperiod 1875-79, it was 0.1 percent. 2. Monetary contractions are classified severe if the stock of money (or high-powered money in the case of the U.K. prior to 1870) either declined in absolute terms in the period during or immediately preceding the cycle peak or underwent a substantial decrease in its rate of growth.

Not Severe

Severe

Severity of Cycle

-------------------------------- United Kingdom --------------------------------

The Interrelation of Panics, Money and Cyclical Contractions in the United Kingdom and United States, 1836-1933

---------------------------------- United States ----------------------------------

Table 10.A.l

500

Wallace E. HutTman and James R. Lothian

1907, 1920, 1929 as severe cycles for both countries. For the United States we then added 1837, 1839, 1847, and 1882; and for the United Kingdom, 1839 and 1883. We had some doubts about the degree of severity of 1837 and 1847 in the United States and their two counterparts in the United Kingdom. On the basis of the output data in tables 10.2 and 10.3, we classified the two U.S. cycles as severe and the corresponding U.K. cycles as mild. Following Burns and Mitchell we did not include the 1913-14 contraction in the severe category. On the basis of the real output data alone, the phase clearly was severe. Had we so classified it, the case we make below would have been weakened but hardly overturned. Moreover, as Cagan (p. 223) points out, the phase is not very informative in any event since the panic was a "rather mild affair." By including both the United States and the United Kingdom in the period prior to 1875, we have thirteen severe cyclical contractions in addition to those Cagan examined-twelve accompanied by panics, one not, and three additional mild cyclical contractions, none accompanied by panics. As a glance at the table indicates, the deciding factor in a cycle's severity is the existence or absence of a monetary contraction rather than the existence or absence of a panic. Panics took place in only ten of eighteen severe cycles; severe monetary contractions took place in fifteen. In three of the five mild cycles during which a panic took place, the monetary contraction was also mild, and in only one (1845 in the United Kingdom) was there an absolute decline in money. The other interesting aspect of these data is the light they shed on the question of transmission. In only three of the common cycles-1836-38 (1837-38 in the United Kingdom), 1847-48 (1845-48 in the United Kingdom), and 1856-58 (1857-58 in the United Kingdom)-were there panics in both countries. In the first two, the fluctuations in output were a good deal more severe in the United States than in the United Kingdom. The importance of panics as a direct channel of transmission of cyclical fluctuations does not appear to have been great. As an indirect channel, that is through their effects on money supplies, panics appear to have exerted a more important influence.

Notes 1. Considerable debate has centered on this topic. See Pippenger's paper in this volume for evidence supporting this statement. 2. In Friedman and Schwartz's (1963a) study of business cycles, for example, changes in monetary growth were the causative variable. Fisher (e.g., 1935) related cyclical movements in real variables to distributed lags of past prices, the latter being identified by Fisher as an indicator of monetary pressure. 3. For a discussion of transmission in the Meade-Mundell framework see Mussa's excellent survey article (1980).

501

The Transmission of Business Cycles, 1833-1932

4. See Darby and Lothian 1983 for a discussion of how these various channels of transmission operated during the fixed-exchange-rate Bretton Woods period. 5. In appendix B we evaluate these two hypotheses. We conclude that the monetary decline, rather than the panic itself, was the major factor leading to cyclical contractions in output. 6. If the specie circular had been the cause of the monetary contraction, we would expect to see the ratio of money to specie rather than specie itself account for the decline in monetary growth. 7. Between 1876 and 1879, the money stock decreased at an average annual rate of 2.6 percent while high-powered money increased at an average annual rate of 1.7 percent. 8. Ultimately, however, some reduction in the U.K. money supply and price level would have had to occur given the reduction in both the United States and the rest of the world. A largely domestic-induced decrease in the money stock in this instance was the equilibrating factor. Had the decrease not occurred, an outflow of gold presumably would have been the main avenue through which monetary deflation would have taken place. 9. We stress the word "almost." The onset of the 1920-21 cycles poses a partilcular problem in this regard. Both countries experienced substantial monetary decelerations beginning at roughly the same time. The increase in the discount rate by the Federal Reserve and subsequent reaction by the Bank of England may have been the key factor here. 10. See Choudhri and Kochin 1980 for evidence drawn from a number of countries during the 1930s and Jonung 1981 for Sweden. Lothian 1981 contains a further discussion of the U.K. vs. U.S. comparisons presented here. 11. In the presentation and discussion of the empirical results, we concentrate exclusively on the Granger tests. An autoregressive system is difficult to describe succinctly. Moreover, it is difficult to make much sense of individual coefficients of the regressions equations since coefficients on successive lagged values of a given variable tend to oscillate in sign, and there tends to be a complicated pattern of cross-equation feedback. Additional insights into the performance of the system of equations could be obtained by analyzing the system's responses to typical random shocks. 12. Cassese and Lothian 1982 contains a discussion of some of these issues, in particular the relation between timing and causation in the context of international transmission of economic disturbances. C. Hernandez-Iglesias and F. Hernandez-Iglesias (1981) provide examples of models where economic causality may be difficult to verify with tests based upon Granger's predictive concept of causality. 13. The F-statistic is fairly robust to relaxing the assumption of normality of the errors (Judge et al. 1980). Estimation and testing with lagged endogenous variables rely on asymptotic distribution theory. Autocorrelated error terms are a serious potential source of problems. 14. Dhrymes (1970, pp. 34-40) presents the basic form of the test. Under the null hypothesis, we impose q linear restrictions on the coefficients of 1T(L). Applying the likelihood-ratio principle, we then arrive at the test statistic (T - k) 1n I!oo/!n I, which has an asymptotic X~ distribution where !oo and !n are estimates of the variance-covariance matrix of the error terms under the restricted system associated with the null hypothesis and of the general system respectively. This form of the statistic is due to Sims (1980) who argues that standard tests are biased toward rejecting the null hypothesis when q approaches or exceeds T - k in size. He therefore suggests treating the sample size as (T - k) rather than T in these cases. 15. Ideally we would have liked to have had quarterly or perhaps even monthly data. Annual data can of course mask the timing relationships that are central to our analysis. Unfortunately, however, no such intrayear data are available in continuous form for anything even close to our full sample period.

502

Wallace E. Huffman and James R. Lothian

16. Tests presented in the earlier version of this paper based on a slightly different body of data indicated possible heterogeneity of the model across these two subperiods. For this reason, we included the dummy variable in each of the equations. Additional tests of lag length were not inconsistent with the two-year distributed lags used here. 17. Since only lagged values of the variables appear on the right-hand side of the equations, these tests are likely to understate the importance of the arbitrage relationships. We therefore ran additional regressions, in the first case, of the contemporaneous percentage change in one country's price level on the other's and, in the second, of the contemporaneous first difference of the one country's interest rate on the other's. In both instances we also included a dummy variable for intercept shift in the second subperiod. The partial correlation between the price variables was .41 and between the interest-rate variables .52. Both are significant at better than the .01 level. The statistical significance uncovered in certain of the Granger tests, however, suggests that neither process was complete within the year. For the interest-rate relationship the lagged adjustment is suggestive of an asset-market transmission mechanism of the type posited by Branson (1968, 1970). For the price relationship, differences in the adjustment of prices of traded and nontraded goods are a possible explanation. Lags in adjustment in the goods and the bond markets, together with the successful intervention techniques followed by the Bank of England, suggest that the simplest monetary-approach models are inappropriate for the period. Similar conclusions for both the United States and the United Kingdom, as well as six other industrial countries during the post-World War II era, are presented in Darby and Lothian 1983. 18. Brittain 1981 and Miles 1978 contain evidence derived from post-World War II data for the U.S. and several other industrial countries and the U.S. and Canada, respectively, that is consistent with the currency-substitution hypothesis. Darby and Lothian, in summarizing the findings of the National Bureau study The International Transmission of Inflation (1983), present evidence that largely contradicts it. In their study of the United Kingdom and United States, Friedman and Schwartz (1982) find for the gold standard portion of their period that other-country money did not affect own-country nominal income in either instance but did affect the own-country price level in both. Since their data are averages taken over reference-cycle phases, the possibility of shorter-term effects on nominal (and real) income exists. Further compounding the problem is the evidence they present that such effects were significant for the United Kingdom but not for the United States post1914. The standard comment that further study of the question is required is, therefore, more appropriate than usual. 19. Friedman and Schwartz 1963 for the United States after 1870, Warburton 1958 and 1962 for the U.S. antebellum period, Huffman and Lothian 1980 for the United Kingdom in the nineteenth century contain results consistent with this conclusion.

References Anderson, T. W., and Taylor, John B. 1976. Conditions for strong consistency of least squares estimates in linear models. Technical report no. 213, Institute for Mathematical Studies in the Social Sciences, Stanford University. Ayres, Leonard P. 1939. Turning points in business cycles. New York: Macmillan. Berry, Thomas, Sr. 1968. Estimated annual variations in gross national product, 1789 to 1909. Richmond: University of Richmond.

503

The Transmission of Business Cycles, 1833-1932

Bordo, Michael. 1981. The classical gold standard: Some lessons for today. Federal Reserve Bank of St. Louis Review 63: 1-17. Branson, William H. 1968. Financial capital flows in the U.S. balance of payments. Amsterdam: North-Holland. - - - . 1970. Monetary policy and the new view of international capital movements. Brookings Papers on Economic Activity 2: 235-62. Brittain, Bruce. 1981. International currency substitution and the apparent instability of velocity in some western European economies and in the United States. Journal of Money, Credit, and Banking 13: 135-55. Burns, Arthur M., and Wesley Mitchell. 1946. Measuring business cycles. New York: National Bureau of Economic Research. Cagan, Phillip. 1965. Determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. Cassese, Anthony, and James R. Lothian. 1982. The timing of monetary and price changes and the international transmission of inflation. Journal of Monetary Economics 10: 1-23. Choudhri, E., and L. Kochin. 1980. International transmission of business cycle disturbances under fixed and flexible exchange rates: Some evidence from the Great Depression. Journal of Money, Credit, and Banking 4: 565-74. Clapham, (Sir) John. 1945. The Bank of England (1797-1914). New York: Macmillan. Collins, Michael. 1981. The English banking sector and monetary growth, 1844-80. Discussion paper, University of Leeds School of Economic Studies. Darby, Michael R., and James R. Lothian. 1983. Conclusions on the international transmission of inflation. In The international transmission of inflation , ed. Michael R. Darby, James R. Lothian, and Arthur E. Gandolfi, Anna J. Schwartz, and Alan C. Stockman. Chicago: University of Chicago Press. Darby, Michael R., and Alan C. Stockman. 1983. The mark III international transmission model: Estimates. In The international transmission of inflation. See Darby and Lothian 1983. Davis, L. E., and J. R. T. Hughes. 1960. A dollar-sterling exchange, 1803-1895. Economic History Review 13: 52-78. Deane, Phyllis. 1968. New estimates of gross national product for the United Kingdom, 1830-1914. Review of Income and Wealth 14: 96112. Dhrymes, Phoebus. 1970. Econometrics. New York: Harper and Row. Feinstein, Charles H. 1972. National income, expenditure and output of the United Kingdom, 1855-1965. Cambridge: Cambridge University Press. Fisher, Irving. 1920. Stabilizing the dollar. New York: Macmillan. - - - . 1935. Are booms and depressions transmitted internationally

504

Wallace E. HutTman and James R. Lothian

through monetary standards? Bulletin deL'Institut International de Statistique (London) 28: 1-29. Francis, John. 1862. History of the Bank of England. New York: Banker's Magazine. Friedman, Milton. 1968. The role of monetary policy. American Economic Review 58: 1-17. Friedman, Milton, and Anna J. Schwartz. 1963a. A monetary history of the United States, 1867-1960. Princeton: Princeton University Press. - - - . 1982. Monetary trends in the United States and the United Kingdom: Their relation to money, prices, and interest rates, 1874-1975. Chicago: University of Chicago Press. - - - . 1963b. Money and business cycles. Review of Economics and Statistics 45, suppl: 32-64. Gallman, Robert E. 1966. Gross national product in the United States, 1834-1909. Output, employment, and productivity in the United States after 1800. Studies in Income and Wealth, no. 30. New York: National Bureau of Economic Research. Gayer, Arthur D., Walter W. Rostow, and Anna J. Schwartz. 1976. The growth andfluctuation of the British economy, 1790-1850. 2d ed. New York: Barnes and Noble. Granger, C. W. J. 1969. Investigating causal relations by econometric model and cross-spectral methods. Econometrica 37: 424-38. Hawtrey, R. G. 1947. The gold standard in theory and practice. London: Longmans, Green and Company. Hernandez-Iglesias C., and F. Hernandez-Iglesias. 1981. Causality and the independence phenomenon. Journal of Econometrics 15: 247-63. Huffman, Wallace, and James R. Lothian. 1980. Money in the U.K., 1833-1880. Journal of Money, Credit, and Banking 2: 155-75. Johnson, Harry G. 1976. The monetary approach to the balance-ofpayments theory. In The monetary approach to the balance of payments, ed. J. Frenkel and H. G. Johnson, 147-67. Toronto: University of Toronto Press. Jonung, Lars. 1981. The depression in Sweden and the United States: A comparison of causes and policies. In The Great Depression revisited, ed. Karl Brunner, vol. 2, 286-315. Rochester Studies in Economics and Policy Issues. Boston: Martinus-Nijhoff. Judge, George, William Griffith, R. C. Hill, and T. Lee. 1980. The theory and practice of econometrics. New York: John Wiley. Juglar, Clement, and W. Thomas DeCourcy. 1916. A brief history of panics and their periodical occurrence in the United States. New York: Knickerbocker. Keynes, John Maynard. 1930. A treatise on money. London: Macmillan. Kindahl, James K. 1961. Economic factors in specie resumption: The United States, 1865-1879. Journal of Political Economy 59: 30-48.

505

The Transmission of Business Cycles, 1833-1932

Kindleberger, Charles. 1978. Manias, panics, and crashes: A history of financial crises. New York: Basic Books. Leiderman, Leonardo. 1980. Relationships between macroeconomic time series in a fixed exchange rate economy. European Economic Review 14: 61-77. Lothian, James R. 1976. Interwar Britain: A comparison with American experience. Mimeo. New York, Citibank, N.A. - - - . 1981. Comments on "monetarist interpretations of the great depression." In The Great Depression revisited, vol. 2, pp. 134-47. See Jonung 1981. Macaulay, Frederick. 1938. The movement of interest rates, bond yields, and stock prices in the United States since 1856. New York: National Bureau of Economic Research. Macleod, Henry Dunning. 1896. A history of banking in Great Britain. InA history ofbanking in all leading nations, vol. 2. New York: Journal of Commerce and Commercial Bulletin. Macesich, George. 1960. Sources of monetary disturbances in the United States, 1834-1945. Journal of Economic History 20: 407-34. Matthews, R. C. O. 1954. A study in trade-cycle history: Economic fluctuations in Great Britain, 1833-42. Cambridge: Cambridge University Press. Miles, Marc A. 1978. Currency substitution, flexible exchange rates, and monetary interdependence. American Economic Review 68: 428-36. Mitchell, B. R., and Phyllis Deane. 1962. Abstract of British historical statistics. Cambridge: Cambridge University Press. Mussa, Michael. 1980. Macroeconomic interdependence and the exchange rate regime. In International economic policy: Theory and evidence, ed. R. Dornbusch and J. A. Frenkel. Baltimore: Johns Hopkins University Press. Officer. Lawrence H. 1981. The floating dollar in the greenback period: A test of theories of exchange-rate determination. Journal of Economic History 41: 629-50. Rockoff, Hugh. 1971. Money, prices, and banks in the Jacksonian era. In The reinterpretation of American economic history, ed. R. W. Fogel and S. L. Engerman. New York: Harper and Row. Rutner, Jack L. 1974. Money in the antebellum economy: Its composition, relation to income, and its determinants. Ph.D. diss., University of Chicago. Saidi, Nasser H. 1980. Fluctuating exchange rates and the international transmission of economic disturbances. Journal of Money, Credit, and Banking 12: 575-91. - - - . 1982. Expectations, international business cycles, and the balance of payments. Journal of Money, Credit, and Banking 14: 327-46.

506

Wallace E. HutTman and James R. Lothian

Sargent, Thomas J. 1976. A classical macroeconomic model for the United States. Journal of Political Economy 84: 207-37. Sayers, R. S. 1976. The Bank of England, 1891-1944. Cambridge: Cambridge: Cambridge University Press. Sheppard, David K. 1971. The growth and role of U.K. financial institutions, 1880-1962. London: Methuen and Co. Silberling, Norman J. 1943. The dynamics of business. New York: McGraw-Hill. Sims, Christopher A. 1980. Macroeconomics and reality. Econometrica 48: 1-48. - - - . 1982. Policy analysis with econometric models. Brookings Papers on Economic Activity 1: 107-64. Smith, Walter Buckingham, and Arthur H. Cole. 1935. Fluctuations in American business, 1790-1860. Cambridge: Harvard University Press. Sumner, William G. 1896. A history of banking in the United States. New York: Journal of Commerce and Commercial Bulletin. Temin, Peter., 1969. The Jacksonian economy. New York:' W. W. Norton. - - - . 1968. The economic consequences of the bank war. Journal of Political Economy 76: 257-74. - - - . 1974. The Anglo-American business cycle, 1820-60. Economic History Review 27: 207-21. - - - . 1975. The panic of 1857. Intermountain Economic Review 6: 1-12. Timberlake, R. H., Jr. 1978. The origins of central banking in the United States. Cambridge: Cambridge University Press. United Kingdom. Parliament. 1857. Report from the select committee on the Bank Acts. London: HMSO. U.S. Department of Commerce. Bureau of the Census. 1960. Historical statistics of the United States: Colonial times to 1957. Washington, D.C.: Government Printing Office. - - - . 1973. Long-term economic growth, 1860-1970. Washington, D.C.: Government Printing Office. Van Vleck, George W. 1943. The panic of 1857. New York: Columbia University Press. Warburton, Clark. 1958. Variations in economic growth and banking developments in the United States from 1835 to 1885. Journal of Economic History 18: 283-97. - - - . 1962. Monetary disturbances and business fluctuations in two centuries of American history. In In search of a monetary constitution, ed. Leland B. Yeager. Cambridge: Harvard University Press. Williamson, J. G. 1961. International trade and United States economic development, 1827-1843. Journal of Economic History 21: 372-83.

507

The Transmission of Business Cycles, 1833...,.1932

Zellner, Arnold. 1979. Causality and econometrics. In Three aspects of policy and policymaking: knowledge, data and institutions, ed. K. Brunner and A. H. Meltzer. Carnegie-Rochester Conference Series on Public Policy 10. Amsterdam: North-Holland.

Comment

Michael Connolly

The major finding of this study is that for each country the proximate determinant of output fluctuations are sudden, unanticipated changes in domestic monetary variables. Transmission between countries occurs via specie flows and the monetary reactions they induce, either on the part of the monetary authorities or on the part of the banking system. In my comments, I will argue that some evidence, particularly of the historical-narrative kind, is provided in support of this finding. But I am less convinced by the econometric evidence. Evidence on Real-Income Interdependence Two bodies of evidence are offered in support of the international transmission of business cycles. The first is an anecdotal narrative of the major contractions in the United States and the United Kingdom over the one hundred-year period; the second is econometric, involving GrangerSims autoregressive tests of causality for the combined period 1837-59, 1882-1914. The historical narratives are to some extent convincing, the econometric testing is less so. The major U.K. recessions were transmitted to the United States, it is argued, during the antebellum period, principally by the United Kingdom's raising Bank rate, thus triggering a slowdown of growth or an outright loss of gold in the United States. Each recession is documented and was frequently shared by both countries during this gold standard period. The extent to which recessions were transmitted from the United Kingdom to the United States, however, is undoubtedly exaggerated by the use of the Smith and Cole and the Ayres indexes of total trade for U.S. income from 1833 to 1861. As described in appendix A, these indexes contain two-thirds domestic trade and one-third foreign trade. This series is much too volatile and, more importantly, very likely to be biased toward procyclical movements with U.K. income. The reason is simple: When the rate of growth of U.K. income expands or contracts, Michael Connolly is professor of economics at the University of South Carolina.

508

Wallace E. HutTman and James R. Lothian

U.K. imports from the United States will expand or contract. Consequently, the U.S. output index will, I suspect, reflect too dramatically the decline in U.K. income simply because it includes such a high proportion of foreign trade. These problems are illustrated in figure C10.1 which I have drawn from data supplied by Wallace Huffman and James Lothian (hereafter H-L). Expansions and contractions in the U.K. growth rate are accompanied by multiplied expansions and contractions in the U.S. growth rate the same year or a year later. Personally, I suspect a problem with the U.S. income index, although I agree with H-L that the Gallman index is unsuitable since it excludes inventories, a particularly sensitive boom-and-slump item. In any event, to the extent that the U.S. output index is biased towards procyclical behavior, the findings will be also. I will not review each contractionary episode and the extent to which H-L argue it was shared under the gold standard. In most cases it is clear recessions were shared; in others it is not so clear. I am puzzled, however, by their interpretation of the "controlled experiment" of the 1915-25 interwar preiod when the United Kingdom was off the gold standard. Following the sharp 1919-21 U.K. slump, U.S. income fellS percent in 1920 and 4 percent in 1921, suggesting that the United States shared in the slump despite the nonexistence of the gold standard. It may be that the rebound in the U.S. economy was both rapid and strong; the U.K.

o o N !")

UK GROWTH IN REAL GNP

PERCENT

uS GROWTH IN REAL GNP o

o

(YEARLY DATA 1834 TO 1933)

o o N

!")+-----r---.-----r----.,---,..----r---r---r-----r---~-___,

'i83500

1844·00

Fig. CIO.l

1853.00

1862.00

u.s.

1871.001880.001889.00

i898.00

1907·00

1916.00

192500

1934.00

and U.K. real GNP, yearly growth rates, 1834-1933.

509

The Transmission of Business Cycles, 1833-1932

economic rebound was weaker and came later. But the fact is that both economies slumped together-the United States was not insulated from the British drive to restore the gold standard at the prewar parity by deflation. I am convinced by most of the narrative stories, but in some the authors use poetic license. As for the weak Granger-Sims causality tests reported in table 10.5, there appear to be some relationships between U.S. income and price variables and U.K. income and price variables. U.S. real-income growth is Granger-caused by U.K. monetary variables (gold, Bank rate, and high-powered money), but not by U.S. gold nor U.K. income. In turn U.K. income growth is Granger-caused by U.S. high-powered money, prices, and discount rate, but surprisingly not by U.K. prices or Bank rate. The strong causality tests reported in table 10.6 show U.S. income Granger-caused by only U.K. prices and U.S. population growth. U.K. income is not Granger-caused by any variable. These tests have problems to which I will return. Evidence on the Transmission Mechanism In general, two types of transmission mechanisms are possible: one through the direct effects of international-price and interest-rate arbitrage, the other through flows of specie and consequent effects on money growth rates. The first is not tested in this paper because the autoregressive tests impose a one-year lag on variables for the purpose of identifying causality. Since synchronous variables are excluded, the test cannot pick up rapid arbitrage. In the H-L tests, U.S. inflation is not related to lagged U.K. inflation, but U.K. inflation is weakly caused by past U.S. inflation. Similarly, changes in U. S. interest rates are not Granger-caused by lagged changes in U.K. interest rates, but changes in U.K. Bank rate are weakly caused by changes in lagged U.S. interest rates. Not much information on arbitrage can be drawn from this evidence because of the long lags involved. To get at some of these relationships, it might be useful to use a priori information on the direction of causation to justify the inclusion of some contemporaneous variables. For example, the authors argue that contractions prior to 1870 in the United States are a dance to the tune of U.K. Bank rate. In clearcut cases, contemporaneous U.K. income, prices, and Bank rate might be included in the regressions of U.S. income. (Choudhri 1983 uses this technique.) In any event, this would give greater evidence of association if not causality and, in particular, would shed light on the price- and interest-arbitrage relationship. The narrative evidence presented for 1833-70 focuses upon the importance of jumps in U.K. Bank rate in inducing slowdowns or declines in U.S. specie and consequently in U.S. money growth, which then slows

510

Wallace E. HutTman and James R. Lothian

u.s. income. A glance at figure CIO.2, also reproduced from H-L's data, does suggest this pattern in the antebellum period. Declines in U.K. specie led to increases in Bank rate, provoking shortly thereafter either a slowdown or decline in U.S. specie. In the 1836-37 period, U.S. gold growth slowed; in 1840 it declined 7 percent; it slowed in 1857 and in 1882 and declined 4 percent in 1889. In the Great Contraction, the U.K. break with gold in September 1931 probably enabled money growth to expand there sooner and more rapidly (see figure C10.3). This type of evidence is quite strong-in many of the major contractions, the slowdowns and declines in U.S. gold had similar effects on the U.S. money supply. In H-L's terms, however, the movements in gold often fail to account for anything close to the full movement in M2. Furthermore, in several instances there was little correspondence between gold and M2, as reported in tables 10.2 and 10.3. Nevertheless the evidence they report on some specific episodes appears quite strong. I would like to conclude with some remarks on the possible pitfalls of the autoregressive tests reported in the H-L paper. First, as mentioned above, lagged relationships using annual data suppress too much information and do not allow for rapid price and interest arbitrage and/or o

o o

[~C[IIT

If)

.

UK CHANGE IN GOLD uS CHANGE IN GOLD

o o

o

(YEARLY DATA 1834 iO 1933) I'

" I'

g o

I""l

I

1\

I' I'

I

I

'

,\

I I' ,,1

I I

\1 I

I I

o

o o o

(")

o

o

(")

o

N

o(") o

I""l+-----.---_--...,.......---......---r-----....--........---...-------.---~-____,

1.835.001844.001853.001862.00187\.001880.00

Fig. CI0.2

i889·00

i898.00

.9:)'7.00

1916.00

i92S00

i9J4.0{1

U.S. and U.K. annual percent change in monetary gold stock, 1834-1933. The United States was not on the gold standard 1861-79; the United Kingdom was not on the gold standard 1915-25 and from 1931 on.

511

The Transmission of Business Cycles, 1833-1932

o o o PERUNI

..

UK MONEY GROWTH LJS MONEY GROWTH

o o

(YEARLY DATA 1834 TO 1933)

,..,N 1

g\

.. '

NI

I

,

g

I

~

I

,

1

I

g :

\/1

I,.

I

: "

\,'

~ "~

1 J)

:

: J

'

I

I ,

o

,'1'1 '1:1 I " I I 11 I LJ I 'I

I ,

I

o

I~ II I,

,":::

1

g

II

~J

~;

I I

\

I A I 111I 11 '1

\

\

I'"

o o

I I , I

I II

ID

II

",

~

"

I o o

..

N+--___,---.,----,----~-___,--__r_--_r_--~-__,.--__r_-___,

1'1835.00

18H.OO

Fig. CI0.3

1853.00

1862.00

1871.00

1880.00

1889.00

1898.00

1907.00

1916·00

1925.00

1934·00

U.S. and U.K. yearly money growth rates, 1834-1933.

causation. tests. Second, reporting only the F-statistic, as is common in these tests, gives us no idea about the magnitude or even the sign of the coefficients in the regressions (on this, see note 11). Third, as specified, the test clearly picks up spurious causality; for example, U.S. population growth Granger-causes U.K. domestic-credit growth, and U.S. domestic-credit growth causes U.K. population growth. To sum up, I found the paper interesting and some of the evidence convincing. Reference Choudhri, Ehsan. 1983. The transmission of inflation in a small economy: An empirical analysis of Canadian inflation, 1962-80. Journal of International Money and Finance 2 (Aug.): 167-78.

11

Real Output and the Gold Standard Years, 1830-1913 Stephen T. Easton

The topic "real output and the gold standard" describes in some sense the ultimate objective of our quest to balance the costs and benefits of the gold standard. Per capita real income, the aggregate embodiment of the national standard of living, is a touchstone of any broad macroeconomic policy. If we can show, for example, that real income is higher or grows faster under a set of rules termed a gold standard than under alternative monetary arrangements, then we are well on the way toward new policy prescriptions and a truly new economic order. This paper makes no such claims. Instead I focus on a much narrower and less dramatic set of issues that may ultimately, but not immediately, shed light on the grander question. My task, here, is to characterize the behavior of real output in several nations that were linked in several ways during the most sustained period of a worldwide operating gold standard. The paper consists of two sections. Section 11.1 deals with eight nations in Europe and North America during the nineteenth and early twentieth centuries to determine whether there was an Atlantic econollJY in the sense that changes in real output in one country were eitb~r correlated with or caused changes in real output in other nations. In addition, I try to discover whether the general move to the gold standard made an appreciable difference to the links among national incomes. Section 11.2 asks whether there is any evidence to support the naturalrate hypothesis during the gold standard years in much the same way as Lucas (1973) examined the output-inflation tradeoff of the post-World Stephen T. Easton is associate professor of economics at Simon Fraser University, Burnaby, British Columbia. The author is indebted to John Chant, Peter Coyte, and Anna Schwartz for helpful comments, and to Kelly Busche, Reg Cameron, and Jim Irwin for able research assistance. The author of course bears responsibility for remaining errors.

513

514

Stephen T. Easton

War II period; then the model is extended to test whether the price level is more appropriately characterized as endogenous or exogenous to most countries during the period. To state the conclusions most succinctly: There is little evidence of an Atlantic economy as far as real-income movements are concerned, there is some evidence in favor of the naturalrate hypothesis, and in some countries the elasticity of aggregate demand is very large. 11.1 The Behavior of Real Output before and after 1879 11.1.1

Dating the Gold Standard 1

The years between 1879 and 1914 are usually referred to as the gold standard period. In this era the United States joined the United Kingdom and most of the rest of the economic world in adopting the gold standard. Britain eliminated restrictions on gold exports and required the Bank of England to redeem its notes in gold (or coin) in 1821. By 1850 both the United States and France had effectively moved to the gold standard.2 During the 1860s, the United States joined Germany, Italy (1866), Russia, and Austria-Hungary with inconvertible currencies. In July 1873 Germany moved to the gold standard, and by 1879 most nations had more or less adopted policies consistent with gold. The exceptions were Japan, India, Russia, and Austria-Hungary. These, too, joined the rest of the world in 1895 or so, and the regime persisted until the start of World War I, although Italy formally adhered to the gold standard only during the decade 1884--94.3 11.1.2 Models of Real Output during the Gold Standard Years Most discussions of the gold standard tend to focus on how institutions or economic variables worked under that regime-the behavior of banks and banking systems, relative prices and price levels, interest rates, foreign trade, the balance of payments and gold flows.4 In most macroeconomic models these variables are associated with changes in real income. National macroeconomic models typically have several windows to the rest of the world. In the case of Keynesian income-expenditure models, these windows include terms-of-trade effects, the direct effect of foreign demand on domestic goods, and a relationship between foreign and domestic interest rates that affects capital flows. Recent monetarist models tend to stress a natural-rate hypothesis coupled with assumptions about the way in which expectations are formed, making a sharp distinction between anticipated and unanticipated magnitudes of exogenous variables.5 Since these models typically focus on quarterly or yearly fluctuations over relatively short periods, it is not surprising that they ignore other channels by which real output may be altered. Most notably

515

Real Output and the Gold Standard Years, 1830-1913

in the nineteenth century, the movement of labor internationally and the opening of new land for settlement had some effects on the behavior of real income.6 The usual model of the nineteenth-century gold standard has tended to examine the interrelationships among exogenous variables in one country and the way in which, say, the balance of payments, trade, and capital account behaved in the other. The question posed here is a different one. It is not whether exports, for example, increase when there is an exogenous rise in income abroad, but rather whether the total effect of the increase in income abroad is enough to raise income significantly in another country. By implicitly aggregating across all the channels by which international excess demands are transmitted, this study seeks to discover whether national incomes, nominal or real, are linked. Two works that stand as major efforts to explain the behavior of real output during the gold standard are those by Oskar Morgenstern (1959) and Brinley Thomas (1973). Morgenstern studies the behavior of four countries-the United States, the United Kingdom, France, and Germany-by looking at NBER-reference-cycle peaks and troughs. His analysis suggests that during the gold standard period (prior to World War I), the association of the business cycle of the three European nations was relatively close, and that between Europe and the United States was less close. To reach these conclusions, Morgenstern forms a contingency table of months in which various countries were in similar phases of their business cycles-up phases or down phases-and finds a low probability that the observed number of concurrent up and down phases would be observed by chance (pp. 51-73). This methodology is flawed in several ways. First, the reference-cycle data do not abstract from the underlying growth rate observed in all countries. The approach yields a closer association of business cycles, measured as months of shared up or down phases, than would be true if the cycles, net of the underlying growth rates, were measured. And second, with only six to ten business cycles observed in the four countries, any relationship that relies on the cycle itself as a fundamental unit of observation has very few degrees of freedom. Since only up and down movements of the cycle are examined, amplitude of movements is ignored. In addition, the average duration of the cycles among the four countries-between forty-three and seventy months-makes it difficult to accept inferences about the intercountry relationships among the cycles? Brinley Thomas's Migration and Economic Growth (1973) is probably the most detailed analysis of the period. Thomas's model organizes the data about a more or less informal open Keynesian multiplier-accelerator view of the world in which the United Kingdom is the hub. Exports and capital formation depend upon the level of population and migration which in turn depend upon both the natural growth of population and

516

Stephen T. Easton

relative income levels among countries. The gestation period for investment and the natural cycle of population provide for the complex lags observed in a myriad of time series that Thomas correlates. The gold standard provides the international regime, but interest rates set in the United Kingdom mark the tempo to which the nominal monetary aggregates dance. He rejects the monetary approach to the balance of payments out of hand. Thomas finds interrelationships between a variety of time series among nations. He builds a coherent explanation primarily from observations that particular series peaked and troughed together, or that one followed another at a reasonable lag. The various peaks and troughs are often strikingly apparent, but little is done to relate the series to one another in a statistical sense. Thomas is usually concerned with long cycles----of ten or twenty years in duration-but he does not estimate systematic behavior explicitly, and the series he studies have wide variations in periodicity. My task is considerably less ambitious. I focus on a single aggregate measure, real GNP, and ask whether there is any evidence of the Atlantic economy in the behavior of that measure among countries. By breaking the time period in two-pre-1879 and post-1878-1 try to assess possible differences introduced by the United States' move to gold. The strategy is first to look at simple correlations among country real outputs, much as Thomas does, to determine whether the relationships that he finds can be observed, and then to apply Granger-Sims tests to determine whether movements in income of one country cause income to change in another. 11.1.3 Simple Correlations Tables 11.1 and 11.2 provide the simple correlations among the outputs of a number of nations. The unit of real output is the annual deviation of the log of actual GNP from the log of (exponential) trend GNP. Using this smoothing device is customary, although more sophisticated filters to achieve mean and covariance stationary processes might prove fruitful. The data limitations are rather severe. The gold standard period lasted at most a mere thirty-five years, and some have argued that it lasted only fifteen. Since real-output series for several countries are available only since 1870, the data are limited in the other direction as well. Tables 11.1 and 11.2 give a bird's-eye view of what a simple year-byyear comparison among real outputs reveals. In the tables, a blank means that the correlation was not statistically significant at the 0.10 level. Although I corrected for autocorrelation in the residuals of the correlations, I show the uncorrected simple correlations in table 11.1. Eyeball comparisons of peaks and troughs would not correct for such autocorrelations and the correction might suggest a misleadingly sanguine view of the simple relationships. In table 11.2, I note shifts from significant to insig-

517

Real Output and the Gold Standard Years, 1830-1913

Table 11.1 No. of Observations US CN UK GER ITL DEN NOR SWD

Simple Correlations of Real-Output Deviations from Trend,

183~79

US

CN

UK

GER

ITL

DEN

NOR

SWD

(37)

(12)

(50)

(30)

(19)

(10)

(15)

(10)

X

- .6** X X

.3* X

-.5*

.65**

.5* .6**

.7**

X

.6* X X

.6* .6* X

Sources: US = United States values using Gallman (1968) data from 1830; no observations from 1860 to 1869. CN = Canadian values from 1869 (Dick 1978). UK = United Kingdom based on Mitchell (1975) from 1830 and Feinstein (1972) from 1855. GER, ITL, DEN, NOR, SWD = Germany, Italy, Denmark, Norway, and Sweden based on Mitchell 1975. Notes: The table shows correlation coefficients between country pairs of log deviations from the log of the trend of real GNP (or NNP). No entry = no significant correlation at the 0.10 level; * = significant at the 0.10 level; ** = significant at 0.05 or better.

Simple Correlations of Real-Output Deviations from Trend, 1879--1913

Table 11.2

US CN UK GER ITL DEN NOR SWD

US

CN

UK

GER

ITL

DEN

X

.5** X

.4**

- .7** -.4' - .4** X

.3* - .5**

-.02*'

X

.3* X

X

NOR

SWD

.3' - .3*

.3' .5' .4** X

.6** .5** .6**

X

Sources: Same as for table 11.1. Notes: There are 35 observations for each correlation. No entry = no significant correlation at the 0.10 level; * = significant at the 0.10 level; ** = significant at the 0.05 level; , = not significant at the 0.10 level when autocorrelations accounted for although apparently significant before the correction; *' = significant at 0.10 level after correction for autocor· relation.

nificant coefficients depending on whether correction is made for autocorrelation in the residuals.

1830-79 During the years prior to the United States' return to the gold standard excluding the Civil War decade, there is no evidence of a simple associa-

518

Stephen T. Easton

tion of real output in the United States and the United Kingdom. The U.S. measure is negatively associated with Canadian output and with no other country measure although, with the exception of the United Kingdom, there are few observations for correlations with other countries. Canadian output, like that of the United States, appears to be independent of the measure for the United Kingdom and the rest of Europe. Fluctuations in output in the United Kingdom are associated more or less strongly with output changes in Norway, Denmark, and Germany, and are negatively related to those changes in Italy. Changes in real output in Germany also appear to be related to those in the Scandinavian countries and, to a leser extent, to those in the United Kingdom. A degree of positive association exists among real outputs of the Scandinavian countries. Thus in this early period some association of outputs is evident within Europe, with the United Kingdom displaying the most significant number of associations in the group of countries I examine. Little association of output is evident between Europe and North America. 1879-1913 During the gold standard era, U.S. output appears to be positively associated with that in the United Kingdom, Canada, and Norway, and negatively associated with that in Germany. Canadian output, however, as in the earlier period, remains unrelated to output in the United Kingdom, but is positively associated with output in the United States. Output in the United Kingdom now is negatively correlated with output in Germany, Italy, and Norway, whereas the association in the earlier period had tended to be positive. That negative relationship is the one that Thomas finds and upon which several of his hub-periphery interpretations of the time series depend. Real output in Germany is negatively associated with that in both Canada and the United States and in the United Kingdom and unrelated to other European outputs. Real output in the Scandinavian countries and Italy appears positively associated more strongly than before 1879, and the negative relationship between Italy's and the United Kingdom's real output persists as in the earlier period. Real output in Italy also displays a slight positive correlation with that in Canada. Note that the output links among Denmark, Norway, and Sweden become much stronger in the later period. In 1885 central-bank reserves of anyone country could be located in central banks of the other two and still serve as a basis for domestic currency issue. Significant simple correlations among real-output changes in many countries provide no evidence of causal impetus. They do not by themselves support the notion that the United Kingdom was the hub around which at least some nations revolved. Common cyclical movements

519

Real Output and the Gold Standard Years, 1830-1913

among the European nations are not sufficient evidence that they were in a state of causal dependency upon one another.8 Although the associations in tables 11.1 and 11.2 provide a method for analyzing the behavior of real-output movements-one that underlies the casual comparison of time series by many authors-there is an alternative methodology-the Granger and Sims tests for causality. 11.1.4 The Granger Test One method of defining causality has been to say that X causes Y if past values of X significantly affect Y, once past values of Y have been considered. In a regression framework, if the bis of equation (1) are significant, then X is said to cause Y (Granger 1969). 00

00

Yr = ao + I ai Yr-i + I biXt - 1 + Ut ,

(1)

i= 1

i= 1

where U t is white noise. By placing X on the left-hand side one can also test to see whether Y causes X. The Sims test (1972), an alternative that amounts to the same thing asymptotically, is to regress Y on both past and future values of X. If the coefficients on the future values of X prove significant, i.e., the ais in equation (2), the conclusion is that Y causes X. 00

(2)

Yr

00

= ao + i=l I aiXt+l + I biXt - 1 + Ute i+O

Sargent (1979, pp. 277-92) provides a helpful discussion of the theory underlying the test and some examples. Table 11.3 reports the results of a Granger test, equation (1) with four lags, on the bivariate relationships among changes in national outputs.9 Table 11.3 should be read as follows. Row-head country output causes Granger Causal Links between National Real Outputs, 1881-1913

Table 11.3

US US eN UK GER

UK

GER

ITL

x

.25 .25

.25

.10

X

.25 .25

X

.25 .05

DEN

NOR

SWD

.10 .25

.25 .05 .01

.25

X

.25 .25

ITL

DEN NOR SWD

eN

.10

.25 .10

X

.05

X

.25 X

.25

.05

.25 .05

.25

X

.25

Notes: Read across the rows for causal direction. Row-head country causes column-head country at the significance level in the table. Each column-head country is caused by the row-related elements in the column. A blank indicates a significance level greater than 0.25.

520

Stephen T. Easton

column-head country output at the significance level reported in the table. In the first row there is no causal connection between output changes in the United States and Canada at the 0.25 significance level or better. There is, however, a causal relationship flowing from the United States to the United Kingdom at the 0.25 significance level. According to the United States column, output in the United States is caused by output in the United Kingdom and in Norway, each at the 0.25 significance level, and in Germany at the 0.05 confidence level. These unusually lax significance levels have been chosen so as to give as much latitude as possible for the display of bivariate relationships. The key feature of table 11.3 is that of the 56 possible bivariate causal relationships, 27 are significant at the 0.25 confidence level, of which 10 are significant at the 0.10 level, 6 at the 0.05 confidence level, and 1 at the 0.01 confidence level. This pattern casts considerable doubt on any systematic causal relationships among the outputs. tO At the usual 0.05-orbetter significance level, Canada and the United Kingdom cause Norway, Germany causes the United States, Denmark and Sweden cause Italy, and Sweden causes Denmark. These relationships are not very attractive since the output of small countries appears to be determining that of large countries more often than the other way round. Certainly there is almost no evidence of a systematic hub-periphery relationship between the United Kingdom and other countries and little evidence to suggest a large number of bivariate causal relationships during the gold standard period. The simple correlations of tables 11.1 and 11.2 may arise either from a common external force or from concurrent domestic conditions, but they do not indicate significant bivariate causation between national outputs. To see what patterns might emerge if the period is not limited to the classical gold standard years, 1881-1913, table 11.4 examines the same patterns of bivariate causal links for different dates. In table 11.4-as in table 11.3, row-head nation causes column-head nation-there is little evidence of systematic relationships, although the United Kingdom does seem to display more causal links at more exacting significance levels than any other nation, and its links with Germany and Canada are not unreasonable. The reasons for such poor causal links among nations may be due to measurement errors in the data that raise the noise-to-signal ratio beyond acceptable bounds. Relatively few observations are available at best. To supplement the Granger causal relationships, table 11.5 uses a Sims test to examine the full range of data available for each country.11 Once again the pattern of causations reflected in table 11.5, like that of tables 11.3 and 11.4, shows little evidence of systematic causal relationships. It may be reasonable that U.S. and U.K. real outputs caused Canadian real output, but it seems odd that Denmark's real output should have a significant effect on that of Italy.

1873-1913: .05

-

-

CN UK GER

ITL

DEN NOR SWD

-

X 1872-1913: .05

1873-1913:

-

-

X

-

1838-59: .10 1873-1913:

UK(1830)

-

1872-1913: 1854-79: .10 X 1865-79: .25 1865-1913: .25

GER(1850)

1877-1913: .25

X

-

1874--1913: .05 X 1877-1913: .05

1869-1913: .25

-

-

1873-1913: .25 1869-1913: .10 1869-1913: .25

NOR(1865)

1874-1913: 1874-1913 1874-1913:

1873-1913: .10

DEN(1870)

1874-1913: .25

X

1872-1913: .25 1865-1913: .25

1871-1913:

ITL(1860)

1877-1913: .25 1877-1913: .25 X

1877-1913: 1877-1913: .10 1877-1913: .25

1877-1913: .25

SWD(1873)

Notes: Read across the rows for causal direction. The figure beside the country header is the year data become available. If the significance level was not at least 0.25, only the years covered by the test appear. In each case the test covered dates common to both sets of country data.

1873-1913: .25

X

CN(1869)

Granger Causal Links between National Real Outputs, Various Periods

US

US(1830)

Table 11.4

522

Stephen T. Easton Sims Causal Links between National Real Outputs, 1830-1913

Table 11.5

US eN UK GER

US

eN

X .22

x

DEN NOR SWD

GER

ITL

DEN

NOR

SWD

.17

.06 .05

.06

.05 .05

ITL

UK .12 X .10

X .10

.15

.10

.16

X .05

X

.15 X

.10

X

Notes: Read across the rows for causal direction. Row-head country causes column-head country at the significance level in the table. Each column-head country is caused by the row-related elements in the column. A blank indicates that the significance level was not at least 0.22.

A natural objection to the above approach is that it imposes too rigid a relationship between the outputs of the countries in our sample. In particular the Granger-Sims tests yield a causal association between outputs only if the same causal pattern is observed throughout the series. For this to be the case in the context of national-output movements, a disturbance in the United Kingdom, say, must be transmitted to the United States in every episode in the same fashion. This relationship may be reasonable if the United Kingdom is thought to be the leading nation, and all shocks have the same effect upon the participants. For example, an increase in U.K. output always raises or lowers output in the other country. But consider the possibility-suggested by Geoffrey Wood in commenting on an earlier version of this paper-that in one episode a domestic shock to. demand tends to stimulate output abroad, and in another episode the shock to domestic supply tends to reduce output abroad. That possibility is the one most likely to obscure the international causal links between real outputs-though one can easily imagine a model in which a positive supply shock stimulates the foreign economy as well. When there is no simple positive or negative association between national incomes, the Granger-Sims tests show no relationship. Yet one country has caused the output in another to change. One way to check whether positive or negative changes in one country's output are transmitted to those of another country is to examine the links among the absolute values of output deviations. The question is whether a shock, in either demand or supply, say, that affects domestic output, measured by the absolute value of the change in domestic output, is associated with a change in the absolute value of output abroad. Thus the same methodology as before serves, only now the units of observation are the absolute values of the deviations from trend levels of national outputs.

523

Real Output and the Gold Standard Years, 1830-1913

The results for both the gold standard years and the entire sample period are consistent with the earlier finding that there is no systematic evidence that changes in one country's output are related in a GrangerSims fashion to outputs in other countries (tables 11.A.1 and 11.A.2 of the Appendix). Another possible reason that no evidence emerges of significant bivariate relationships is that the true underlying relationship may be between groups of countries. The technique used here does not reject the possibility that blocs of nations, the Scandinavian countries, for example, are more relevant. for causal connections than each nation in the bloc individually.12 The lack of causal dependency of national outputs upon one another is an important issue in several contexts. The notion that one country was an engine of economic expansion or contraction for an Atlantic economy is dubious at the aggregate level. This is not to say that particular components of national output were not strongly related to one another. Investment in the United States, for example, might have been affected by investment in the United Kingdom. But at the level of aggregate outputs, the relationships do not appear to have been particularly strong. Money-demand studies such as those that underlie the monetary approach to the balance of payments typically assume the independence of real national incomes. At least on a pairwise basis, the assumption appears to be justified. Finally, whatever the many channels by which disturbances were transmitted from economy to economy, the sum of those disturbances (bilaterally) from one national output to another appears to have been weak during the gold standard period. Changes in a nation's output were determined by domestic and possibly international variables, but not in a simple leader-follower fashion. 13 The results of this section are largely negative, showing little indication of bilateral relationships among the real outputs .of various nations. Another approach is to ask whether real-output changes during the gold standard years correspond to current experience and theory. Section 11.2 explores the output-inflation tradeoff in the past and asks whether there is any evidence for the historical period of a "natural rate" of real-output growth. 11.2 Output, Inflation, and the Domestic Determination of Prices

In section 11.1 little evidence emerged of bivariate relationships among national outputs during the gold standard period. Section 11.2 examines the interconnectedness of real output from a different perspective. The perspective is a model of each economy based on a maintained hypothesis that both the price level and the level of output are endogenously determined by domestic conditions. The closed-economy model, developed

524

Stephen T. Easton

originally by Lucas (1973), assumes that the elasticity of the aggregate demand schedule is unity. The model is then recast (Arak 1977) to allow for the possibility of a nonunitary aggregate-demand elasticity. The presumption of the analysis is that very high estimated price elasticities of demand are consistent with the view that countries were small. The relevant price level in that case is an international price level exogenous to the home country. In the event, the evidence is mixed. Although high elasticities of demand are present-some are infinite-the estimates are not sufficiently precise to rule out a unitary elasticity in most cases. Since the Lucas model has been of great interest as a device for exposing and testing a rational-expectations approach to the Phillips curve /4 section 11.2.1 reports results of tests similar to those used by Lucas. Lucas's data were drawn from the post-World War II era, and it is at least of passing interest to see how the same tests fare with data from the gold standard era. Section 11.2.3 estimates the elasticity of the aggregate demand schedule and describes a model that allows the price level to become, in effect, an exogenous variable.15 Estimates of the elasticity of demand give some information on the degree to which countries were able to determine their own price levels. Although the evidence is mixed, as was apparent in the conference papers and the discussion from the floor, the speed with which nations adjusted to international prices is still an unsettled issue (McCloskey and Zecher 1976, and their paper in this volume). 11.2.1

Lucas's Model

The cyclical behavior of aggregate supply, Yct, is assumed to depend upon the discrepancy between the actual and expected price level, plus a lagged value of cyclical real output (where all variables are measured in natural logarithms) so that I6 (3)

Yct = ~[Pr - Et(Pt I It)] + ~Yc,t-l·

Using information about the average price level and the observed price in the local market, the aggregate-supply function can be written as: (4)

Yt = Ynt + e~(Pr - ~) + ~[Yt-l - Yn,t-l] ,

where Ynt refers to the secular level of real output common to all markets, Pt , the known mean level of prices, and 1'2

e=--(12 + 1'2 ' where 'T 2 is the variance of deviations of local prices from Pt and (12 is the variance of the overall price level. Assume that the demand for goods can be represented in a simple form:

525

Real Output and the Gold Standard Years, 1830-1913

Yt + Pr = X t ,

(5)

where X t refers to the (log of) nominal income. Assume that changes in X t are normally distributed and are independent with mean 8 and variance (1;. The reduced forms for Yet and 1i.Pt are:

(6)

Yet

(7)

1i.Pr = - ~ + (1 -

where

~

= -11'8

+ 11'1i.xt + AYe,t-l , 11' )1i.xt

+ 11'1i.xt - 1 - A1i.Ye,t-l ,

= the (exponential) trend growth rate of real output, 'IT =

11'

~ 1 + e'Y

or, substituting for

e,

72'Y

= ------.;...---(1 - 11')2(1; + 72 (1 + 'Y) .

From equations (6), (7), and the definition of 11', it follows that when (1; becomes large, 11' approaches zero, and demand shocks have little effect on real output, being increasingly absorbed as price-level changes. Thus, a prediction of the theory is that values of 11' should decline as the sample variance of nominal income changes (1i.x t ) increases. 11.2.2

The Evidence

Table 11.6 reports a summary of the country-by-country regression results. In the table the first two columns display values of 11' and Awhich are drawn from the regression results of equation (6), the reduced form for cyclical income. The R 2 associated with regression equation (6) is in column 3, and the R 2 associated with equation (7) is in column 4. Column 5 contains the variance of nominal-income changes for each of the countries and relevant time periods. The time periods have been constructed so that they cover interesting periods of a reasonable length. In each case the three decades 1881-1913 are distinguished from the preceding years. For most countries at least a few observations are available for years before the world gold standard era. The table gives the estimates for the full run of data and for subperiods. Two basic tests of the model are presented. The first test, whether the country-by-country regressions fit well, gives generally favorable evidence. The R 2 ,s of the different equations range from .33 to .94 with most of the values falling around .75-the same order of magnitude as those Lucas (1973) obtained for more recent data. A second encouraging finding is that the values of 11' and A with two exceptions fall within the interval zero-to-unity which is consistent with theoretical expectations.I7 Although certainly not conclusive, these features of the model are generally supportive of the approach. A less encouraging picture is presented in figure 11.1 where 11' is plotted

526

Stephen T. Easton

Table 11.6

Canada 1870-1913 1881-1913 Germany 1854-1913 1854-1879 1879-1913 Italy 1866-1883 1884-1894 1895-1913 1881-1913 Norway 1893-1913 Sweden 1863-1913 1863-1879 1881-1913 United Kingdom 1832-1913 1832-1859 1881-1913 United States 1834-1859 1881-1913

Summary Statistics

Coefficient on NominalIncome Change 7T (s.e.) (1)

Coefficient on Lagged Real Cyclical Output A 2 RYet (s.e.) (2) (3)

(4)

Variance of NominalIncome Percentage Changes (5)

.69 (.07) .73 (.09)

.91 (.07) .98 (.08)

.82

.33

.00864

.84

.33

.00758

.32 (.05) .36 (.10) .31 (.06)

.76 (.06) .83 (.18) .64 (.09)

.61

.74

.00498

.49

.74

.00749

.70

.74

.00328

.22 (.07) .32 (.15) .52 (.07) .46 (.05)

.59 (.16) .71 (.26) .97 (.06) .94 (.05)

.56

.83

.00603

.55

.68

.00356

.95

.79

.00541

.92

.78

.00569

.15 (.06)

1.01 (.14)

.75

.92

.00293

.51 (.06) .51 (.09) .54 (.07)

.82 (.07) .84 (.13) .74 (.10)

.79

.66

.00511

.79

.70

.00811

.79

.63

.00378

.19 (.03) .16 (.04) .23 (.08)

.82 (.06) .70 (.11) .86 (.10)

.73

.87

.00410

.70

.94

.00832

.71

.77

.00209

.48 (.08) .76 (.06)

.65 (.10) .96 (.06)

.76

.54

.00400

.92

.56

.00318

Rip

Sources: Real output as in table 11.1. Prices: Canada (Dick 1981); Germany, Italy, Norway, Sweden, United Kingdom before 1855 (Mitchell 1975, table 11); United Kingdom after 1855 (Feinstein 1972, table 61); United States (U.S. Bureau of Census 1960).

527

Real Output and the Gold Standard Years, 1830-1913

0";

against the variance of nominal income for the years of the gold standard, 1881 to 1913. If the model adequately characterizes the economy's response to nominal-income shocks, a negative relationship should be displayed. No such relationship emerges.I8 One possible explanation is relatively little variabililty in the nominal-income variances across the countries in the sample (table 11.6, col. 5), in comparison, say, to Lucas's study. There are no outliers like Argentina or Paraguay with an order-ofmagnitude dIfference in nominal-income variance from the rest of the sample. If one excludes these two observations from Lucas's sample, then for his results is obscure as the negative relationship between 11' and well. I9 Another way to view the data is to break the sample for each country into subperiods and examine the pattern of 11' and income variances. The advantage of this procedure is that the value of 'Y, the elasticity of supply, is more likely to be stable within a country than across countries. The disadvantage is the unavailability of many periods so that the result of the comparison must be informal rather than statistically rigorous. Figure 11.2 plots the results of the exercise. In each case the first observation plots 11' against the variance of nominal income in a period prior to the years 1881-1913. The second observation, the point to which the arrow is drawn, represents the combination of 11' and the nominalincome variance in the period 1881-1913. Splitting the data in this fashion means assuming some systematic difference between a world on the gold

0";

IT 1.0

us C

SWD

0.5

I G

UK 0.1 100 Fig. 11.1

500

1000

a;, a;

The relationship between 7T and 1953-67, and during the gold standard years, 1881-1913. = variance of nominal income x 10- 5• Source: Table 11.6.

528

Stephen T. Easton

n 1.0

us

C..--4 C

\ SIWD.. . . . .- -__ SWD US -----"1

0.5

UK UK 0.1 100

Fig. 11.2

1000

500

(1;

'I--

(1;

Values of 7r and before and after the gold standard years. = variance of nominal income x 10- 5. Source: Table 11.6.

standard and a world on various alternative monetary arrangements; some countries are formally on a bimetallic standard, some are floating, and some are on a gold standard. For countries on a gold standard throughout the period, as the United Kingdom was, splitting the data before and after 1881 should reveal no change. Figure 11.2 shows that in most cases the fall in the variance of nominal income during the gold standard period gave rise to an increase in 7T. The exceptions, Canada and Germany, are also the countries in table 11.6 for which the fits of the equations are the weakest. Italy is treated slightly differently than the other countries in figure 11.2 since the exchange rate was fixed only from 1884 to 1894. The first arrow indicates the change from pre-gOld standard years, 1863 to 1883, to the fixed-rate period, and the second arrow points to the whole 1881-1913 era. Although there are in Italy is consistent only a few observations, the behavior of 7T and with the changes in nominal-income variance observed during the subperiods. Unlike the general decline in nominal-income variance during the 1881-1913 decades, nominal-income variance in Italy increases before and after the period of the Italian gold standard decade. The two countries for which 7T and move in the "wrong" direction are Germany and Canada, but in both countries the change in 7T is rather small. Thus even this informal test yields mixed results. For most of the countries in

0';

0';

529

Real Output and the Gold Standard Years, 1830-1913

the sample, behavior is consistent with the predictions of the theory, but the pattern is not overwhelming. In sum, there is some evidence to support this form of the model, and some reasons why the evidence might be expected to be weak, but the major predictions are disappointing. However, at least one more element of the model deserves investigation-the assumption of a unitary elasticity of aggregate demand, embodied in the notion that nominal income or, more precisely, the change in nominal income, is the relevant exogenous variable. 11.2.3

International Interconnectedness

The major limitation of the Lucas model from an international perspective is that it ignores the possibility that countries are linked by international arbitrage (McCloskey and Zecher 1976). In a gold standard environment-one in which the exchange rate is rigid-domestic prices depend upon foreign prices, and if the country is small and open, the domestic price level is exogenous. In terms of the Lucas model, in equation (3)-the aggregate-supply equation-the domestic price level should be replaced by the foreign price level. Deviations of actual from expected foreign prices would drive the model, making it necessary to develop a price index for each country that reflected the relevant foreign prices. An alternative would be to recast the theory, using world nominal-income shocks as the relevant exogenous demand variable, and then predict world-price-Ievel changes and world-cyclical-output responses. To avoid such constructions based on tenuous data, I pursued the closed-economy version of the theory which assumes that there is exogeneity in domestic nominal-income shocks and that domestic prices are at least in some measure determined by domestic demand and supply. Since in section 11.1 no systematic causal relationship between nominal or real incomes in our several countries was evident, that findinggranted that only bilateral relationships were examined-serves as some justification for eschewing the "world-scope" approach. If domestic prices are rigidly linked to world prices, then equations (6) and (7) are irrelevant, since world prices are determined by world conditions and not simply by domestic nominal income. Let demand be characterized by Yt = - ~Pt + X t where ~ becomes very large. In that case neither equation (6) nor equation (7) is relevant, since 1T = e'Y/(~ + e'Y), which means equation (3) must include some index of foreign prices. This concern leads to a test of the model. Does the specification in equations (6) and (7) fit the data? In particular, is there evidence that changes in measured price levels for each country are set in international markets so that the demand schedule facing each country is highly elastic? If we enforce the restriction that ~ = 1, then the RZ's in table 11.6

530

Stephen T. Easton

provide a general test of goodness of fit. If we estimate ~, assuming an exact fit (i.e., the errors are exactly zero) for equations (6) and (7), we can test the restriction that ~ = 1 (Arak 1977) and also test whether the exact form of the model is appropriate (Lucas 1977). If ~ =F 1, then.li~ and Yct may be written as: (8)

IiPt

= ~o +

(9)

Yct

-

(1 - A)

Ut

+ --,

---Yct-l

~

e~

+~

_ AYct-l + -e~- U , e~

+~

t

where Ut is the error associated with nominal disturbances.2o The coefficients of Yct-l in the two equations allow us to construct an estimate of~, the price elasticity of demand for real output; regressing the residuals of equation (8) on the residuals of equation (9) gives an estimate of e~, the supply response to an unexpected change in prices; and the R 2 of the regression which regresses the residuals of equation (8) on the residuals of equation (9) allows us to test the assumption that the exact version of the model is appropriate since the errors should be proportional and the R 2 high. Table 11.7 presents the values of A and ~ and the approximate standard errors. The final column reports the R 2 of the residuals regression. The R 2 ,s of both equation (8) and (9) are uniformly low, as expected, and are not reported. The R 2 of regressing the residuals of equation (8) and on the residuals of equation (9), reported in the final column, are also uniformly low and sometimes negative. Recall that it is a homogenous regression. Table 11.7

Estimates of Price Elasticity of Demand for Real Output

(~)

R 2 of residuals

Canada 1869-1913 1869-1881 1881-1913 Germany 1853-1913 1853-1879 1879-1913

A (s.e)

(1- A)/~ (s.e)

.63 (.12) .52 (.29) .70 (.13)

-.04 (.08) .18 (.19) -.10 (.09)

-9.25 (18.26) 2.7 (2.38) -3.0 (2.37)

.58 (.10) .44 (.17) .71 (.11)

.00 (.16) -.40 (.27) .43 (.20)

**** ****

~

(s.e. )a

-1.65 (1.03) .67 (.62)

of eq. (8) on residuals of eq. (9)

.0 .04 .0

.01 -.02 .02

531

Real Output and the Gold Standard Years, 1830-1913

Table 11.7 (continued)

Italy 1862-1913 1866-1883 1884-1894 1895-1913 Norway 1892-1913 Sweden 1862-1913 1862-1879 1881-1913 United Kingdom 1832-1913 1832-1859 1832-1879 1881-1913 United States 1831-1859 187{}-1913 1881-1913

R 2 of residuals of eq. (8) on residuals of eq. (9)

A (s.e)

(1- A)/~ (s.e)

.87 (.08) .51 (.19) .61 (.31) .89 (.11)

-.01 (.11) .10 (.43) .06 (.46) -.32 (.11)

-13.0 (142.78) 5.0 (26.20) 6.4 (48.0) -.47

.96 (.14)

.30 (.43)

.13 (.47)

.01

.68 (.11) .65 (.18) .71 (.15)

.05 (.11) -.04 (.20) .22 (.15)

6.4 (13.91) -8.75 (43.52) 1.3 (.56)

.07

.79 (.07) .73 (.13) .79 (.24) .78 (.19)

.06 (.17) .31 (.49) .18 (.09) - .16 (.10)

3.5 (9.85) .87 (1.31) 1.2

.61 (.15) .67 (.11) .68 (.12)

.12 (.14) .07 (.07) -.05 (.06)

~

(s.e.)a

.09 .01 .10 .20

(-)

.09 .06

.04 .20 .07

-1.3

.0

3.25 (3.58) 4.7 (4.42) -6.4 (7.3)

.02 .16 .05

Sources: Same as for table 11.6.

aSE(O ""

C~ J[VARe ~ A) - ~ VAR(l- A)jV2.

The exact specification of the model is thus rejected by the data. I have not reported estimated values of the coefficient e~ that were statistically significant at the usual confidence level in only two of the twenty-five or so regressions.

532

Stephen T. Easton

In general, the values of ~ are quite erratic, and although they do not reject ~ = 1, they tend to be uncomfortably large, especially for the smaller countries, Canada and Sweden. 11.3 Conclusion

What is the upshot of this study? The Lucas model, a model that does not explicitly take account of international links among countries, is only broadly conformable to the evidence. The most powerful implicationthe relationship between the coefficient on nominal-income change (7r) and the variance of nominal-income percentage changes (a;)-is not supported by simple cross-sectional evidence. Higher and lower nominal-income-variance episodes, however, in individual countries do lend some support to that relationship. Our test of the restrictions that demand elasticity was unitary is not very powerful. Although the result is consistent with the restriction, it is also usually consistent with highelasticity estimates. A high elasticity would be expected if prices in different countries were closely linked to one another-if the world were filled with small open economies. Price behavior may conform to the rule of international arbitrage, since a measure of that behavior discussed above responds in a predictable fashion to shocks in nominal income. In section 11.1 little evidence emerged that either real or nominal output was causally related in a bivariate fashion among countries. The evidence suggests that nominal (and real) shocks may be generated in part domestically (or the evidence suggests no single international source). The attempt to capture international links in section 11.2 was only partly successful. There are two possible routes to follow in future work to explain real- or nominal-income behavior. One route is to construct a set of world prices and aggregate shocks that may impinge on individual countries, each of which would be small. The other route is to gather additional data-perhaps quarterly or semiannual observationsand use vector autoregressions to discover whether blocs of countries provide the links among nations. Finally, one general conclusion to be drawn from this study is that during the gold standard years the bilateral links among national incomes were not particularly strong in a causal sense and that at least some observed domestic-price and real-output behavior is explicable on the assumption that the price elasticity of aggregate demand is not infinite. Economies appear to have been less small than might have been anticipated, but the issue is still open.

533

Real Output and the Gold Standard Years, 1830-1913

Appendix

rr;

Figure 11.A.l contains a plot of values of 'IT and from both Lucas 1973 and table 11.6. Tables ll.A.l and 11.A.2 illustrate the results for some of the countries in the sample when the absolute value of output deviations of one country are assumed to Granger-cause the output deviations of another country. Table ll.A.l reports the F-value of the effects of the restric-

n 1.0

• •

• , •

o

o

••

0.5

,..

o

o

• • ••

o

o o

0.1 '------I.

---L.-

~/

-...L

500

100

••

I

1000

(1;, (1;

Fig. II.A.I

Values of 7T and 1953-67, and during the gold standard years, 1881-1913. = variance of nominal income x 10- 5. The black circles are drawn from Lucas 1973, tables 1 and 2; the white ones are drawn from fig. 11.1. Source: Lucas 1973, table 6.

Table II.A.l

Granger Causal Links between National Absolute Values of Output Deviations, 1881-1913: Values of the F -test

U.S. U.K. Germany Canada

Source: Same as table 11.3.

u.s.

U.K.

Germany

X 0.85

2.74 X 0.83 6.69

1.14 X

0.75

Canada 0.56 1.24 X

534

Stephen T. Easton

Table 11.A.2

U.S. U.K. Germany Canada Italy

Granger Causal Links between National Absolute Values of Output Deviations, Full Sample Years: Values of the F -test

U.S.

U.K.

Germany

Canada

X 1.68 0.89 0.16

0.19 X

0.62 1.10 X

0.78 1.95

4.69 1.43

Italy

1.66 0.24

X

1.07

X

Source: Same as table 11.4.

tions. A high F means that the row country Granger-causes the column country. In tables II.A.I and II.A.2, the only significant F, at the 95 percent confidence level, is for Canada Granger-causing the United Kingdom. Thus, as is the case when causation is based on the algebraic value of output deviations from trend, the absolute values of the deviations also reveal little binational linkage among countries. Although potentially a weaker restriction of Granger-causation between national outputs, the test based on the latter set nonetheless reveals no significant pattern.

Notes 1. For a summary of the gold standard years see Yeager 1966, pp. 251-65. A more detailed discussion may be found in Hawtrey 1935 and, for the United States and the United Kingdom, in Bordo 1981. 2. Both France and the United States were officially on a bimetallic standard-gold and silver-but large changes in available quantities made official ratios, approximately 15:1 in the United States and 15.5:1 in France, too great to be sustained; thus the United States was effectively on a gold standard for much of the period 1834-59. Since we were unable to unearth real output data for France, no more will be said about this interesting interaction of France with the rest of the world. 3. Countries joined the gold standard at different times and some dropped out for various periods. Russia and Japan both joined in 1897. Italy was a member of the gold standard "club" only between 1884 and 1894 (Fratianni and Spinelli, this volume), although Bloomfield (1959, p. 13) has dated the period as 1883-91. Austria-Hungary was never legally on gold, although Bloomfield suggests that its exchange rate was stable relative to gold from 1900 to 1914; Argentina left gold in 1885 and returned in 1900. Bulgaria left in 1899 only to return in 1906. Other countries dropped out at various times: Portugal (1890), Chile (1898), and Mexico (1910). Spain was not associated with the gold standard, and China, EI Salvador, and Honduras remained on a silver standard. 4. See McCloskey and Zecher 1976, pp. 357-85 in which they cite much of the standard literature associated with these questions. 5. Choudhri and Kochin 1980 have a good discussion of the different models available.

535

Real Output and the Gold Standard Years, 1830-1913

6. There is an extensive literature on "staple" models of growth and development. See, for example, North 1966, Chambers and Gordon 1966, Dales 1966, and Williamson 1980. 7. For a more detailed discussion of efforts to identify business cycles, see McCulloch 1975 and Neftci 1979. 8. The United Kingdom may still have been a hub for some countries. We have no evidence on the association of output between the United Kingdom, Australia, Argentina, and India, to name a few. It is important to stress that Thomas, in particular, is concerned with long swings whereas I examine year-to-year fluctuations. 9. One of the key problems of this methodology is to determine the appropriate length of the lags. In principle they may be infinite. In tables 1L3-1 L5 reported here, lags up to six periods, and, when relevant, leads of six periods are reported. An F-test on additional lags (and leads) indicated no significant effects beyond those reported in the tables. 10. To assess the degree to which the nations shown in table 1L3 are causally independent, assume that each causal interaction-each of the 56 elements of table 11.3--is independent, then at the 0.05 confidence level, 0.05 x 56 of the elements would be statistically significant by chance. We use the binomial distribution-approximated from the central-limit theorem by a normal distribution-to calculate the probability that the 6 significant elemerits actually found are due to chan~e. The null hypothesis is that the number of significant elements is 2.8( = 0.05 x 56); the alternative is that the observed number is greater than 2.8. Calculate the standard value : = 6/56 - 2.8/56 = L96.

"I

V

(.05) (.95) 56

The hypothesis that the observed pattern is random at the 0.05 confidence level is rejected when L96 is compared to a normal distribution. But there are two obvious problems with each procedure: only integer values of the number of significant relationships is possible, and the trials may not be independent. Allow 3 to be the "expected" number of random significant relationships, then = 1.84. Choose the confidence level consistent with 3 significant relationships, .054, then = L78. These calculations still give some evidence supporting the hypothesis that the pattern of observed causal links is nonrandom. There is a sort of "catch-22" quality to the experiment, however, since the calculations assume that the trials were independent. That assumption is surely at risk if income in country A causes income in country B and country C, and income in country B appears to cause income in country C as well. Consider in table 11.3 that Sweden causes both Italy and Denmark. But Denmark causes Italy too. The observed causation is very possibly spurious and surely reflects an interdependence. Eliminating the potentially spurious relationship and computing , which is equal to L35, means that the observed distribution of significant elements does not differ from that expected from a random sample at the 0.05 confidence level. (At the 0.10 confidence level there are 8 untainted interactions and again falls below the critical level. ) A vector autoregression to see whether Denmark explained any additional variation once Sweden was accounted for was negative, but there were relatively few degrees of freedom. The fact that loss of a single observation drives the observed pattern well into the range consistent with random behavior at the 0.05 level provides additional support for the claim that the pattern of table 11.3 is not significant. Carrying out the same procedure at the higher confidence levels 0.10 and 0.25 yields similar conclusions. Because it is so dependent on the independence of the trials, this form of "test" really gives only an upper bound on the number of causal links. The conclusion is that with any sort of simple correction for interdependence, table 1L3 does not support the hypothesis that real outputs of countries were firmly linked during the gold standard period. 11. The results of the Sims test are usually similar to those found by the Granger methods. I find no reason to prefer one test over another with these data although Feige and

536

Stephen T. Easton

Pearce 1979 report the Sims test to be very sensitive to the prefilter used, and Geweke, Meese, and Dent 1979 report other drawbacks. 12. Neftci 1979 reports experiments along these lines using contemporary data. With the small amount of historical data available, however, vector autoregressions quickly use up the limited number of degrees of freedom. 13. Table 11.3 was recast using nominal income instead of real income, consistent with the approach in section 11.2 that considers nominal income the exogenous impulse to which output and prices respond. Unfortunately, the patterns of causation (not reported here) were every bit as faint and erratic as those in table 11.3. 14. The natural-rate debate still burns hotly. See Phelps 1971, Barro and Fischer 1976, and Azariadis 1981 for recent discussions. 15. For some criticisms of Lucas's (1973) model, see Arak 1977 and Lucas 1977. 16. The notation corresponds to Lucas 1973 for easy reference, but for simplicity omits the market index Z. 17. Norway is an exception with a 1.01 value of A.. But there are observations for Norway only from 1892 to 1913, and the standard error of the estimate is .14. 18. This is true even if we ignore Canada, for which the output data are notoriously poor, and Norway, for which the data, based on fewer observations than the rest, fit least well. 19. The Appendix shows a plot of 1T and A. from Lucas's (1973) post-World War II estimates and from table 11.6 for 1881-1913. 20. If demand is Yt = - ~Pt + X t then nominal income is Yt + Pr so that Yt + Pt = (1 - ~)Pr + X t • Only if ~ = 1 is the exogenous shock ~t exactly equal to changes in nominal income. In the context of equations (8) and (9), ~t = 8 + f.Lt, where 8 is a constant, the rate of growth of nominal output.

References Arak, Marcelle. 1977. Some international evidence on output-inflation tradeoffs: Comment. American Economic Review 67 (Sept.): 728-30. Azariadis, Costas. 1981. A reexamination of natural rate theory. American Economic Review 71 (Dec.): 946-60. Barra, Robert J., and Stanley Fischer. 1976. Recent developments in monetary theory. Journal of Monetary Economics 2 (Apr.): 133-67. Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard. New York: Federal Reserve Bank of New York. Bordo, Michael David. 1981. The classical gold standard: Some lessons for today. Federal Reserve Bank of St. Louis Review 63 (May): 1-17. Chambers, E. J., and D. F. Gordon. 1966. Primary products and economic growth: An empirical measurement. Journal of Political Economy 74 (Aug.): 315-32. Choudhri, Ehsan, and Levis Kochin. 1980. International transmission of the business cycle: Spain and the gold standard countries in the Great Depression. Journal of Money, Credit, and Banking 12 (Nov.): 56574. Dales, John H. 1966. The protective tariff in Canada's development. Toronto: University of Toronto Press.

537

Real Output and the Gold Standard Years, 1830-1913

Dick, Trevor J. O. 1978. Output, prices, and real wages: The Canadian experience, 1870-1915. Mimeo. - - - . 1981. Canadian balance of payments, 1896-1913: Mechanisms of adjustment. Mimeo. Feige, Edgar L., and Douglas K. Pearce. 1979. The causal relationship between money and income: Some caveats for time series analysis. Review of Economics and Statistics 61 (Nov.): 521-33. Feinstein, C. H. 1972. National income, expenditure, and output of the United Kingdom, 1855-1965. Cambridge: Cambridge University Press. Firestone, O. J.1958. Canada's economic development, 1867-1953. Studies in Income and Wealth, series no. 7. London: Bowes and Bowes. Gallman, Robert. 1968. Estimates of national income. Mimeo. Geweke, John, Richard Meese, and Warren T. Dent. 1979. Comparing alternative tests of causality in temporal systems: Analytical results and experimental evidence. SSRI Workshop series no. 7928, University of Wisconsin. Granger, Clive W. J. 1969. Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37 (July): 424--38. Hawtrey, R. G. 1935. The gold standard in theory and practice. 5th ed. London: Longmans Green. Lucas, Robert E., Jr. 1973. Some international evidence on outputinflation tradeoffs. American Economic Review 63 (June): 326-34. - - - . 1977. Some international evidence on output-inflation tradeoffs: Reply. American Economic Review 67 (Sept.): 731. McCloskey, D. N., and J. R. Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance ofpayments, ed. J. Frenkel and H. G. Johnson. Toronto: University of Toronto Press. McCulloch, J. Huston. 1975. The Monte Carlo cycle in business activity. Economic Inquiry 13 (Sept.): 303-36. Mitchell, B. R. 1975. European historical statistics, 1750-1970. New York: Columbia University Press. Morgenstern, Oskar. 1959. International financial transactions and business cycles. Princeton: Princeton University Press. Neftci, Salih N. 1979. International transmission of business cycles: An empirical study of five economies. Mimeo. North, Douglass C. 1966. The economic growth of the United States, 1770-1860. New York: W. W. Norton and Co. Phelps, Edmund S. 1971. The "natural rate" controversy and economic history. In Inflation and the Canadian experience. Kingston, ant.: Queen's University. Sargent, Thomas J. 1979. Macroeconomic theory. New York: Academic Press.

538

Stephen T. Easton

Sims, Christopher A. 1972. Money, income, and causality. American Economic Review 62 (Sept.): 540-52. Thomas, Brinley. 1973. Migration and economic growth. 2d ed. Cambridge: Cambridge University Press. U.S. Bureau of Census. 1960. Historical statistics of the United States: Colonial times to 1957. Washington, D.C.: Government Printing Office. Williamson, Jeffrey. 1980. Greasing the wheels of sputtering engines: Midwestern grain and American growth. Explorations in Economic History 17 (no. 3): 189-217. Yeager, Leland B. 1966. International monetary relations: Theory, history, and policy. New York: Harper & Row.

Comment

Geoffrey E. Wood

Professor Easton's paper begins by telling us that the choice between alternative monetary standards should depend on which standard leads to the highest available level, or rate of growth, of output per head. That notion (apart from any minor qualifications that might be made to regarding output per head as the only measure of welfare) is surely correct. What can money, neutral and therefore affecting only the price level in the long run, contribute to the level or rate of growth of real output? In view of its neutrality, money's only contribution can be to increase the predictability of the future price level. The crucial test that should guide us in choosing between the gold standard and all other monetary standards is therefore a comparison of price-level predictability under different monetary systems. This test is not among the large number of tests whose results are reported in Professor Easton's paper. Its absence suggests an uncertainty of aim which led to failure to link economic analysis with an examination of the data. This failure, it will be argued, is the crucial one which produced the defects in Professor Easton's paper. The comments that follow are divided into four sections. First, the questions Professor Easton addresses are summarized. The answers he produces are then discussed. Third, some details in the paper are examined. The comments then conclude with a summary of the lessons of this paper. The Questions Professor Easton uses data drawn from a wide range of countries, over a long run of years, to address the following questions. Geoffrey E. Wood is Reader in Banking and International Finance at the City University, London. He is indebted for discussion of these comments to Charles Goodhart and Forrest Capie.

539

Real Output and the Gold Standard Years, 1830-1913

1. Is there any evidence of correlation, or causal influences, or both, between real-income movements in these countries? 2. Does whatever relationship existed between real-income movements in different countries before 1879 change after that year, in which the United States formally adopted the gold standard? 3. Is the output-inflation relationship hypothesized by Robert E. Lucas, and tested in his 1973 paper, found in the data of this period? 4. Is the price level "more appropriately characterized as endogenous or exogenous to most countries during the period" (p. 514)? Briefly, Easton's answers to the first two questions are "no," to the third a tentative "yes," and to the fourth, "in some countries the elasticity of aggregate demand is very large" (p. 514). The Answers in Detail Output Fluctuations

Easton's examination of the relationship among output fluctuations is best considered in two separate sections. First he looks at correlations among various countries of annual deviations of the log of actual GNP from the log of trend GNP. He finds that by this measure there is, during 1830-79, a good degree of association of output fluctuations within Europe, but no association of any significance between Europe and North America. After 1879 (when the United States resumed the gold standard), the United States becomes involved in this pattern of correlation, as to a lesser extent does Canada. Sometimes the correlations are negative and sometimes positive. This result was noted earlier by Brinley Thomas (1954), and is indeed what one would expect in a world where fluctuations in output about trend are produced by a mixture of demand shocks and supply shocks. The reason for this can be summarized very briefly. Consider first a demand expansion. An expansion in demand in one country resulting from, for example, a drop in the savings rate, would under fixed rates "spill over" into the balance of payments as imports rose and exports fell, thus increasing demand overseas.! So long as the short-run aggregate supply curve were not vertical, there would be a positive output change in both countries, and hence a positive correlation between output movements. We should not, however, expect a similar response to a supply fluctuation. Consider the example of an unexpected, favorable, supply shock in one country. The shock would raise output and thereby increase the demand for money. With a fixed exchange rate, funds are drawn in from overseas, thus producing a monetary contraction overseas (unanticipated because the supply shock was unanticipated). This squeeze will in turn,

540

Stephen T. Easton

because of its being unanticipated, produce a recession overseas. Hence if the income fluctuation originates in a supply shock, there will be a negative correlation between income movements in different countries under a fixed-exchange-rate regime. Therefore Easton's results, because they essentially replicate Brinley Thomas's earlier work, and on analytical grounds, should so far be no surprise. So much for correlation. Easton next proceeds to look for "causal" relationships, using Granger-Sims methods to identify timing patterns. He looks first at relationships among fluctuations in national income and then at relationships among the absolute values of these national-income fluctuations. He carries out these tests because significant simple correlations among real output changes in many countries provided no evidence of causal impetus. They do not by themselves support the notion that the United Kingdom was the hub around which at least some nations revolved. (P. 518) His tests reveal some bilateral relationships, but no strong, overwhelming connections. What do these negative findings indicate? A first and basic point is that his tests have absolutely nothing to do with the "engineof-growth" hypothesis. That hypothesis is concerned with nationalincome trends; Easton's tests examine deviations from trend. If the engine-of-growth hypothesis were the focus of these tests (as is implied in the above quotation), then it can only be said that the tests were not well chosen. Do the tests reveal anything? The answer is th~t they do pro~ide a little information. The first half of this section, which mixes together positiveand negative-income correlations and looks for timing relationships between them, is totally pointless. Since we knew before carrying out the tests that the sample period contained a mixture of positive and negative correlations between income fluctuations, we could have anticipated that these would roughly offset each other, thus concealing whatever relationship might have existed. Looking at absolute values, as is done in the second half of this "causality" section, avoids this problem. By finding no clear lead-lag relationship among these fluctuations, Easton shows that cycles in one country did not, by and large, have much of a systematic timing relationship with cycles in another. That information is useful. It must be stressed, however, that these tests have revealed nothing about causality in the sense that A causes B, meaning that the occurrence of A is necessary and sufficient for the occurrence of B. There are very few occasions in economics when the isolation of a timing relationship, or failure to isolate such a relationship, can confirm or refute a hypothesis. The occasion on which Easton has used these tests is not one of them.2

541

Real Output and the Gold Standard Years, 1830-1913

The Lucas Hypothesis

It is useful first to quote from Lucas (1973). The positive association of price changes and output arises because suppliers misinterpret general price movements for relative price changes. It follows from this view, first, that changes in average inflation rates will not increase average output, and secondly, that the higher the variance in average prices, the less "favorable" will be that observed trade off. Before this idea can be tested by cross-country comparisons as Lucas did (and Easton does),3 there must be a good amount of inflation variability between countries. Is that likely to be found under the gold standard? Here the arguments of McCloskey and Zecher, advanced both at this conference (chap. 2) and in an earlier paper (1976), might become important. They argue that commodity arbitrage was an important link between countries under the gold standard. If that argument is correct, it is clearly impossible for countries in that period to differ significantly in either their inflation rates or in their inflation-rate variability. It is not, however, necessary for the purposes of examining Professor Easton's paper to decide whether McCloskey and Zecher are correct. The price indexes used by Easton are dominated by internationally traded goods in which there was, certainly after the telegraph revolution of the early 1870s, quite indisputably very close arbitrage. The behavior of these price indexes could not diverge substantially across countries. Hence whether it would be possible to test the Lucas hypothesis on this period's data by use of some other price indexes remains an open question. What is clear, though, is that given the indexes used by Easton, it is impossible for the gold standard years after 1870 to give confirmation of the Lucas hypothesis. Even if that hypothesis described the world perfectly, the world could by its nature not generate data on which the hypothesis could be tested. In view of this, what do Professor Easton's findings tell us? He found some modest confirmation of the Lucas hypothesis before 1879, but essentially none after. This is encouraging for the hypothesis. Before 1879 prices were tied together less closely than they were afterwards; before 1879 not all the countries in his sample were on the gold standard, and because of poorer communications, arbitrage in traded goods occurred more slowly than it did after that date. His findings can therefore be described as being consistent with the Lucas hypothesis, although it cannot be claimed as confirmation of it.4 Easton's informal work, in which he uses .graphical methods and (very sensibly) breaks the period into before and after the gold standard, provides some support for this view. His tests for the price elasticity of demand for each country's output

542

Stephen T. Easton

are, as he admits, so inconclusive as to be consistent with almost any prior expectation. Points of Detail Three details of Professor Easton's paper require comment. First, the choice of countries can best be described as eclectic. It is curious to include the Scandinavian countries and exclude Belgium, the Nether.lands, France, Argentina, India, and Australia. These were important countries in the system-a fact surely outweighed by the relatively poor data available. In particular, Australia was a most important member of the periphery. It is just not good enough to write (footnote 8), "We have no evidence on the association of output between the United Kingdom, Australia, Argentina, and India, to name a few." The data to seek that evidence do exist; the omission of, certainly, Australia and Argentina is of sufficient importance that it should have been either rectified or justified. Second, his rather harsh strictures on Morgenstern (1959) and Thomas (1954) are totally unjustified. Morgenstern, he writes, did not "abstract from the underlying growth rate"-but that is surely sensible if one is interested in relations among economies over a long period of years. With regard to Thomas, the assertion that it is difficult to accept inferences about the relationships among cycles because Thomas looks at cycles of varying amplitude and duration is certainly not self-evidently true-but it is presented as if it were. Third, some of the results he obtained are (as in some cases he notes himself) distinctly odd. For example, from 1879 Canada's income fluctuations appear associated with those of the United States, and those of the United States are associated with those of the United Kingdom. But no association is found between Canadian and United Kingdom income fluctuations. Turning to patterns of causality, there are ·some most curious findings. Germany "causes" U.K. fluctuations, Denmark and Sweden "cause" Italian fluctuations, but Norwegian fluctuations, although correlated with those in Denmark and Sweden, do not appear to join fluctuations in those two countries in "causing" Italian fluctuations. These findings may result from the omission of some third variable from the bivariate causality tests. But it is much more likely that they are the chance result of extensive manipulation of poor-quality data. The Lessons of the Paper What does this paper tell us? The author himself does not seem sure; if we read the concluding section, we find proposals for future work, but no clear statement of what has been achieved. In fact, four lessons emerge very clearly from the paper. First, it confirms that a fairly predictable monetary system imposes no particular systematic behavior pattern on

543

Real Output and the Gold Standard Years, 1830-1913

the real economy. That confirmation is useful. Second, it demonstrates that in the gold standard period looking at individual episodes is helpful, while examining the period as a whole on balance conceals information. Third, the paper is a useful reminder of the dangers of taking at face value the results of the extensive and elaborate statistical manipulation of not particularly reliable data. Fourth, and perhaps most important, it shows very clearly just how necessary it is to consider analytically the question being addressed before rushing at the data with a battery of statistical techniques.

Notes 1. This alleviation of demand pressure was known in the United Kingdom during the Bretton Woods era as the "problem" of the balance of payments. For discussion of the cyclical behavior of the United Kingdom's balance of payments in this period, see Williamson and Wood 1976. 2. Two further points should be noted on the use of Granger-Sims tests in this paper. First, although two references are cited that suggest Granger's methods are superior, a Sims test is used in one instance without an indication of why it was chosen. Second, "To supplement the Granger causal relationships, table 11.5 uses a Sims test to examine the full range of data available for each country" (p. 520). Now one thing Granger-Sims methods have revealed quite unambiguously is the importance of the exchange-rate regime for relationships between countries. See e.g., Mills and Wood 1978. Thus simply flinging together data from different exchange-rate regimes could certainly not help reveal any "causal" relationships; indeed, it is more likely to conceal them. 3. It should be pointed out that Easton takes no account of the comments of Neil Wallace on Lucas (1973), which Lucas acknowledged in a note (1976) to be of some importance to the interpretation of his paper. As Easton's paper is affected by a different and prior flaw, however, this issue is not pursued here. 4. Easton points to the lack of cross-country inflation variability as the reason for the failure of the test, but does not seem to grasp why it occurred or what it implies.

References Lucas, Robert E. 1973. Some international evidence o'n output-inflation tradeoffs. American Economic Review 63 (June): 326-34. - - - . 1976. Errata. American Economic Review 66 (Dec.): 985. McCloskey, Donald N., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance ofpayments , ed. J. Frenkel and H. G. Johnson. Toronto: University of ' Toronto Press. Mills, T. C., and Geoffrey E.Wood. 1978. Money-income relationships and the exchange rate regime. Federal Reserve Bank of St. Louis Review 60 (Aug.): 22-27. Morgenstern, Oskar. 1959. International financial transactions and business cycles. Princeton: Princeton University Press.

544

Stephen T. Easton

Thomas, Brinley. 1954. Migration and economic growth. Cambridge: Cambridge University Press. Williamson, John, and Geoffrey E. Wood. 1976. The British inflation: Indigenous or imported? American Economic Review (Sept.): 520-31.

Discussion of Huffman-Lothian and Easton Papers MELTZER commented on the view that seemed to prevail in the preceding sessions to the effect that institutions didn't matter: Sweden's growth didn't matter; it didn't matter if Sweden was on the gold standard or not; Italy not being on the gold standard didn't matter; nothing seemed to matter. This session's papers and the comments on them raise an important issue because they ask the question whether in fact and in which way did institutions matter. Suppose the fluctuations are entirely random, random mixtures of supply and demand shocks. What is random and what is not random with respect to prices is not independent of the perceptions that people have, what they expect prices are going to do, and how long these movements are going to persist; surely that set of beliefs depends very much on the nature of the monetary system. The proposition may be a difficult one to test, but it certainly is not impossible to test-that the monetary rule, the fiscal rule, and the agreements that people have made among themselves and between countries have something to do with the way in which cycles, price movements, exchange-rate movements, monetary movements occur. The fact that we don't find in patterns of shocks exactly what might be expected from a rigid interpretation of these arrangements should not immediately lead to the view that the arrangements don't matter at all-that what the rule is or what kind of monetary system we have doesn't matter. It would seem a little early to dismiss the importance of rules, one reason being that the nature of what is considered a permanent deviation and what is considered simply transitory white noise might be very different. For example, compare the Irish potato famine and its effect on prices in England, where the price of wheat apparently doubled or trebled during the course ~f a very short period of time and then fell by 50 percent, to the uncertainty generated in the modern economy by a similar rise in the price of oil. In the earlier episode, under the prevailing institutions of Britain, no one expected price controls to be put on wheat, no one expected any of the kinds of policies that would be very likely to be high on the list of what one might expect when similar shocks take place today. So I hope that the monetary and fiscal rule will be considered to

545

Real Output and the Gold Standard Years, 1830-1913

have more importance than was being assigned to it earlier in the conference. BARRO, commenting on Easton's paper, pointed out an error in the version of the Lucas model that was utilized. In point of fact, the output and inflation equations in the model are not independent. The inflation equation is just the first difference of the output equation. Thus, there is no point to estimating separately the output equation and the inflation equation. DORNBUSCH indicated that in reading the Lothian-Huffman paper, he was unable to link the theoretical and empirical sections of the essay. MCCLOSKEY started with a general comment. There comes a point, he suggested, when the results of fitting hyperplanes to time-series data are so strange that one must question the advisability of the exercise. In the Huffman-Lothian paper, for example, there is a one-year lag in the adjustment of interest rates to external shocks. One can imagine certain circumstances in which such an adjustment lag might exist, but the foregone profit opportunities seem tremendous. Thus, this result must be treated skeptically. McCloskey went on to quarrel with the way the word "significance" was used in these papers as a basis for making inferences. In fact, all the "significance" level indicates is the probability of type-one error. LOTHIAN accepted McCloskey's first point. Their paper was not an attempt to present an explicit test of interest arbitrage. The authors merely wanted to have interest rates in the model. THOMAS commented on Easton's paper. In his introduction, Easton states that there is little evidence of an Atlantic economy as far as real income movements are concerned. That statement is based on an analysis of the United States, Canada, the United Kingdom, Germany, Italy, Denmark, Norway, and Sweden. Yet the term "Atlantic economy" was chosen to represent the United Kingdom and a periphery of developing countries, namely, countries of recent settlement overseas-not all of them in or around the Atlantic ocean. It is quite clear in the literature that the Atlantic economy includes countries such as Canada, the United States, Argentina, Australia, and New Zealand. The rather eccentric set of countries considered in Easton's paper could not possibly warrant a conclusion about the existence of an Atlantic economy. ABRAMOVITZ elaborated upon Thomas's point. There is another sense in which the selection of countries and the uniform treatment of the whole period is unrepresentative of the notion of the Atlantic economy. Consider the question of migration from European countries to the United States. The links between the United States and the various European countries were changing over time. Immigration occurred as European nations entered their periods of industrialization and the movement from their farms began. For example, before the U.S. Civil

546

Stephen T. Easton

War German immigration began, followed after the war by immigration from Scandinavia, then Italy, and so on. Changes in the linkages among countries must be taken into account. EASTON commented on his usage of the phrase "Atlantic economy." He argued that all the countries he dealt with were closely related to one another. His footnote 8 carefully distinguishes his hypothesis from the hypothesis with which Brinley Thomas was concerned. Thomas is concerned with the long swing, whereas Easton is looking at year-to-year fluctuations. While the United Kingdom may still have been an economic hub for some countries, there is no evidence of any association of output between the United Kingdom, Australia, Argentina, and India. LOTHIAN acknowledge the problematic nature of the real-income estimates used in the Huffman-Lothian paper. Regarding the use of vector autoregressions, he acknowledged that the authors had reported some peculiar correlations. However, they had been careful not to make strong assertions on the basis of such results. Lothian also addressed one of Dornbusch's comments regarding the paper's emphasis on monetary variables. The authors had begun their inquiry with the presumption that monetary shocks were important but that real shocks might have mattered also. Dornbusch and Frenkel may be right that a harvest failure was the cause of a particular decrease in the gold stock. Yet the question remains: Is there then feedback from the endogenous changes in money to the real side of the economy? Are major monetary fluctuations associated with subsequent real-income fluctuations? Granger's and Sims's methodology provides a technique for distinguishing the influence of real and monetary variables. HUFFMAN commented that note 11 of the paper describes at some length the problems of interpreting Granger tests. It is well known that coefficients on successive lag values of a given variable will oscillate in sign. Huffman and Lothian report the results of a wide range of tests with the aim of providing different pictures of the relationships among variables. Not all of the pictures necessarily lead to the same conclusions. Huffman suggested that one's priors make a difference in how one interprets the results. However, in Huffman's view, there are plausible interpretations for most of the results the authors present.

12

Canada without a Central Bank: Operation of the Price-Specie-Flow Mechanism, 1872-1913 Georg Rich

12.1

Viner's Analysis of the Price-Specie-Flow Mechanism

Viner's (1924) celebrated study of the Canadian balance of payments before 1914 represents one of the first attempts to verify empirically the classical adjustment mechanism under the gold standard, as described by , the venerable price-specie-flow (PSP) model. In Viner's opinion, the Canadian experience between 1900 and 1913 provides an almost perfect opportunity for studying the' operation of the classical adjustment mechanism. Towards the end of the nineteenth century, Ca,nada experienced a remarkable surge in economic growth, triggered by a strong expansion in exports of mineral and agricultural products. Around 1903, the economic boom also led to an enormous acceleration of capital imports from Great Britain and the United States. Viner felt that the massive capital inflow created conditions highly favorable to an empirical test of the PSP model. On the basis of Canadian experience, it was possible to investigate how the capital inflow had been translated into a real transfer of goods and services and how efficiently the classical adjustment mechanism had operated. After a laborious investigation of the relationship observed between the Canadian current-account balance, prices, the money stock, and the capital inflow Viner concluded that Georg Rich is director of the Swiss National Bank, Zurich. The author would like to thank Keith Acheson, Andre Burgstaller, Derek Chisholm, Jack Galbraith, Ed Neufeld, Gilles Paquet, Soo Bin Park, Ron Shearer, Anna Schwartz, Alexander Swoboda, and Tom Rymes for their very helpful suggestions and comments on this project. The author would also like to thank the Canada Council for the award of two research grants to support this study. Moreover, he is indebted to Carleton University for financial support in the form of a GR-6.

547

548

Georg Rich

Canadian borrowings obtained transfer into Canada smoothly and without noticeable friction in the form of a net commodity and service import surplus, as the result of relative price changes (and shifts in demands) which were of the character indicated as to be expected by the older writers. (Viner 1937, p. 413) In order to summarize Viner's account of the adjustment mechanism, suppose that there occurred a once-and-for-all increase in the Canadian foreign debt. Moreover, assume that the current-account balance was zero initially. According to Viner, the capital flow led to an increase in the Canadian money stock as the Canadian borrowers converted their foreign-exchange receipts into domestic money and deposited these funds with the chartered banks, that is, Canada's commercial banks. As a result of the capital inflow, the chartered banks acquired additional reserves. However, despite the increase in their reserves, the banks did not augment their lending to Canadian residents; instead, they allowed their reserves to rise relative to their liabilities. The increase in bank-reserve ratios implied that the capital inflow was associated only with a primary round of domestic money creation-a direct result of foreign borrowing-but not with a secondary round, due to an expansion of the banks' Canadian loans (Viner 1924, chap. 8).1 As the Canadian money stock increased, Canadian demand for goods and services also rose. Since the prices of Canadian imports were largely set abroad, the increase in demand elicited primarily a rise in the relative prices of nontraded goods. The relative prices of exportable goods also rose, but less than those of their nontraded counterparts. These relative price changes were responsible for a shift in demand from nontraded to imported and exportable goods. Thus the current account tended to deteriorate in response to the capital inflow. Balance-of-payments equilibrium was restored when the cumulative value of the ensuing current-account deficits exactly matched the increase in the Canadian foreign debt. As a result of these deficits, the initial change in the Canadian money stock and relative prices was reversed. Since the money stock and prices only changed temporarily, the current-account balance in the new equilibrium was once again zero (1924, chaps. 9-11)? In one respect, Viner found that the Canadian evidence did not conform to the classical PSF analysis. The students of the pre-1914 Canadian financial system (Johnson 1910, pp. 49-50; Viner 1924; Beckhart 1929, pp. 416-17, 430; Shearer 1965, p. 331) generally agree that the chartered banks held their reserves not only in the form of cash, but also in the form of foreign short-term assets-in particular, call loans extended in New York and London, as well as deposits with foreign banks. These foreign short-term assets, which I shalI'call secondary reserves, could be readily converted into gold if the chartered banks faced an unexpected drain of the precious metal. In Viner's opinion, the secondary reserves played an

549

Canada without a Central Bank, 1872-1913

important role in the adjustment mec~anism. In the initial phase of the adjustment process the capital flow was not accompanied by an inflow of monetary gold, as the proponents of the PSF doctrine would have argued. Instead, the chartered banks augmented their secondary reserves. According to Viner, monetary gold was imported only as a result of the subsequent increase in the Canadian money stock since the chartered banks strove to maintain a stable ratio between their cash reserves and their liabilities. Thus flows of monetary gold across the Canadian border were not directly related to the capital inflow, but were a consequence of the change in the Canadian money stock. Viner's conclusions imply that the pre-1914 Canadian monetary system resembled a goldexchange rather than a pure gold standard. However, -in his view, this did not render the PSF mechanism inoperative since "fluctuations in the [secondary] reserves played the same role in the Canadian mechanism as that assigned to gold movements in the classical doctrine" (1937, p. 414). Viner's account of the Canadian adjustment mechanism did not go unchallenged and elicited a large number of critical comments. Two shortcomings of his analysis are especially noteworthy. First, his method of verifying the PSF model is valid only if capital flows can be regarded as a truly exogenous variable. Viner's verdict as to the speed and smoothness of the adjustment mechanism depends crucially on a close positive correlation he uncovered between the current-account deficit and capital inflows. In the presence of endogenous capital flows, it is conceivable that the correlation was merely the consequence of common factors impinging on the two variables. Therefore, the positive correlation does not necessarily indicate that the current account would have adjusted quickly if an exogneous capital inflow had occurred. A number of authors treating capital flows as an endogenous variable (Carr 1931; Meier 1953; Ingram 1957; Stovel 1959; Borts 1964; Cairncross 1968) have cast doubt on Viner's conclusions and have demonstrated convincingly that the evidence is consistent with alternative interpretations of the adjustment mechanism. For example, both the acceleration of the capital inflow and the deterioration of the current-account balance could have been explained by the shift to rapid economic growth observed around the turn of the century. However, these studies do not refute the Viner analysis; they merely suggest that Viner failed to furnish sufficient empirical support for the PSF model. Viner himself later on admitted that the evidence was consistent with alternative interpretations.3 Second, Viner's analysis of the link between the Canadian money stock and the balance of payments leaves much to be desired. A number of his critics have called into question his conclusion as to the unimportance of secondary money creation in the adjustment mechanism.4 In reply to his critics, Viner conceded that "primary and secondary expansion of means

550

Georg Rich

of payment both contributed to the creation of a situation in which necessary import surpluses could develop" (1937, p. 429). However, he did not completely change his mind in this regard, for he continued to insist that primary money creation had played the dominant role in the adjustment mechanism (p. 431). If we consider the procedure Viner adopted for analyzing money creation by the chartered banks, his conclusions are hardly surprising. Interestingly enough, in his 1924 study he did not examine the relationship between the aggregate liabilities to Canadian nonbanks and the reserves of the chartered banks, but between a series that he called foreign-loan deposits and reserves. He defined foreign-loan deposits as the difference between the banks' aggregate liabilities and loans to Canadian nonbank residents (1924, p. 187).5 As Goodhart (1969, pp. 148-51) has convincingly demonstrated, the Viner procedure is largely tautological since it effectively eliminates from the data on aggregate bank liabilities much of the variation due to secondary money creation. In his 1937 study Viner evidently sensed the inadequacies of his procedure, for he decided to compare bank reserves with aggregate bank liabilities, as well as with foreign loan deposits. Although even a cursory glance at the data suggests that aggregate liabilities and reserves were not closely correlated, Viner stuck firmly to his conclusions as to the speed and efficiency of the Canadian PSF mechanism. In his opinion, the adjustment mechanism had operated efficiently despite the existence in Canada of a fractional reserve banking system. However, it is doubtful whether Viner correctly interpreted the available evidence. If he had carefully scrutinized the data, he would have noticed that over the period 1900-1913, out of thirteen pairs of annual changes in aggregate .liabilities and reserves, five exhibited opposite signs.6 Thus it does not appear that the link between the capital inflow, bank reserves, and the money stock was as close as Viner suggested. Since Viner's conclusions are not entirely convincing, this paper reexamines the Canadian evidence on the operation of the PSF mechanism. Specifically, the objectives of the paper are twofold. First, the link between the Canadian balance-of-payments surplus (or monetary flows in the balance of payments) and the Canadian money stock is analyzed. A study of this link is hampered by the fact that the existing data on monetary flows are marred by serious omissions and inconsistent reporting. In an effort to improve the quality of pre-1914 Canadian balance-of-payments data, I reestimated monetary flows on the basis of unpublished and hitherto unused published evidence available for the major chartered banks. For a discussion of the estimation procedure, the reader is referred to Rich 1983. I show that the evidence contradicts Viner's account of the way the PSF mechanism operated in Canada. In the long run, the money stock was determined chiefly by the

551

Canada without a Central Bank, 1872-1913

balance of payments, as suggested by the classical PSF-doctrine. However, in the short run, the two magnitudes were but loosely related. Over the business cycle, in particular, the money stock was negatively correlated with the balance-of-payments surplus. The money stock displayed a distinctive procyclical pattern, while an inverse relationship existed between the surplus and the business cycle. Thus, contrary to Viner's view, the PSF mechanism did not work well in the short run. Second, I attempt to demonstrate that the failure of the PSF mechanism to operate in the short run was an important cause of cyclical instability in economic activity under the gold standard. The defects in the PSF mechanism implied that a cyclical change in the demand for goods and services could be accommodated, at least to some extent, by procyclical movements in the money stock, despite the severe constraints imposed by the gold standard on the ability of central and commercial banks to create money. The PSF mechanism acted as an effective stabilizer in the longer run, but it did not work quickly enough to prevent destabilizing movements in the money stock. The reasons that various defects in the PSF mechanism could generate procyclical movements in the money stock are discussed in section 12.2 of the paper. In section 12.3 the relationship between the Canadian money stock and the balance-of-payments is examined. The empirical work is based on the period 1872-1913 for which adequate data are available, rather than the much shorter period underlying Viner's study. The principal conclusion of section 12.3 is that changes in the reserve ratios of the chartered banks were an important source of the procyclical movements in the money stock. Section 12.4 demonstrates how bank-reserve management was responsible for the failure of the PSF mechanism to operate in the short run and thus contributed to cyclical instability in Canadian economic activity. 12.2 Procyclical Movements in the Money Stock and Defects in the Price-Specie-Flow Mechanism If the PSF mechanism had operated without any flaws, it is likely that the gold standard would have acted as an effective automatic stabilizer of cyclical fluctuations in economic activity. In order to illuminate the stabilizing role of the PSF mechanism, I first assume that cyclical disturbances were confined to Canada and then move on to the more realistic case in which business cycles were closely synchronized among the various gold standard countries. Suppose that there was a cyclical surge in Canadian economic activity, while foreign economic activity remained unchanged. Moreover, assume that the current-account balance was zero initially. If we abstract from international capital flows, the cyclical upswing was bound to generate a

552

Georg Rich

balance-of-payments deficit. Since, by assumption, the cyclical disturbance only affected Canada, foreign demand for Canadian exports was liable to remain unchanged, while the cyclical increase in Canadian economic activity induced a rise in imports. If the adjustment mechanism had worked along the lines postulated by Viner, the adverse balance of payments would have caused secondary reserves of the chartered banks, the money stock, and the stock of monetary gold to drop. The adjustment in the money stock would have acted as an automatic stabilizer designed to dampen the cyclical surge in economic activity. The PSF mechanism would not have played a stabilizing role if the cyclical upswing in Canada had triggered an increase in domestic interest rates and capital imports. In that case the balance of payments would have shown a surplus despite the deterioration of the current account. Thus, the cyclical increase in economic activity could have been accommodated by a rise in the money stock and an outflow of monetary gold. The possibility of capital flows playing a destabilizing role was first recognized by Taussig (1927, pp. 207-9). In what follows, I attribute procyclical movements in the money stock to a Taussig effect if they were caused by destabilizing capital flows? In practice, cyclical disturbances were not confined to Canada, but affected several or all the gold standard countries. In a study ofpre-1914 business cycles, Morgenstern (1959, chap. 2) uncovered a high degree of correlation between cyclical activity in Britain, France, Germany, and the United States. Moreover, a number of authors have shown that the Canadian economy was highly sensitive to cyclical fluctuations in U.S. economic activity. Prior to 1914, there was also a good correspondence between Canadian and British cycles, but the relationship was less close than with fluctuations in the United States (Chambers 1964; Hay 1966; Bonomo and Tanner 1972). In general the parallelism of cyclical movements was very close within North America and Europe, but less so between the two continents. However, North American and European cycles were not out of phase in any fundamental sense. The turning points of major cycles coincided closely, but North America witnessed a number of minor cycles that were not transmitted to Europe. The high degree of synchronization among national busines cycles implies that the gold standard would have acted as an automatic stabilizer, as long as each country was prepared to maintain a rigid link between its money stock and reserve of monetary gold. Since the world stock of monetary gold was unlikely to vary procyclically ,8 it would have been impossible for all countries to import simultaneously additional monetary gold in order to accommodate a cyclical surge in economic activity through an expansion in their money stocks. Balance-ofpayments surpluses or deficits would not have altered the world stock of monetary gold; they would merely have redistributed that stock among

553

Canada without a Central Bank, 1872-1913

the various countries. The money stocks could have moved procyclically in some countries, but only at the expense of countercylical fluctuations in the rest of the world. Thus the gold standard would have played a stabilizing role in the sense that it would have served as a prophylactic against procyclical movements in the money stock. The available evidence for the pre-1914 period suggests that the prophylactic function of the gold standard left much to be desired . .Although the cyclical pattern of the money stock has not been examined for all gold-standard countries, it is safe to argue that procyclical movements tended to dominate the scene. Existing research indicates that a variant of the Taussig effect was responsible for procyclical movements in the British money stock (Beach 1935; Ford 1962, chap. 3; Goodhart 1972, pp. 205-7; McCloskey and Zecher 1976). A persistent procyclical pattern has also been shown to exist for the U.S. and Canadian money stocks (Friedman and Schwartz 1963; Hay 1967). Similarly, it appears that the German and French money stocks were positively correlated with their respective business cycles (Rich 1983, chap. 8). The pervasive procyclical pattern of the money stock documented for a variety of gold standard countries lends support to the view that the PSF mechanism suffered from serious defects which loosened the links between the various national money stocks and the respective reserves of monetary gold. At least three factors explain why the gold standard did not forestall procyclical movements in the money stock. First, monetary authorities frequently did not observe the gold standard rules of the game (see Bloomfield 1959). Even before 1914, governments or central banks were empowered to issue notes under carefully specified conditions. Compliance with the rules of the game implied that the monetary authorities did not attempt to offset the impact of international gold flows on the supply of their notes. If they failed to adhere to the rules of the game, they effectively detached the money stock from international gold flows. Two other sources of destabilizing movements in the money stock were, respectively, countercyclical changes in the reserve ratios of the commercial banks and cyclical shifts in the composition of that stock as between notes issued by the monetary authorities and liabilities of the commercial banks. The problem arising from cyclical movements in commercial-bank reserve ratios, in particular, was extensively discussed in the older literature on the gold standard (Hawtrey 1928, 1947; Taussig 1927, pp. 200-203; Beach 1935, chap. 2). As far as Canada is concerned, the existing literature does not shed any light on the reasons for the procyclical pattern of the pre-1914 money stock. However, the above analysis points to two possible explanations. The first one is that the Canadian PSF mechanism worked without any flaws, but defects afflicted the mechanisms of other countries. In this

554

Georg Rich

event, procyclical movements in the balance-of-payments surplus would have been responsible for the observed cyclical pattern of the money stock. Alternatively, defects in the Canadian mechanism itself might have been at the root of the problem. In the following section, the sources of the procyclical movements in the Canadian money stock are traced, but before proceeding with this analysis we can dispense with one possible source. In pre-1914 Canada, destabilizing behavior on the part of the monetary authorities was not a significant cause of cyclical changes in the money stock. Although a central bank was not established until 1935, the government was entitled to issue notes, circulating under the name of Dominion notes. These notes served both as media of exchange for the nonbank public and as cash reserves for the chartered banks. They were convertible into gold at a fixed exchange rate and subject to a minimum gold-reserve requirement. However, while the government was obliged to maintain a minimum gold reserve, Canadian legislation did not impose minimum reserve requirements on the chartered banks. Nonetheless, the banks held ample cash reserves in the form of gold, subsidiary coin, and Dominion notes in order to safeguard the convertibility of their liabilities. A large fraction of the Dominion notes was backed by gold in government vaults, but occasionally notes were issued on an uncovered basis. During the pre-1914 period, changes in uncovered Dominion notes, though quite important prior to 1886, were not a significant source of cyclical variation in the Canadian money stock. Aside from a temporary issue of uncovered Dominion notes at the end of 1907, the Canadian government from 1886 onwards was exemplary in its adherence to the rules of the game.9 12.3 Cyclical Movements in the Canadian Balance of Payments, the Monetary Base, and the Money Stock 12.3.1

Balance of Payments and Monetary Base

Table 12.1 presents data on the three major components of the Canadian balance of payments . The current account covers merchandise- and non-merchandise-trade flows, excluding net interest and dividend receipts for which the available data are extremely unreliable. The overall surplus or net monetary inflows equal the first differences in Canada's stock of international monetary assets, embracing monetary gold in the hands of the government, as well as monetary gold and secondary reserves held by the chartered banks. Data on monetary gold in the hands of private nonbank residents are not available, but it is generally agreed that in Canada holdings of gold coin outside the government and the chartered banks were negligible, at least prior to 1914. The difference

555

Canada without a Central Bank, 1872-1913

between the overall and current-account surplus is defined as residual inflows. In the absence of any errors and omissions, the latter would cover net interest and dividend receipts, as well as net inflows of foreign capital, excluding changes in secondary reserves of the chartered banks. Residual inflows are here employed as a proxy for nonmonetary capital flows. It is safe to assume that over the business cycle, the residual was closely correlated with nonmonetary capital flows since net interest and dividend receipts, in all probability, did not display much cyclical variability.1O The data on the current account are drawn from the standard sources (Viner 1924; Hartland 1955, 1960), save for the inclusion of new estimates of nonmonetary gold flows. The series on international monetary assets is compiled from the official banking statistics (Curtis 1931), but incorporates new data on monetary gold held by the chartered banks and the government, as well as revised estimates of secondary reserves. The available data on international monetary assets appear to be reasonably accurate for the post-1900 period, but not for the earlier years. However, the quality of the pre-1900 data is adequate for analyz~ng the cyclical attributes of the overall surplus. In order to identify the cyclical characteristics of the Canadian balance of payments, the three components are related to the Canadian reference-cycle turning points (table 12.1). As indicated by that table, the overall surplus exhibits a cyclical pattern that is remarkably regular. The largest (smallest) surpluses tended to coincide with reference-cycle troughs (peaks), that is, the overall surplus was negatively correlated with the reference cycle throughout the period under study. The only exception to this finding was the cyclical upswing from 1896 to 1900, during which exports of gold (Yukon gold rush) and other mineral products grew very rapidly. The current-account surplus also moved countercyclically, but table 12.1 suggests that its cyclical pattern was not as regular as that of the overall surplus.ll For the subperiod from 1894 to 1904, in particular, no distinctive pattern can be observed. Residual inflows, by contrast, displayed a complex pattern. From about 1885 to 1895 and 1900 to 1913, they also seem to have varied countercyclically, but during the remainder of the period the fluctuations were irregular. The conclusions drawn from a cursory examination of table 12.1 are confirmed by a comparison of the average overall and current-account surplus, as well as average residual inflows, observed during the boom and depression phases of the reference cycle. The boom (depression) phase is defined as the period between the midpoint of a cyclical expansion (contraction) and the midpoint ofthe subsequent contraction (expansion). For each of the three series shown in table 12.1, the yearly observations are assigned to either the boom or depression phase, and averages are calculated for each of the two phases (table 12.2). Since the

-8.2 -1.7

-9.9 -21.9 -28.6 -28.5 -33.3 -19.7 -10.0 -8.8 -6.9 4.3

5.6 -3.3 -15.4 -21.0 -15.6 -12.5

1870 1871 1872 1873 1874 1875 1876 1877 1878 1879

1880 1881 1882 1883 1884 1885

0.4 -2.9 1.4 31.3 12.3 15.2

20.6 29.1 32.8 18.9 7.0 10.2 4.9 16.4

CurrentAccount Surplus (excl. interest Residual and dividends) Inflows

6.0 -6.2 -14.0 10.3 -3.3 2.7

-8.0 0.6 -0.5 -0.8 -3.0 1.4 -2.0 20.7

Balance-ofPayments Surplus

T/03

PI 07

T/05

Pilla

Ref. erence .Cycle

Canadian Balance of Payments (millions of dollars)

1868 1869

Table 12.1

1900 1900 1901 1901 1902 1903 1904

1890 1891 1892 1893 1894 1894 1895 1896 1897 1898 1899 1899 18.5 23.3 21.3 21.3 22.9 -14.4 -39.0

-21.4 -12.3 -6.5 -0.5 3.1 4.2 3.9 12.5 42.4 19.2 15.0 14.9

CurrentAccount Surplus (excl. interest and dividends)

-7.5 -12.3 8.7 -1.4 -16.2 18.2 70.4

15.6 28.6 8.4 -2.8 10.4 9.3 -9.4 -11.7 -30.8 -21.4 -15.2 -15.1

Residual Inflows

11.0 11.0 30.0 19.9 6.7 3.8 31.4

-5.8 16.3 1.9 -3.3 13.5 13.5 -5.5 0.8 11.6 -2.2 -0.2 -0.2

Balance-ofPayments Surplus

T/06

P/12

T/02

PI 04

P/08 a T/08 a

PI 02 T/03

T/03 a

Reference Cycle

-14.5 -13.7 -17.5 -17.5 -22.9 -20.8

10.5 11.5 26.9 28.8 8.3 15.0

-4.0 -2.2 9.4 11.3 -14.6 -5.8 P/07 a

PI 02 T/02

1905 1906 1907 1908 1909 1910 1911 1912 1913

-33.2 -52.1 -119.5 -29.7 -66.8 -141.9 -213.6 -316.9 -254.5 36.9 53.1 102.4 147.9 94.8 112.5 248.8 310.3 281.7

3.7 1.0 -17.1 118.2 28.0 -29.4 35.2 -6.6 27.2 PI 03 T/07 Pill

T/07

P/12

Sources: Rich 1983, table 2.4. For data on the Canadian reference cycle, see Hay 1966. Notes: The turning points are identified by P for peak and T for trough. The number following the dash indicates the month in which a turning point occurred. Discontinuities in the data, shown by entries above and below a horizontal line for the same year, are accounted for as follows: 1888: The discontinuity in the overall balance-of-payments surplus is due to discontinuities in the available data on monetary gold and secondary reserves of the chartered banks. The official statistical sources on monetary gold (Curtis 1931, p. 36) omit part of the gold the chartered banks held at their foreign branches. For this reason, I compiled a new series on monetary gold including most of the omitted items. The new series is available for the period from the end of 1887 to the end of 1913. For the pre-1887 period, I employed the Curtis data. As far as secondary reserves are concerned, no information is available on the chartered banks' foreign call loans prior to 31 December 1887. 1890: Minor discontinuity in the current-account surplus, due to a break in the data on net exports of nonmonetary silver. 1894: Discontinuity in the current-account surplus. Prior to 1894, reasonably reliable data on merchandise trade are available only for fiscal years. For the pre-1894 period, the data shown in the table are calendar-year estimates, derived from fiscal-year data. 1899: Minor discontinuity in the current-account surplus. Data for 1899 onwards include my own estimates for net exports of nonmonetary refined gold. 1900: Major discontinuity in the current-account surplus, due to a discontinuity in the Hartland (1955, 1960) estimates of nonmerchandise trade. 1901: Major discontinuity in the overall balance-of-payments surplus due to a discontinuity in the data on the chartered banks' secondary reserves. The official sources (Curtis 1931) do not provide data on foreign call loans of the chartered banks for the period prior to 31 July 1900. On the basis of unpublished information, I was able to make fairly crude annual estimates of foreign call loans back to 1887. aS ome uncertainty exists about the exact timing of these turning points.

1886 1887 1888 1888 1889 1890

558

Georg Rich

Table 12.2

Cyclical Characteristics of the Balance of Payments and the Monetary Base (millions of dollars)

Averages of Annual Data

Boom Phase

Depression Phase

1872-87 Current-account surplus Residual inflows Overall surplus d monetary base

-18.8 16.0 -2.8 -2.6

-5.6 9.9 4.3 4.8

-9.5 4.0 -5.5 -5.6

-3.3 13.8 10.5 10.4

-117.9 119.6 1.7 2.4

-58.8 100.5 41.7 41.0

1888-96 Current-account surplus Residual inflows Overall surplus d monetary base

1901-13 Current-account surplus Residual inflows Overall surplus d monetary base

Sources: Table 12.1 and Rich 1983, tables 2-1 and 2-2. Notes: The following years are assigned to the boom phase: 1872-76, 1881-83, 1886-87,

1889-90, 1892-93, 1895, 1898-1900, 1902-3, 1906-7, 1909-10, 1912-13. The remaining years are assumed to belong to the depression phase.

data in table 12.1 suffer from a number of major breaks, the period under study is divided up into subperiods delineated by these breaks. The subperiod extending from 1897 to 1900, for which the normal cyclical pattern of the overall surplus does not obtain, is ignored altogether. If the current account and overall surplus had reached a peak (trough) near a reference-cycle trough (peak), we should find that the corresponding averages were consistently higher during depressions than during booms. Table 12.2 brings out clearly the expected pattern. Not surprisingly, for residual inflows the data do not reveal a clear-cut pattern. From 1872 to 1887 residual inflows moved procyclically, if at all, while in the second subperiod the observed pattern was countercyclical. In the third subperiod no marked cyclical pattern seems to have prevailed. However, the average for booms is strongly influenced by the exceedingly high figure for 1912. If the averages are recalculated for the shorter subperiod from 1901 to 1911, they take on values of $60.8 and $100.5 million for booms and depressions, respectively. Thus, the evidence confirms our earlier observation that during much of the post-1900 period, residual inflows varied countercyclically. On the basis of quarterly data available for the post-1900 period, it is possible to test the null hypothesis that on average, the overall surplus or the growth in international monetary assets was the same during booms and during depressions. The test of the null hypothesis involves rates of

559

Canada without a Central Bank, 1872-1913

change, rather than first differences in international monetary assets, in order to eliminate the scale effect arising from the rapid growth in that variable after 1900. If the scale effect were not eliminated, the test would be biased against the cyclical fluctuations in the overall surplus observed in the early part of the 1900s. As indicated by table 12.3, the null hypothesis is clearly rejected at the 99 percent level of significance, that is, the rate of growth in international monetary assets displayed a statistically significant countercyclical pattern. The countercyclical movements in the overall surplus also produced countercyclical movements in the growth of the monetary base, assumed to embrace the bank's cash (in the form of monetary gold and Dominion notes) and secondary reserves, as well as Dominion notes in the hands of the nonbank public. Alternatively, the monetary base may be defined as the sum of international monetary assets and uncovered Dominion notes. Since I pointed out earlier that uncovered Dominion notes did not vary much over the business cycle, the cyclical pattern of the monetary base was almost identical to that of international monetary assets (tables 12.2 and 12.3). As indicated by figure 12.2, the growth in the monetary base invariably reached a peak (trough) near reference-cycle troughs (peaks). In the subsequent analysis, I assume that the monetary base consisted entirely of international monetary assets. 12.3.2 Money Stock The cyclical pattern of the balance of payments can be contrasted with that of the money stock. In this study I employ a broadly defined concept of the money stock, embracing Dominion notes outside the banking system, demand and notice deposits in the hands of the private nonbank public and the provinces, as well as notes issued by the chartered banks. Notice (or time) deposits are included in the money stock since they differed from demand deposits by degree rather than substance. A large fraction of notice deposits consisted of funds in savings accounts that were endowed with limited checking privileges. In table 12.3 various significance tests are performed in order to ascertain the cyclical attributes of the money stock. The tests are based on quarterly rates of growth in the money stock for the period from 1874 III to 1913 IV. The evidence suggests that the difference in the average growth between booms and depressions was not statistically significant. However, a distinctive cyclical pattern can be discerned from the data if the reference cycle is split up into an expansion and a contraction phase (extending from trough to peak and peak to trough respectively). Throughout the period under study, the average growth in the money stock was significantly higher during expansions than during contractions, with the exception of the subperiod from 1887 to 1895 for which no statistically significant pattern can be observed.

0.80

1.39 2.92

0.80

1.00 1.18 2.28

1901 111-1913 I

1888 111-1894 IV 1901 111-1913 I

1901 111-19131

1874 111-1886 IV 1887 1-1895 III 1895 IV-1913 I

Boom Phase

0.62 1.56 2.40

4.14

1.99 4.06

4.62

Depression Phase

Expansion Phase

International Monetary Assets 5.14 (52) 2.76 Monetary Gold 0.43 (16) 1.67 2.51 (52) 2.83 Monetary Base 2.55 5.00 (52) Money Stock 2.24 0.82 (52) 1.79 (39) 1.37 0.55 (78) 2.65

t-values

0.24 (52) 7.86 (51) 0.66 (40) 4.20 (78)

-0.19 1.21 1.86

1.00 (15) 3.11 (52)

0.14 (52)

t-values

2.35

3.10 4.25

2.62

Contraction Phase

Cyclical Characteristics of International Monetary Assets, the Monetary Base, and the Money Stock (arithmetic means of quarterly rates of change)

Source: Rich 1983, tables 2-1 and 2-2. Notes: The data are smoothed by a seven-quarter moving average in order to eliminate intracyclical variation. Rates of change are calculated from the smoothed series. The significance tests for monetary gold for the period 1888 111-1894 IV rest on semiannual rates of change since quarterly data are unavailable. The semiannual rates of change are calculated from data smoothed by a 1V2-year moving average. The quarterly and semiannual observations, respectively, are assigned to the various reference-cycle phases as follows: Boom phase-quarterly: 1874111-1876 III, 1880 IV-1883 IV, 18861-1887 III, 188911-1890 IV, 18921-1893 III, 1984 IV-1896 I, 189811-1900 III, 1902 1-1903 III, 1905 111-1907 III, 1909 11-1910 IV, 1921 1-1913 IV. Boom phase-semiannually: 1889 first half-1890 second half, 1892 first half-1893 second half, 1894 second half. Depression phase: remaining and overlapping observations. Expansion phase-quarterly: 187911-1882 III, 1885 1-1887 I, 18881-1890 III, 1891 1-1893 I, 18941-1895 III, 1896111-1900 II, 1901 1-1902 IV, 1904 11-1906 IV, 1908 111-1910 I, 1911 111-1912 IV. Expansion phase-semiannually: 1888 second half-1890 second half, 1891 first half-1893 first half, 1894 first and second halves. Contraction phase: remaining and overlapping observations. The figures in parentheses denote degrees of freedom. For the significance test, see Yamane 1973, pp. 661-69.

Table 12.3

561

Canada without a Central Bank, 1872-1913

The evidence suggests that in general, the growth in the money stock reached a cyclical peak (trough) near the midpoint of an expansion (contraction). From the cyclical pattern displayed by the growth rate of the money stock, it is possible to make inferences about the cyclical pattern of its level. If the money stock had been characterized by regular oscillations, the turning points in its level would have followed the corresponding turning points in its rate of growth (or, more precisely, in its first differences) by a lag amounting to one quarter of the length of a full reference cycle. Since the growth rates typically peaked at the midpoint of expansions, it is likely that the peaks in the level coincided with reference-cycle peaks. An analogous pattern would have obtained for the contraction phase. Consequently, the results of table 12.3 can also be interpreted to imply that the level of the money stock was positively correlated with the level of economic activity. It would be useful to test directly the proposition that the cyclical turning points in the levels of the money stock and the reference cycle tended to coincide. As it is well known, however, all the available techniques for identifying the cyclical turning points in a series subject to a strong trend are somewhat arbitrary (see Mintz 1969; Beveridge and Nelson 1981). Besides drawing inferences from the observed variation in growth rates, we may analyze the cyclical attributes of the money stock on the basis of a detrended series. Figures 12.1 and 12.2 show how the deviations in the money stock from its trend are related to the reference cycle. The message conveyed by the two charts tends to confirm the conclusions drawn from the analysis of growth rates. Save for the subperiod from 1887 to 1895, the detrended money stock as a rule, reached peaks (troughs) near reference-cycle peaks (troughs) with little evidence of systematic leads or lags between the turning points in the two series.12 The only exceptions to this rule were the expansion of 1885-87 and the contraction of 1910-11. During the expansion of 1885-87, the detrended money stock decreased, but at a much lower rate than during the preceding contraction. Similarly, the contraction of 1910-11 witnessed a decline in the growth of, but not an absolute decrease in, the detrended money stock. During the subperiod from 1887 to 1895, by contrast, the detrended money stock did not always move procyclically (1894-95) or the cyclical turning points in that stock led cyclical turning points in the reference cycle by several quarters (1887, 1889, and 1890). 12.3.3

Relationships between the Money Stock and the Balance of Payments

Since the balance of payments was virtually the only source of change in the monetary base, the relationship between the money stock and the balance of payments can be analyzed by examining the link between that stock and the monetary base. Considering the cyclical attributes of the

Fig. 12.1

'I

•

'

I'

I

'

I'

,

..........

••••••••••••

••••••••••••

.' • I,

.

••••••••••••

••••.•••••••••••

."

••••••

••••••

'I

I,

•

I,

• I

•

•

I,

••••

•••• , ' , I' •••••••• •

••••••••••

•••••••

'. "

'

'

"

1

75

••••••

•

'

••••••••••••

.

•

'

'.'

••••••••••••

',1 • • • • • • • • • • • • • • 11. 1 •••••••••••••

'

•

••

••••••••••••••••

I

' '

I

•

•

. . ................... . ... ..... .....

..

:::::::::::::Jili .... • I, •••••••••••••••

• ' I •••••••••

II

· I, I' •••••.••

II

'I I,

I

•

· I, .••.•.•.. · 'I " " ..•.. · I, ••••.. ','

••••••••••

,',

• I, ,',

•

· I, I' •••••.• · I, ,', .•..••

· ' I ' ••••••.••••.•• • I, ,I ' , ' • • • • • • •

'

••.•••...•••••. .·· I,I, .•.••..•••..•.. .

.'

I'

• I, I' I' •••••••.•••••

"

•

• I, I ' , ' ••••••••.•.•

80

I

I °

85

0,. · · i·· °1•

. .........

.:::::::::T

, .' ,

..

I · . i·

P::: ::T

I

'°1'

... 90

1

p:: :T

I

t

'-,'

..

... ..

.:

: ::,:

"

···.......-. ··.....' , '

p::::r ....

... ..

'

95

, ::1:::

~.

.. .......... ...... ..... . ...... .. ..

.:':'

... ..

.- .- . .....

.....

"

"

·, :.: ',' ·" '.'.'

1900

· :.:

.' .' . ' .' .' "

I ' .'

I'

"

.' .'

P: ::h'

5.40 + 0.0241t, R 2 = 0.999. (888.7) (155.0)

1nM =

M and t denote the money stock and time, respectively; numbers in parentheses are t-values. Shaded areas identify reference-cycle contractions. Source: See Rich 1983, table 2-1.

Subperiod 1896 111-1913 I:

4.34 + 0.0121t, R 2 = 0.958. (308.9) (44.9)

Subperiod 1874 11-1896 III: 1nM =

Cyclical movements in the money stock, 187411-1901 II. The data were adjusted by a seven-quarter moving average. The percentage deviations represent the difference between the natural logs of the actual and trend values of the money stock. Since the growth in the money stock accelerated considerably with the onset of the natural-resources boom around 1896, trend values were determined by two regression equations:

-20

-10

o

10

20

Quarterly deviation f rom trend ...-----":"".--:-.--:-.-:-.-,..--r.-----..."r--r-u-ro . ...........,.,- - - - . - ,:-r'":'",r-----.....----"""':'..-;--~.:_:""._-----~.:_:_:.:-'I (percent>:::::::::: T p P::::~

Fig. 12.2

-20

-10

o

10

20

·

I

I-':~::: :)~

,

.

•••••

,. ... . . . ..

~""'"

., ,

\

\

.

-

-

/

..

,.J..·

Moneystock

.Monetary

, · ,......

\ ..

r

• ••••••••• •

~~~~~~~~~~ ~

. .

-

. ....

\ \

..... ..... . .. \ •••••••

<~ ~ ~ ~ ~ ~ ~

P:

."-. .

..

•••• •••••••• •

••••••

:1. : : · : ·

• • • • '.

• : .:

• • •• ••• ' • •••••• • ••

e

\

'..

.•.•.•.••.••

: : : :: :. : :. : •• , • ••.• • •• .... • •••••••• • •

• : • : : -: :.

.•.•••••

••••••••

• •••• ••• •• • •

l~u::~~~:: :r~trendl i....•.•.... f:::::!::: .....•... ~1N!:::. .•.. ."",.1-1 .... ./ ' :::::::~::::? e, ::::::::::: ::::::::: :::::..:::::: .:::::::;:::/\. ::::::::: :::,.::::::::

•••••••••" . . • (deviation

~ ~ ~ ~ ~ ~ ~ ~ ~ (;

.

.

.~.' ~~

e,• :·::::,:::: :'. \ :::::

\

y : • :. : • : • • :onetar base • • ••• • •• • ••• ,. rate •. . ,,(quarterly ' · · · · · · · t · · · \. ~f change)

:::::::::::: ~:

...•••••..• •

:::::!WT

• •••••••••• •

..... :\. '.. ~ ~ ~ ~ ~ ~ :f: ~ ~ ~

; ;

~;~ ~ ~ ~~~~

·· •..•,\........ .. -J.. ·

-

•••••••••••• •

-

PiiiiiiiiiW r I/-:\~iimiii ..x........ iiiiiiiiiiiii/~~~!~~~)i i:t~mi i ~ ji;Tj~~~ ~1 ;;~ ~ i~ i~~~ ~ -,

-

2

4

6

10 8

Quarterly rate of change (percent)

=

4.47 + 0.0281!, R 2 = 0.950. (167.9) (29.8)

Source: See fig. 12.1.

InH

Cyclical movements in the monetary base and the money stock, 1900 IV-1913 I. The data were adjusted by a seven-quarter moving average. Trend values of the monetary base (H) were determined by the following regression equation for the period 1901 11-1913 I.

•,

\

\.

• •••••••••• •

-

PiiiiiiWiiiT ~-, iii Wim i

.,

I

1901

• T

Quarterly deviation from trend (percent)

564

Georg Rich

two monetary aggregates, it is evident that the cyclical turning points in the money stock tended to lag the corresponding turning points in the monetary base. The growth in the base typically peaked near referencecycle troughs, while the growth in the money stock attained its highest value near the midpoint of the subsequent expansion (tables 12.2 and 12.3). The observed growth patterns imply that the level of the monetary base and the money stock, respectively, peaked near the midpoint and the end of reference-cycle expansions (figure 12.2). Thus, under the pre-1914 Canadian gold standard, the money stock and the monetary base, at least in the short run, were not closely correlated. The reason that there was only a very loose short-run link between the two variables must be sought in bank-reserve management. In order to analyze this link, the money stock is related to the monetary base by the well-known identity: (1)

M

= H(l + e)/(c + e),

where M, H, c, and e denote, respectively, the money stock, the monetary base, the aggregate reserve ratio (cash and secondary reserves -;liabilities, i.e., deposits and notes) of the chartered banks, and the currency-liability ratio (Dominion notes in the hands of the nonbank public -;- liabilities of the banks). Identity (1) captures the various possible sources of change in the money stock. Since H varied countercyclically, the procyclical pattern in the money stock was not attributable to changes in H, but to marked countercyclical movements in the aggregate reserve ratio. That ratio typically rose during the depression phase of the reference cycle, reaching a peak as it approached the midpoint of an expansion. The opposite pattern can be observed for the boom phase (table 12.4.). During the first half of an expansion, a cyclical increase in the money stock was accompanied by a cyclical increase in the monetary base, but part of the latter was absorbed by a drop in the aggregate reserve ratio. During the second half, however, the cyclical increase in the money stock was due entirely to a cyclical rise in the aggregate reserve ratio. The currency-liability ratio, by contrast, did not make any contribution to the cyclical variation in the money stock. Holdings of Dominion notes outside the banking system were neither a large component of the money stock,13 nor did e fluctuate cyclically (table 12.4).14 Prior to 1914, the Canadian money stock consisted almost exclusively of assets supplied by the chartered banks. An analysis of the link between the Canadian money stock and the balance of payments also yields intriguing results concerning the role of gold flows in the adjustment mechanism. During the period under study, the monetary gold stock typically varied in a countercyclical fashion, very much like total international monetary assets. The only exception to this pattern was once again the subperiod from 1897 to 1900. Prior to 1897,

565

Canada without a Central Bank, 1872-1913

Table 12.4

Cyclical Characteristics of the Aggregate Reserve Ratio and the Currency-Liability Ratio (percent) Boom Phase

Aggregate Reserve Ratio 1873-1887 30.2 1887-1900 22.9 190G-1913 25.8 Currency-Liability Ratio 1873-1913 3.0

Depression Phase

Expansion Phase

Contraction Phase

Total

29.1 24.3 25.9

37.5 24.3 28.7

26.0 22.5 23.8

29.6 23.7 25.9

3.0

3.0

3.0

3.0

Source: Calculated from Rich 1983, table 3-3. Notes: The data in the table represent arithmetic means of end-of-year ratios. The annual observations were allocated to the various reference-cycle phases in the following way. Boom phase: 1873-75, 188G-82, 1886, 1889,1892,1894-95,1898-99,1902,1905-6,1909, 1912-13. Expansion phase: 1879-81, 1885-86, 1888-89, 1891-92, 1894, 1896--99, 1901, 1904-5,1908-9,1911. The depression and contraction phases cover the respective remaining years.

however, the countercyclical movements in the monetary gold stock were completely dwarfed by secondary-reserve flows. After 1900 the amplitude of these movements increased considerably relative to that of secondary reserve flows, but the latter continued to fluctuate more strongly than the former. Table 12.3 sheds light on the cyclical attributes of the monetary gold stock. Over the subperiod from 1888 III to 1894 IV, for which reasonably reliable semi-annual data are available, the monetary gold stock tended to grow most rapidly (slowly) during the contraction/ expansion phase of the reference cycle. The observed cyclical pattern, however, was not statistically significant. In the post-1900 period, by contrast, the growth in the monetary gold stock displayed a statistically significant countercyclical pattern. IS In summary, the Canadian evidence on the link between the money stock and the balance of payments does not accord well with the classical PSF-analysis. Over the business cycle, the growth in the money stock was not closely correlated with the overall balance-of-payments surplus. During the latter stages of a reference-cycle expansion and contraction, in particular, the balance of payments failed to act as an effective prophylactic against procyclical movements in th~ money stock. These movements were not due to a procyclical pattern of the current-account surplus or to the influence of a Taussig effect, but were explained by a defect in the Canadian PSF mechanism. Consequently, the evidence clearly contradicts Viner's assertion that the Canadian mechanism operated efficiently both in the short and long run. At least for the post-1900 period, the evidence is also inconsistent with Viner's contention that payments imbalances were settled by flows of

566

Georg Rich

secondary reserves, rather than monetary gold. Admittedly, over the business cycle, secondary reserves fluctuated more strongly than monetary gold, but after 1900 the roles played in the adjustment mechanism by the two categories of monetary flows were no longer fundamentally different. Interestingly enough, during cyclical upswings, the deterioration in the overall balance of payments implied that Canada curtailed the growth in imports of monetary gold. In this manner, precious metal was made available to the rest of the world and helped to accommodate to a modest extent cyclical upswings through increases in the money stock abroad. 12.4 The Role of Bank-Reserve Management in the Canadian Business Cycle Considering the importance of bank-reserve management as a source of procyclical movements in the Canadian money stock, the question arises how profit-maximizing banks, operating within the constraints of the gold standard, could contribute to cyclical instability in economic activity. In this section of the paper, I sketch in a nontechnical manner a model designed to elucidate the significance of bank-reserve management as a destabilizing force. 16 The theoretical analysis rests upon two building blocks. First, following the majority of other investigators, I assume that cyclical movements in Canadian economic activity were induced by disturbances from abroad. For the sake of simplicity, the foreign cyclical disturbance is depicted by an exogenous change in Canadian exports. Furthermore, drawing on the research by Bryce (1939) and Rosenbluth (1958), I assume that a cyclical change in exports resulted in a deterioration rather than an improvement in the Canadian currentaccount balance, owing to an accelerator effect on imports of capital goods. Second, the model incorporates the observation that the chartered banks accommodated cyclical movements in economic activity by varying the aggregate reserve ratio. Countercyclical changes in the aggregate reserve ratio could have been caused by a number of factors, but as a first step I assume that they were the result of procyclical movements in domestic interest rates or in the opportunity cost of holding reserves. Now suppose that a cyclical expansion in foreign economic activity elicited a rise in Canadian exports. Provided the Canadian current and capital-account balance was zero initially, the impact effect of the exogenous disturbance was to augment economic activity, the demand for money, and interest rates in Canada. As interest rates rose, the chartered banks were prompted to expand the money supply by lowering the aggregate reserve ratio. But the impact effect of the exogenous disturbance was not the end of the story. The cyclical surge in Canadian economic activity was associated with a deterioration in the current-

567

Canada without a Central Bank, 1872-1913

account balance. Abstracting for the moment from capital flows, the current-account deficit should have led to a reduction in the monetary base, pushing up domestic interest rates still further. The surge in economic activity tapered off when exports 'reached a cyclical peak. The theoretical analysis suggests that a subsequent downturn in economic activity, induced by a cyclical drop in exports, did not cause domestic interest rates to decline instantaneously. At this stage of the cycle, domestic interest rates were influenced by two opposing forces. On the one hand, the cyclical contraction in economic activity and the attendant decrease in the demand for money tended to lower domestic interest rates. On the other hand, the overall balance-of-payments deficit, inherited .from the previous expansion, induced a decrease in the monetary base, putting upward pressure on domestic interest rates. Since the monetary base ceased to fall at the end of a boom, domestic rates should have reached a cyclical peak at some point during the first half of a reference-cycle contraction. In other words, the cyclical turning points in domestic interest rates should have followed the corresponding reference-cycle turning points with a time lag amounting to less than half the length of the respective reference-cycle expansion or contraction. Thus, the model yields a testable proposition about the cyclical pattern of Canadian interest rates. Since there is a dearth of statistics on pre-1914 Canadian interest rates, I limit myself to an examination of the cyclical movements in the Montreal call-loan rate, for which data are available for the period 19001913.17 Figure 12.3 suggests that the Montreal call-loan rate tended to peak during the first half (1903) or near the midpoint (1911) of referencecycle contractions. During the contraction of 1907-8, it attained a peak of 6 percent in the first quarter of 1907 and stayed at that leveJ for half a year (on a seasonally adjusted basis). Therefore, the timing of that peak is somewhat uncertain. The evidence is consistent with the view that the Montreal call-loan rate peaked either during the first half or near the midpoint of the contraction of 1907-8. Similarly, the cyclical troughs in that rate tended to occur during the first half (1901, 1905) or near the midpoint (1909, 1912) of reference-cycle expansions. On the whole the cyclical peaks (troughs) in the Montreal call-loan rate were closer to the midpoints of reference-cycle contractions (expansions) than to the reference-cycle peaks (troughs). For the period extending from the trough of 1901 to the peak of 1912, the turning points in the Montreal call-loan rate lagged the corresponding turning points in the reference cycle by an average of 2.5 quarters and led the corresponding midpoints by an average of 0.7 quarters. I8 Thus, the theoretical analysis is at least partly supported by the empirical evidence. The Montreal call-loan rate clearly lagged the reference cycle, but for roughly half of all the expansions and contractions the lag was longer than predicted by the model.

Fig. 12.3

2

4

5

6

7

8

II.

••••••••

•••••

•• 1.1

•

,',

'.'.

','.

:: ::::\:::.'::

.:::::::::::::. ~ \

:::: I

y.. ..

, ...... ','. :': . .:.: ',. , .

:.: .

•• •

'.' :~: '.'

./

::: :

.. , ~~

....

,I

'

••••••

,',

••••••••••

I,

••••••

, t •••••••

~

It.

II

.'

t'

II

.'

'

. ' I ' •••••••••

............. ·.. . ·· '.' . . ·.'',' ..

:::::::::~::: :::::::::: ~:,

• •• ' •••• \. •••

'

I

Boston

f 1 I I.

I

II'

"

•

\

•

•

•

•

•

•••

•

•

••

\

•

i: :

•

•

•

•

•

•

•

•

i

,~

. .I..........I. •~ :: S : : : : : ::

•

•••••••••• I.

•••••••••••••

............. ............. ............. ............. .. .

••••••••••

.'.

•

...

• ••••••••••

.',

~~~~~~~~~~~~~\_/ ~~~~~~~~~~j .... ....

...........•,~

• ',' t'

.'.'

•

•

•

• '.'

•.:.i.;:

: : ::

:::: • •• •

• • ••

·:::

.. ·

.· · · .·

• •••

: : ::

•

::::

::::

"

•

'

,

'.'

'.'

e','

.~~~~ '.' ..,'.' '.'

: : ::

: ::: ....

.•.•

::::

/ / -.-:..' .... : : :'\

~~~~~~U~~~·' ~:rj

: ::::: :::::

. : : : : : : : : ::

• • • •• ' . •••• ::......... • • •• '. • •••

:::::::::::

·

• ' . • • •• • • '.

· · · · · · · · · ..

• ••••••••••

•

: : : : : : : : : ::

•

: : : : : : : : : ::

:::::::::::

••••••••

.'

II

P::::

•

. ..... .....

• I, ' , ' •••••

P: : : : : : : : : : : T

: : : : : : : : : : : •• e~ : : : :e~ : : : :: •I / /-.,.•. ~.... .' I :............ : : : : : : : : : : : \: : . :: : :: :'~.~ / . ' .,........ / : : : : : : : : : : : : :, e,.,.' I : : : : :':-;':';'~ /

I.......... . : : : : : : : : : :: :

~~~~~~~~)~~ ··········1·

:: :::::: \,::

: : : : : :: ••:

•••••••••••• Montreal

e' .....

.. :::: :\.: :~.'

·: ~: :·::.. : · 1:...· ··

···r· .\.'. ·

· '1·L·,·.·. ... · · .'. · \ ."1 ., r' •••• •

•

:

•

: : : :: :

\.' ::: :/: ,: : : ::

•

::::

::::j:::::: : ~\:

,'.........

, ' , '._ •••••

........... .

P ::::::::::: :T

J ' \ •••••,•••••

I

I

I 1/

I New York

I

. ...........• _e "

~

:::: :~:: :\::::

\

. ~~""'" .•:::t~::::::: . • .' \ • • ~. • '.

.~~:::::::

,', ..". ...•.... I,

.'

•

•

.. ...'.' .. . ... ... .. ... ... ... .' .'.: .. ·.... , '.

.'

P:::::: :::: T

.. :::: '.' . • .' ·

II.

".. '... '... '... '... '... '... '... '... '... I... '... '... .'....

I

•• 1.

•

::: :T

Call-loan rates in Montreal, New York, and Boston, 1901-12. The U.S. short-term rate is an average for the last month of the quarter, while the Montreal rate refers to the first day following the end of the quarter. All the interest-rate data were seasonally adjusted by a thirteen-month moving average. Sources: Montreal call-loan rate: Canada 1915, p. 739. Short-term interest rates in New York: Macaulay 1938, appendix, table 10; and in Boston: Goodhart 1969, pp. 206-19.

Percent

569

Canada without a Central Bank, 1872-1913

Interestingly enough, the cyclical pattern of the Montreal call-loan rate mirrored closely the cyclical pattern of the monetary base (figure 12.4). The strong negative correlation observed between cyclical movements in the two variables suggests that over the business cycle the balance of payments was the chief determinant of the Montreal call-loan rate. During the depression phase of the reference-cycle, for example, the size of the overall surplus was above average. International monetary assets, as well as the monetary base, tended to grow rapidly and caused the Montreal call-loan rate to decline. The model is capable of explaining the cyclical patterns revealed by figure 12.4 if we assume that the aggregate reserve ratio was negatively related to both domestic interest rates and cyclical movements in economic activity. It is plausible to argue that a cyclical expansion (contraction) in economic activity resulted in a drop (rise) in the aggregate reserve ratio even if interest rates remained constant. A cyclical slump in economic activity, for example, opened up the prospect of major bank failures. If the reserve holdings of the chartered banks were inadequate, the prospect of bank failures could result in a run on the banks and an attendant liquidity crisis. In order to be able to weather a potential liquidity crisis, the banks were likely to build up precautionary reserves when the economy pilll1ged into a recession.19 It would have been sensible for the chartered banks to vary their reserve ratios in a countercyclical manner despite the fact that during the period under study, Canadaunlike the United States-did not experience runs on banks and major liquidity crises. The public confidence instilled by strong countercyclical movements in the aggregate reserve ratio could have been a reason for the absence of financial panics in Canada. Provided the aggregate reserve ratio responded to cyclical movements in economic activity, a cyclical increase in exports and the ensuing expansion in economic activity, in the short run, caused both the supply of and demand for money to rise. The supply was boosted as the chartered banks curtailed their precautionary reserves in relation to their liabilities. Given the simultaneous increase in money demand and supply, the Montreal call-loan rate need not have changed in the short run. However, this did not imply that it was completely invariant to the 'cyclical disturbance. In the longer run, the cyclical deterioration in the overall balance of payments, as well as the attendant drop in the monetary base, led to an increase in the Montreal call-loan rate. Thus, it is possible to explain why a cyclical increase in that rate was due chiefly to a cyclical reduction in the monetary base, despite the fact that the ultimate source of the disturbance was a cyclical expansion in exports. Needless to say, the explanation advanced in this paper for the cyclical pattern of the Montreal call-loan rate would not be valid if the Canadian and U.S. money markets had been closely integrated. In that event,

Fig. 12.4

-1

o

4

5

6

7

8

9

10

11

Monetary Base

/..,.,..

/

...

••••••••

••••••••

.... .... I'll.... ........ ........ ........ ........ ........ ........ . ....... .......

'~""

I \ :::::::: ~::::::: I ::::\::: I :::::;:: / ..... ".. ...... "'-......... ........ •••••••• ........ .... ........

,',

I'

I'

' ,•'• •"•

••••

(1.

.',

•••••

I'

1913

'.' "

"".

.....

• 1'1'1'

II'

"

I

•

•

It

..

..• I, 'I,, ' , ' ,.. . '. .. II , ' ,. ... .. . .. ... ,,'' ,, ... . . .·. ... ·· , ' , .. ··· " , . I'.. I.·.·.· · ,', ','

..

:::::::: ::::::::. :::::::: :. :::::::.

,..~,::::::

,',

P:::::::: ... ... . :::::::: 0.4

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Quarterly change in the Montreal call-loan rate and the Canadian monetary base, 1901-13. Sources: See figs. 12.2 and 12.3.

1901

/

/

/

P::::::::T ........ ........ ........ ........ /-, :::::::: \ ::::::::

Call loan rate 0.5

571

Canada without a Central Bank, 1872-1913

Canadian money-market rates would have moved in unison with their U.S. equivalents and would not have been systematically related to the Canadian monetary base. Prior to 1914, however, cyclical movements in Canadian and U. S. call-loan rates were anything but perfectly synchronized. In figure 12.3 the Montreal call-loan rate is compared with the corresponding rates in New York and Boston, the two closest major American financial centers. The evidence clearly indicates that the Montreal call-loan rate lagged its New York and Boston equivalents. Moreover, the cyclical amplitude of the two U.S. rates was much larger than that of the Montreal call-loan rate. It is interesting to note that the average return on Canadian bank loans, as well as Canadian bond yields, on the whole, also displayed a smaller cyclical variance than their U.S. equivalents (Rich 1983, chap. 5). Since the cyclical stability of interest rates was much greater in Canada than in the United States, it is not surprising that the capital inflow to Canada often varied countercyclically. The relative stability of Canadian interest rates explains why the capital account frequently made a significant contribution to the countercyclical movements in the overall balance-of-payments surplus. Evidently, bank-reserve management was not only responsible for procyclical movements in the money stock, but also served to keep the cyclical variance of Canadian interest rates within narrow bounds. Furthermore, the ability of the banks to accommodate a cyclical surge in economic activity through an expansion in the money stock delayed the upward adjustment in interest rates that would normally have occurred at this stage of the business cycle. Interest rates did eventually rise, but not quickly enough to restrain effectively a cyclical expansion in economic activity. Typically, they continued to rise well after economic activity had reached a cyclical peak. In this way, the rise in interest rates was liable to aggravate the subsequent contraction. Consequently, the evidence lends strong support to the view that bank-reserve management amplified the cyclical fluctuations in Canadian economic activity by diluting the stabilizing role of interest rates. 12.5 Summary and Conclusions

In this paper I attempted to show that the PSF mechanism did not act as an effective stabilizer of cyclical fluctuations in economic activity. Although the gold standard constrained the ability of the chartered banks to create money, the restrictions imposed on the banks were not severe enough to prevent procyclical movements in the Canadian money stock. The fetters were sufficiently flexible to allow them to accommodate a cyclical surge in economic activity through an expansion in the money stock. Accommodative behavior on the part of the banks did elicit balance-of-payments deficits (or a reduction in the growth of interna-

572

Georg Rich

tional monetary assets), as suggested by the classical PSF doctrine. However, while the PSF mechanism operated in the long run, the deficits did not effectively curb the growth in the money stock in the short run. Accommodative behavior also delayed the increase in Canadian interest rates that would normally have followed the cyclical surge in economic activity. Interest rates were adjusted only when the overall balance of payments began to deteriorate. Consequently, the PSF mechanism eventually forced up interest rates. But they responded to the cyclical increase in economic activity with such a long lag that they were liable to aggravate the subsequent contraction, instead of restraining the expansion. In summary, the evidence is inconsistent with Viner's contention that the Canadian PSF mechanism operated smoothly not only in the long run, but also in the short run.

Notes 1. Viner did not compare bank reserves with the money stock, but with bank liabilities to Canadian nonbanks, which consisted of notes, demand and notice (or time) deposits. As will be shown later, bank liabilities accounted for the lion's share of the money stock. 2. Viner also analyzed the more realistic case of a permanent increase in the capital inflow. He clearly realized that there was a difference between a once-and-for-all change in Canada's foreign debt and a permanent increase in its growth (1924, pp. 177-80). In his analysis he ignored the implications of changes in interest payments on the foreign debt. 3. According to Viner (1937, pp. 429-30), it was "possible to argue that at times at least" the observed increase in the money stock was due to purely internal factors, rather than a capital inflow. The rise in the money stock caused the current account to deteriorate. Moreover, "the borrowings were engaged in to obtain the foreign funds necessary to liquidate trade balances already incurred." However, he felt this interpretation was "quite consistent with the orthodox explanation." He also admitted that the current-account balance was likely to respond to an exogenous capital inflow with a lag (1937, p. 423). 4. To be precise, in his 1924 study Viner argued that secondary money creation had played a role only towards the end of the period under study (1924, pp. 189-90). 5. Viner (1924, pp. 184--85) appears to have argued that loans to Canadian residents were unrelated to the capital inflow. 6. See Viner's table VI (1937, p. 428). In 1902-3,1906-7,1910-11, and 1912-13 reserves decreased (column 6), while liabilities increased (column 1). For 1907-8 the opposite pattern can be observed. 7. The Taussig effect has been revived and reformulated by Johnson (1972, chap. 9) and others under the heading of the monetary approach to balance-of-payments analysis. For an application of that approach to the gold standard, see McCloskey and Zecher 1976. 8. It is likely that the world stock of monetary gold was negatively correlated, if at all, with the business cycle. During a cyclical upswing, the demand for nonmonetary gold was liable to 'rise, drawing the precious metal away from monetary use. Needless to say, this argument is not applicable to long swings in economic activity which, among other factors, were caused by changes in gold production, resulting from new discoveries. 9. During the entire period under study, the Canadian government and the chartered banks were never forced to suspend the convertibility into gold of Dominion notes and bank liabilities respectively. After the United States returned to the gold standard in 1879, the

573

Canada without a Central Bank, 1872-1913

mint-par values of the U.S. and Canadian currencies were identical, that is, the Canadian dollar was worth US $1.00. For a detailed discussion of pre-1914 Canadian monetary policy, see Rich 1977, 1983, chap. 7. 10. Much of Canada's foreign debt was held in the form of fixed-interest assets. Some information is available on net interest paid to foreigners by the Canadian government. This item did not vary with the reference cycle (Rich 1983, chap. 7). 11. The countercyclical pattern of the current-account surplus mirrored strong countercyclical movements in the merchandise-trade surplus. It is well known (see Taylor 1931, p. 3) that the Canadian merchandise-trade surplus was negatively correlated with the business cycle. 12. Hay (1967) argues that the turning points in the money stock led the corresponding turning points in economic activity. However, his conclusion is not inconsistent with my own since he compares rates of change in the money stock with the level of economic activity. 13. The Canadian government possessed a monopoly of the issue of notes in denominations of $4 and less. The nonbank public did not use the Dominion notes to any great extent unless it was compelled to do so by the government. 14. This evidence contrasts sharply with that presented by Cagan (1965) for the United States for the same period. For the United States, it was movements in the currency ratio that dominated cyclical movements in the money stock, whereas movements in the reserve ratio played a relatively minor role. 15. A more detailed discussion of these points is provided in Rich 1983, chap. 3. 16. For a full discussion of the model, see Rich 1983, chap. 4. 17. Only rudimentary data are available on other Canadian bank-loan rates. The evidence suggests that the average return on Canadian bank loans exhibited the same cyclical pattern as the Montreal call-loan rate (Rich 1983, chap. 5). 18. The turning points in the Montreal call-loan rate are determined on the basis of seasonally adjusted quarterly data (see figure 12.3). The timing of the turning points does not pose any problems, except for the peak of 1907 that I assume to have occurred at the end of the second quarter. For the reference cycle, a turning point or midpoint occurring in, say, March-April and February, respectively, is assigned to the end and the middle of the first quarter. 19. For a similar explanation of cyclical movements in bank-reserve ratios, see Friedman and Schwartz 1963, pp. 449-62 and Morrison 1966, chap. 3.

References Beach, W. E. 1935. British international gold movements and banking policy, 1881-1913. Cambridge: Harvard University Press. Beckhart, B. H. 1929. The banking system of Canada. New York: Henry Holt and Company. Beveridge, Stephen, and Charles R. Nelson. 1981. A new approach to decomposition of economic time series into permanent and transitory components with particular attention to measurement of the business cycle. Journal of Monetary Economics 7: 151-74. Bloomfield, Arthur I. 1959. Monetary policy under the international gold standard, 1880-1914. New York: Federal Reserve Bank of New York. Bono.mo, Victor, and J. Ernest Tanner. 1972. Canadian sensitivity to

574

Georg Rich

economic cycles in the United States. Review of Economics and Statistics 54: 1-8. Borts, George H. 1964. A theory of long-run international capital movements. Journal of Political Economy 72: 341-59. Bryce, R. B. 1939. The effects on Canada of industrial fluctuations in the United States. Canadian Journal of Economics and Political Science 5: 373-86. Cagan, Phillip. 1965. Determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. Cairncross, A. K. 1968. Investment in Canada, 1900-1913. In The export of capital from Britain 1870-1914, ed. R. Hall. London: Methuen. Canada. 1915. Board of inquiry into cost of living in Canada. Report of the board. Vol 2. Ottawa. Carr, R. M. 1931. The role of price in the international trade mechanism. Quarterly Journal of Economics 45: 710-19. Chambers, Ed'Yard J. 1964. Late nineteenth century business cycles in Canada. Canadian Journal of Economics and Political Science 30: 391-412. Curtis, C. A. 1931. Banking statistics in Canada. In Statistical contributions to Canadian economic history, vol. 1. Toronto: Macmillan. Ford. Alec G. 1962. The gold standard, 1880-1914: Britain and Argentina. Oxford: Clarendon Press. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Goodhart, Charles A. E. 1969. The New York money market and the finance of trade. Harvard Economic Studies, no. 132. Cambridge: Harvard University Press. - - .-.1972. The business of banking, 1891-1914. London: Weidenfeld and Nicolson. Hartland (Thunberg), Penelope. 1955. The Canadian balance of payments since 1868. Mimeo. NBER. - - - . 1960. Canadian balance of payments since 1868. In Trends in the American Economy in the Nineteenth Century, ed. W. N. Parker. Studies in Income and Wealth, no. 24. Princeton: Princeton University Press. Hawtrey, R. G. 1928. Trade and credit. London: Longmans, Green and Co. - - - . 1947. The gold standard in theory and practice. 5th ed. London: Longmans, Green and Co. Hay, Keith A. J. 1966. Early twentieth-century business cycles in Canada. Canadian Journal of Economics and Political Science 32: 354-64. - - - . 1967. Money and cycles in post-confederation Canada. Journal of Political Economy 75: 263-73. Ingram, James C. 1957. Growth in capacity and Canada's balance of payments. American Economic Review 47: 93-104.

575

Canada without a Central Bank, 1872-1913

Johnson, Harry G. 1972. Further essays in monetary economics. London: George Allen and Unwin. Johnson, J. F. 1910. The Canadian banking system. Report prepared for the U.S. National Monetary Commission. 61st Cong., 2d sess. Washington, D.C.: Government Printing Office. McCloskey, Donald N., and J. Richard Zecher. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance of payments, ed. J. A. Frenkel and H. G. Johnson. London: George Allen and Unwin. Macaulay, F. R. 1938. Some theoretical problems suggesied by the movements of interest rates, bond yields, and stock prices in the United States since 1856. New York: National Bureau of Economic Research. Meier, Gerald M. 1953. Economic development and the transfer mechanism: Canada, 1895-1913. Canadian Journal of Economics and Political Science 19: 1-19. Mintz, lIse. 1969. Dating postwar business cycles. New York: National Bureau of Economic Research. Morgenstern, Oscar. 1959. Internationalfinancial transactions and business cycles. Princeton: Princeton University Press. Morrison, George R. 1966. Liquidity preference of commercial banks. Chicago: University of Chicago Press. Rich, Georg. 1977. The gold-reserve requirement under the Dominion Notes Act of 1870: How to deceive Parliament. Canadian Journal of Economics 10: 447-53. ---.1983. The cross of gold: Money and the Canadian business cycle, 1867-1913. Mimeo. Rosenbluth, Gideon. 1958. Changing structural factors in Canada's cyclical sensitivity, 1903-1953. Canadian Journal of Economics and Political Science. 24: 21-43. Shearer, Ronald A. 1965. The foreign currency business of Canadian chartered banks. Canadian Journal of Economics and Political Science 31: 328-57. Stovel, J. A. 1959. Canada in the world economy. Cambridge: Harvard University Press. Taussig, Frank W. 1927. International trade. New York: Macmillan. Taylor, K. W. 1931. Statistics of foreign trade. In Statistical contributions to Canadian economic history, vol. 2. Toronto: Macmillan. Viner, Jacob. 1924. Canada's balance of international indebtedness, 1900-1913. Cambridge: Harvard University Press. ---.1937. Studies in the theory ofinternational trade. London: George Allen and Unwin. Yamane, Taro. 1973. Statistics: An introductory analysis. 3d ed. New York: Harper and Row.

576

Georg Rich

Comment

Peter Temin

It is both a pleasure and a burden to comment on Dr. Rich's paperpleasure because the paper is an interesting one that contains an intriguing view of the Canadian economy before the First World War; a burden because it appears to be a summary of a larger work as yet unpublished and inaccessible to me in the time allowed for the writing of comments. I will enjoy the pleasure and try to ignore the burden, recognizing that Dr. Rich may well have all the answers in his larger work to the questions I pose here. My comment has three parts. First, I want to place Rich's argument about Viner into the context of modern theories of the balance of payments. Second, I will review Rich's argument and the tests he performs to substantiate it. And third, I will redo a few of his tests from a slightly different perspective and comment on the results. There seem to be three theories of the balance of payments that have achieved wide currency today. The price-specie-flow theory views capital flows as exogenous and uses a flow concept of equilibrium: trade flows must interact with capital flows to provide a balance-of-payments equilibrium. The monetary theory of the balance of payments views capital flows as endogenous and uses a stock concept of equilibrium: residents of each country must desire to hold the existing stock of money. These two theories differ both in how they view capital flows (exogenously or endogenously) and how they view equilibrium (flow or stock), suggesting that theories of the balance of payments can be arranged in a two-by-two table. Table C12.1 is such an arrangement, where the two theories just discussed form the main diagonal. Only one of the off-diagonal alternatives is represented in the literature. It is what I will call the "real" theory of the balance of payments which uses a Keynesian flow equilibrium and views capital flows as endogenously determined to maintain this equilibrium. The final combination-exogenous capital flows and stock equilibrium-is not yet represented in the literature. Tests of these different theories are hard to design, which may be one reason why they are still extant. In order to discriminate between them, investigators generally have looked at the mechanisms underlying these different theories. The price-specie-flow theory consequently has been tested by a long line of scholars from Viner to Rich by looking at the operation of the banking system to see if the money stock responded to changes in capital flows. McCloskey and Zecher have evaluated the monetary theory by asking whether markets in different countries were Peter Temin is professor of economics at Massachusetts Institute of Technology.

577

Canada without a Central Bank, 1872-1913

Table e12.1

Classification of Balance-of-Payments Theories Type of Equilibrium Flow

CapitalMovements Views

exogenous endogenous

price-specieflow theory "real" theory

Stock ? monetary theory

integrated. And Ingram (1957) verified that for Canada, capital flows could be explained by the needs of domestic investment. It is worth noting that the tests of the price-specie-flow theory have generally been negative, while the tests of the other two theories have tended to be positive. Does this tell us something about the world or about publication bias? Without the ability to test the suggestion, I suspect that the power of the tests used to accept these theories is rather small. Rich supplies in this paper new disconfirmation of the price-specie-flow theory. As with most other tests of this theory, he examines its apparent Achilles heel-the link between the balance of payments and the domestic money supply. Rich first criticizes Viner and then constructs a new data set to perform a better test. He argues that the Canadian money stock did not vary countercyclically. In fact it varied procyclically, but not because of the behavior of the monetary authorities and not because of what Rich calls the "Taussig effect"-induced capital flows that cause the balance of payments to move procyclically. And not, in fact, because the monetary base failed to move countercyclically. The money stock moved procyclically because anticyclical movements in the reserve ratio of the Canadian banking system more than offset the anticyclical pattern of the monetary base. Before discussing Rich's explanation for this phenomenon, it is worth pausing for a moment to look at the evidence, which, in keeping with much of the evidence of the paper, is curiously indirect. Rich first averages the rates of growth of the money stock in all of the booms and depressions from 1874 to 1913. This procedure fails to produce any significant difference between the booms and depressions. So Rich then averages the rates of growth of the money stock in the expansions and contractions of the economy. (A boom includes the period of high economic activity, both before and after a cyclical peak, while an expansion includes the period when the rate of change of economic activity was positive, from a cyclical trough to a cyclical peak.) Here he finds a pattern; the money stock grew faster in expansions than contractions. This pattern suggests, according to Rich, that the rate of growth of the

578

Georg Rich

money stock peaked near the middle of expansions. If the Canadian economy resembled a regular oscillator, the pattern implies that the level of the money stock, which should have followed the rate of growth with a quarter-cycle lag, peaked at the end of expansions, that is, at the cyclical peak. In other words, the level of the money stock was positively correlated with the level of economic activity. This conclusion gains only indirect support from the foregoing chain of reasoning. I do not know why Rich did not compute the correlation directly, and there are not .enough data in the paper for me to do so. Perhaps the book contains the data needed to verify his statement. Having laid the price-specie-flow theory to rest, Rich sets out to explain what was going on in,turn-of-the-century Canada. His model is only sketched out in this paper, but it appears to go like this: An "expansion of foreign activity," which appears to mean an expansion of Canadian exports, sets off the Canadian expansion. Canadian production rose" putting lip.ward pressure on the money supply. Interest rates rose, and banks reduced their reserve ratios both because they were less afraid of failure during a boom than during a depression and because they wanted to take advantage of the profit opportunities offered by the higher interest rates. The reduction in reserve ratios moderated the rise in interest rates, and Rich suggests that Canadian interest rates varied less over the cycle than comparable U.S. ones-a notion verified in Rich's book, although not in his paper. Since the data for replication of this test were available to me, I will comment on it shortly. But it is worth noting that this test is also indirect. Since the core of the theory is about Canadian reserve ratios, it would be nice to see data about them or comparisons between them and, say, U.S. ones. After the next step, a comparison of cyclical movements in the money stock in the two countries, the meaning of a smaller variance of the Canadian interest rates would be clear. Returning to the theory, Rich argues that the Canadian expansion led to a deficit in the current account, a pattern that seems odd in light of the role of export expansion in causing the expansion. Rich appears to be saying that the expansion, once started by an export rise, acquired a life of its own. He claims both that the relative stability of interest rates encouraged the expansion and that it discouraged capital imports, reducing the monetary base. The relative magnitudes of these effects is not clear, but the existence of the latter suggests caution in accepting Rich's principai conclusion that the Canadian banking system aggravated Canadian business cycles. Rich places great weight on the link between interest rates and capital imports in deriving the second test of his theory. He argues that peaks in interest rates followed p~aks in economic activity. Just after the peak, the

579

Canada without a Central Bank, 1872-1913

reduction in economic activity worked to lower interest rates, but the decrease in the monetary base resulting from the cyclical capital exports kept them up. Consequently, interest rates did not decline immediately after the peak. There are two problems with this test. First, since Rich argues that the effect of the fall in the reserve ratio on interest rates was stronger than the effect of the fall in the monetary base during the expansion, it is not clear why it was not stronger also during the early stages of the contraction. Second, again the test is curiously indirect. Rich reports that peaks in the first differences in the interest rate coincided with cyclical peaks in economic activity, so the peak in the interest rate itself must have followed it. A more direct test would have been preferable, and I will report the results of one here. Both of Rich's tests of his positive theory involve interest rates. And since these rates were available in the sources he cites, I replicated them. The rates involved are call-money rates in Montreal, New York, and Boston. I presume this rate was used because it was available, but I wonder how closely it mirrors overall monetary conditions in Canadian financial markets. As before, I can only presume that evidence on this question appears in Rich's longer work. The first proposition is the the Canadian interest rate varied less than the U.S. rate. And it did. The variance of the Montreal rate was 0.34; the New York rate, 7.67; and the Boston rate, 2.76. But while the result of this test is favorable to Rich's theory, examination of the raw data raises some uncomfortable questions. There are five occasions between 1900 and 1913, when the Montreal rate is available, that it was reported as constant for eight months or more. In four of these episodes the interest rate was an integer; one of the episodes lasted for nineteen months (October 1906, through April 1908). This constancy in a call-money rate is puzzling in the extreme. If the rate was accurately reported, then there is more to the Canadian institutions than Rich has allowed in his tale. And if the periods of constancy are the results of gaps in reporting, then the test is weaker than it appears. As an aside, it is worth noting that the New York and the Boston call-money rates were more highly correlated (p = .72) than either was with the Montreal rate. But since the Montreal and Boston rates were more highly correlated (p = .42) than the Montreal and New York rates (p = .26), it is hard to know if this was the effect of distance or of national boundaries. Given the constancy of the exchange rate between Canada and the United States, is an argument needed to justify analyzing Canadian monetary conditions separately from those in the United States? Rich's second proposition is that cyclical peaks in the Montreal callmoney rate lagged behind peaks in Canadian economic activity. Taking the peaks and troughs of economic activity from Rich's table 12.1 and the

580

Georg Rich

peaks and troughs of the interest rate from the rates as reported in Rich's sources yields the results shown in table C12.2. The peaks and troughs of reference cycles and of interest-rate cycles are shown using Rich's notation whereby the number after the slash is the month. Figures in parentheses are the duration of peaks and troughs when the interest rate was constant for several months at its extreme value. The lag of the interest rate behind the reference cycle is shown in months, where a negative number indicates a lead and a range indicates uncertainty about the interest-rate peak or trough. The data offer only limited support to Rich's hypothesis. The interest rate lagged behind general economic activity at the trough of the reference cycle, but it led as often as it lagged at the peak. Only in 1910 did the interest rate clearly peak after the economy. It may have done so as well in 1906-7, but the constancy of the interest rate for nineteen months around its peak precludes any firm conclusion. Considering the short lags at several of the peaks, it might be fair to say that the peaks were roughly contemporaneous at three of the four peaks where determination can be made, while the interest rate clearly lagged at the fourth. In short, the interest rate lagged behind economic activity at troughs, but not (in general) at peaks. Rich's story, therefore, stands as an intriguing hypothesis. It is hard to do justice to a monograph in a brief summary, and the problems of the tests used in this paper may well be solved in Rich's forthcoming book. I hope so for two reasons. First, the story is an appealing one. And second, it offers a "nonmonetary" theory for the lower-right-hand cell in my table C12.1. Rich clearly views capital movements as endogenous, and he appears to say that they move to establish a stock equilibrium. But, in Table C12.2

Timing of Troughs and Peaks in Canadian Interest Rates Compared with Timing of Business Cycle

Reference- InterestCycle Rate Peak Peak

Lead (-) or Lag (+) in Months at ReferenceReference Cycle Peak Trough

1900/04 1902/12

-3 -2-1

1901/02 1904/06

1901/09 1905/04-09

(6)

7 10-15

-2-16

1908/07

1909/02-09

(8)

7-14

9-11

1911/07

1912/06-08

(3)

11-13

Date

1906/12 1910/03 1912/11

1900/01 1902/101903/01 (4) 1906/10 1908/04 (19) 1910/121911/02 (3) 1912/01 (1)

2

Lead (-) or Lag (+) in Months at Reference Trough

Date InterestRate Trough

581

Canada without a Central Bank, 1872-1913

contrast to the monetary theory, the equilibrium is in the bond market, not the money market. That is, capital moves internationally in response to interest-rate differentials in much the same way as capital moves domestically between any two assets of slightly differing characteristics earning quite different returns. It moves until the holders of each asset are happy with the quantities they hold at market prices. Rich's story, therefore, is important both for Canadian history and for balance-ofpayments theory. Reference Ingram, James C. 1957. Growth in capacity and Canada's balance of payments. American Economic Review 47: 93-104.

General Discussion PIPPENGER expressed puzzlement regarding Rich's references to the transfer problem. If a country suddenly finds that four-fifths of its gold stock has disappeared, Pippenger argued, it can merely run balance-ofpayments surpluses until its gold stock is restored to desired levels. This action may entail some change in the. combination of traded and nontraded goods that the country produces, at least during the process of adjustment, but Pippenger saw no reason why this should create a problem, and hence no justification for the phrase "transfer problem." Several participants suggested that the problem potentially lay in the economy's inability to adjust the combination of traded and nontraded goods, or importables and exportables, that was required to facilitate the adjustment process. In this view, countries can experience a transfer problem when the excess demand for money entails a shift in demand away from one type of good and toward another, but markets do not function so as to smoothly effect the transfer of resources. MELTZER argued that an excess demand for commodities would be satisfied both by an excess supply of securities and a money flow. The excess supply of securities would raise interest rates, which would move securities and also cause a movement along the demand-for-money curve, releasing some of the excess supply of money to go along with the excess supply of securities. CAGAN prefaced his remarks with a caution to those who refer to "typical" cyclical patterns or "average" cyclical patterns. In fact, Cagan pointed out, cyclical patterns can vary greatly, and cyclical averages can be a very misleading indicator of the characteristics of a majority of cycles. Before generalizing, it is necessary to examine each cycle, or at least a large number of them, and to analyze the degree to which they

582

Georg Rich

conform to a typical pattern. On the other hand, the pattern Rich documents for Canada is typical also of the United States, and it would not be surprising if it were typical of many other countries as well. Cagan also commented on Rich's discovery that Canada's reserve ratio moves countercyclically: the reserve ratio declines during expansions, which would give rise to an increase in money supply, and rises in contractions. In his study of the U.S. money supply, Cagan did not find that changes in the reserve ratio were the main reason why the money supply fluctuated cyclically. The United States's money supply moved for a variety of reasons, the most important of which was the currencydeposit ratio. Cagan's study of U.S. interest rates showed that they tend to lag the business cycle, particularly at the trough. Thus, interest rates in Canada and the United States exhibit similar cyclical behavior. The parallel is not surprising .because U.S. and Canadian interest rates are connected by interest arbitrage. KOCHIN asked whether it was realistic to treat Montreal as another adjunct to the New York money market, much like St. Louis, San Francisco, or Philadelphia. He wondered whether interest rates in Canada and in the American West exhibited similar relationships to fluctuations in the New York rate. EICHENGREEN responded to Kochin's question by observing that there were in fact connections between money rates in New York and elsewhere in the United States. However, these connections were apparently quite different from those observed by Rich for Canada. For the first decade of the twentieth century, rates in St. Louis and similar midwestern centers typically fluctuated one to one-and-a-half months in advance of the New York rate. The explanation for these fluctuations is found in the harvest cycle: Interest rates in the agricultural regions rose when the demand for cash was stimulated by increasing economic activity during the planting season in the spring and the harvest season in the fall. Higher interest rates in the West attracted funds from the East, as southern and western banks liquidated their correspondent balances in New York in order to exploit the profit opportunities provided by interregional interest arbitrage. The outflow of funds from New York then raised interest rates in the East. It does not appear likely the relationship between interest-rate fluctuations in Montreal and in New York observed in Rich's paper on Canada reflects the impact on local money-market conditions of the agricultural cycle, for that explanation would suggest that Canadian rates should have led rather than lagged rates in New York. The lagging behavior of Canadian rates may therefore reflect the imperfectly competitive nature of the Canadian money market. A number of participants expressed reasons for skepticism concerning use of the call-loan rate in Montreal. For example, LINDERT noted that

583

Canada without a Central Bank, 1872-1913

the call rate was recorded as unchanging from October 1906 through April 1908, a period that spanned the 1907 financial panic in the United States. It is unlikely, therefore, that this series reflects the actual cost of funds. MOGGRIDGE noted that it made little difference which call-loan rate was considered, because the Canadian banking system was much less competitive than the U.S. system. Indeed, the Canadian system was oligopolistic, and the few banks in existence coordinated their administration of call-money rates. RICH acknowledged that the Canadian call-loan rate was probably not comparable with the New Yark call-loan rate because Canadian call loans seldom were called. He also discussed the problem of the constancy of the rate. In the contemporary literature, there is considerable discussion of price-fixing among oligopolistic banks, which may explain why the rate sometimes failed to move with changing market conditions. However, that literature fails to address the linkages between the U.S. and the Canadian markets. Those links were close enough that such price-fixing arrangements probably could not have survived for long. Rich insisted that the Canadian call-loan rate lagged the business cycle and the New York call-loan rate. He had considered the question of whether Canada was not simply a region within a unified North American financial market comparable to various regions in the United States, which is why he included the Boston call-loan rate. The Boston call-loan rate fails to lag the business cycle in the manner of the Montreal call-loan rate. However, rates on longer-term assets in Boston do begin to lag interest rates in New York. Rich also had attempted to consider other interest rates for Canada, but that had proven difficult because of data limitations. His preliminary conclusions were that federal-, provincial-, and municipal-government bond yields in Canada evince no cyclical relationship to yields in the United States. Rich had also considered a series for the average return on bank loans for one Canadian bank and compared it to estimates of the return on U.S. bank loans for various regions. All he could conclude was that the pattern of yields is extremely complex.

PART

v.

The Gold Standard as a Stabilizer of Commodity Prices

13

War, Prices, and Interest Rates: A Martial Solution to Gibson's Paradox Daniel K. Benjamin and Levis A. Kochin

13.1

Introduction

The positive correlation between the price level and the interest rate"Gibson's paradox"-is one of the best known and least understood of all economic regularities. Keynes called it "one of the most completely established empirical facts in the whole field of quantitative economics" (Keynes 1930, 2: p.198). Irving Fisher asserted that "no problem in economics has been more hotly debated" (Fisher 1930, p. 399). Even a casual glance at the literature (see, for example, Shiller and Siegel 1977, and the sources cited therein) suggests that Fisher's assertion is as true today as it was fifty years ago. And Keynes's contention seems amply supported by figure 13.1, which plots the yield on long-term bonds (consols) and the log of the price level in Britain from 1729 to 1931 (the gold standard years for which both series are available).1 The striking visual impression of a positive relationship between the price level and interest rates is confirmed by the following regression, which estimates the yield on consols (R) as a function of the log of the price level (with t-statistics in parentheses): (1)

+ 2.04 log Pt. ( - 8.83) (14.5) R 2 = .51, D.W. = .31. R t = -5.60

As beguiling as figure 13.1 and equation (1) are, both suffer from two serious defects. First, it is well known that many economic time series Daniel K. Benjamin is deputy assistant secretary for policy at the U.S. Department of Labor. Levis A. Kochin is associate professor of economics at the University of Washington, Seattle, Washington.

587

588

Daniel K. Benjamin and Levis A. Kochin 6.0...----------------- 6

5.5

5

5.0

4

4.5

3

4.0 .................I.-..............................--"-....L.......lo.-...-....L...-l...............a....-..............................--"-....L.......lo--........ 1750 1800 1700 1850 1900

Fig. 13.1

Yield on consols and the U.K. price level, 1729-1931.

exhibit nonstationarity, i.e., the series fail to revert to their means. As Granger and Newbold (1974) have shown, if one regresses one nonstationary series on another, the chances are high that one will observe a statistically significant relationship, even though both series are randomly generated and in fact unrelated to one another. We show that this problem is indeed present in the data for the price level and interest rates and that once the series are made stationary the apparent relationship largely disappears. The second defect of figure 13.1 and equation (1) is that prices and interest rates are both endogenous variables. The observed relationship between them might well reflect the effects of some common force that is acting on both. Figure 13.2 suggests what that force might be. Again the log of the price level and the consol rate are plotted for 1729-1931. We have added to the figure a measure of real defense expenditures, labeled WAR. The apparent association between WAR and either prices or interest rates is, to our eyes, at least as impressive as the apparent existence of Gibson's paradox. This apparent association is no accident: large increases in defense expenditures (principally during wars) create incentives to issue fiat money, which in turn tends to generate inflation and high price levels. Wars also create a scarcity of goods currently available for nonwar uses relative to the amounts that will be available in the (postwar) future. 2 The ensuing attempts by individuals to smooth their consumption over time create an excess demand for current goods and thus a rise in the interest rate. Once the nonstationarity of prices and interest rates and the effects of wars are accounted for, Gibson's paradox disappears.

589

War, Prices, and Interest Rates 6.0

r--------------------...,

5.5

5.0

4.5

4.0 L-..i.__'__...Io......II._..6.........I..-..a......-..I---'-......I..-"""--I..-....-L--~__'_~&......L.__'__...Io......II._..6......." 1900 1950 1750 1800 1850 1700 6~------~------------,3

5

2

4

o

3

2 l700

Fig. 13.2

·1 1750

1800

1850

1900

1950

Yield on consols, the U.K. price level, and real defense expenditures, 1729-1931.

13.2 Prices and Interest Rates Many explanations of Gibson's paradox have been proposed; perhaps the best known is the so-called Fisher effect. As first propounded by Irving Fisher (1930), this theory asserts that borrowers and lenders attempt to forecast the inflation rate that will prevail during the life of their debt contracts. Positive inflation rates imply reductions in the real value of debt obligations; hence lenders will demand and borrowers will accede to high nominal interest rates to offset that result. Negative inflation rates similarly will be associated with low nominal interest rates. The result is a positive association between interest rates and inflation

590

Daniel K. Benjamin and Levis A. Kochin

rates. Since inflation rates will be correlated with price levels, the result is an observed correlation between interest rates and the price level. Fisher (1930, chap. 19) found no evidence of a positive association between current interest rates and current inflation rates, nor between current interest rates and future inflation rates. However, he did find a positive association between current interest rates and past inflation. Arguing that past inflation works with a lag on future interest rates via its effects on expectations of future inflation, Fisher concluded that this observed relationship both vindicated his theory and explained Gibson's paradox. As Shiller and Siegel (1977) note, however, a distributed lag on past inflation rates must necessarily be well correlated with the current price level so that Fisher's finding may be nothing more than a restatement of the paradox. Shiller and Siegel go on to present evidence, assuming rational forecasting of inflation, suggesting that the Fisher effect is woefully inadequate in explaining movements in interest rates over the period 1729-1950. Our purpose in this section is to demonstrate the following: (1) during the gold standard period covered by Fisher, his postulated link between past inflation and current interest rates makes no sense; (2) during the full sample of gold standard years, to explain nominal interest rates or Gibson's paradox as resulting from changes in the expected inflation rate makes no sense; (3) there may, in fact, be no Gibson's paradox to explain. To address these issues, we begin by inquiring into the time-series properties of the price level. Table 13.1 shows the autocorrelations of the log of the price level in Britain from 1729 to 1931. The high, slowly decaying pattern revealed in table 13.1 (col. 1) is symptomatic of nonstationary series. Further evidence is contained in column (2), which displays the autocorrelations of the inflation rate (log Pt - log Pt - 1 ). Except at one lag (and at lag 21), none of the autocorrelations is as much as two standard errors from zero. The fact that the autocorrelation at one lag is nearly three standard errors from zero suggests that the inflation rate follows a first-order autoregressive process. However, the price-level data that we are compelled to use are annually averaged data-they reflect the average price level during each year. As Working (1960) has shown, time-averaged data on an underlying process that follows a continuous random walk will tend to produce positive serial correlation at one lag. To see the tendency, consider a rise in the price level during year t. The observed (time-averaged) data will show a rise in prices from period t - 1 to period t. The observed data will also show a rise in the price level from period t to period t + 1, since the observed price level for period t is an averge of the lower price level that prevailed during part of period t and the higher price level that prevailed during the remainder of the period. Working calculated the theoretical value of the autocorrela-

591

War, Prices, and Interest Rates

Table 13.1

Autocorrelations, U.K. Prices, Yields, and Abnormal Defense Expenditures, 1729-1931

Log Pr Order (1)

alog Pt (2)

Rt (3)

aRt (4)

WARt (5)

aWARt (6)

Standard Error of Random Model (7)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

0.206 -0.091 -0.099 -0.087 0.037 -0.060 0.013 0.083 0.050 -0.063 -0.040 -0.028 0.028 -0.081 -0.100 -0.026 0.127 0.072 -0.001 -0.006 -0.155 -0.065 -0.063 -0.043

0.936 0.870 0.823 0.764 0.707 0.666 0.625 0.589 0.563 0.522 0.484 0.461 0.455 0.445 0.435 0.418 0.388 0.346 0.313 0.263 0.214 0.192 0.156 0.112

0.007 -0.140 0.083 -0.009 -0.135 0.009 -0.036 -0.095 0.076 -0.029 -0.085 -0.123 0.021 0.011 0.093 0.125 0.101 -0.063 0.154 -0.009 -0.219 0.119 0.065 -0.139

0.894 0.724 0.548 0.365 0.213 0.111 0.035 -0.015 -0.028 -0.019 0.021 0.085 0.165 0.229 0.268 0.274 0.261 0.221 0.169 0.115 0.057 -0.002 -0.061 -0.116

0.301 0.029 0.026 -0.134 -0.240 -0.127 -0.121 -0.177 -0.104 -0.137 -0.105 -0.050 0.075 0.121 0.161 0.078 0.112 0.059 0.007 0.012 -0.002 0.004 -0.026 -0.086

0.070 0.069 0.069 0.069 0.069 0.069 0.069 0.068 0.068 0.068 0.068 0.068 0.068 0.067 0.067 0.067 0.067 0.067 0.066 0.066 0.066 0.066 0.066 0.066

0.953 0.886 0.826 0.773 0.728 0.679 0.635 0.591 0.538 0.482 0.429 0.379 0.335 0.289 0.251 0.221 0.193 0.153 0.108 0.063 0.019 -0.010 -0.033 -0.052

Sources: Col. (1), annual averages of wholesale prices from Mitchell and Deane 1962, pp. 468-70,474-75; col. (3), annual yield on annuities or consols from Homer 1963, tables 13, 19, and 57; col. (5), reported nominal defense expenditures from Mitchell and Deane 1962, pp. 389-91, 396-99. Notes: Col. (5) adjusted as described in note 6, deflated by prices. The residuals from a log linear trend over the period 1729-1931 were autocorrelated.

tion coefficient to be 0.25 at one lag and zero for lags longer than one period. The autocorrelations revealed in table 13.1 (col. 2) are remarkably close to the pattern predicted by Working. The autocorrelation coefficient at lag one is 0.21, well within one standard error of Working's theoretical value, and only one (at lag 21) of the remaining autocorrelation coefficients is as much as two standard errors from zero. We cannot reject the hypothesis that the price level followed a Martingale process in Britain during the gold standard years. If correct, the Martingale hypothesis implies that there is no evidence of persistent inflation or deflation in Britain during the gold standard years. The secular "trends" in the price

592

Daniel K. Benjamin and Levis A. Kochin

level that most investigators have found are the sorts of tricks of the eye that arise as a result of the human tendency to perceive order-even when none is present.3 The implications of this finding are striking. First, it implies that there should be no Fisher effect present during the years of the gold standard. If the price level follows a Martingale, then the best estimate of the price level in period t + 4 is the price level in period t, given available past information on the inflation rate; equivalently, the best estimate of the inflation rate in period t + i for all i is zero. Thus the "expected inflation rate" that will be rationally incorporated into nominal interest rates is also zero, given the past information on the inflation rate. Second, Fisher's finding that a distributed lag on past inflation rates is positively correlated with current interest rates is utterly unrelated to the Fisher effect. Given the time-series properties of the inflation rate, a distributed lag on past inflation rates is not the rational way to forecast future inflation rates. The distributed-lag method of forecasting will be rational only if inflation rates are positively serially correlated. Since they are not, no rational borrower or lender would use Fisher's proposed method of forecasting. The final issue involves the time-series properties of the long-term interest rate. Table 13.1 (col. 3) displays the autocorrelations of the levels of the consol rate. Again, the high, slowly decaying pattern of autocorrelation coefficients is suggestive of a nonstationary series. Table 13.1 (col. 4), which displays the autocorrelation coefficients of the first difference of the consol rate, reinforces this impression. Except at lag 21, none of the coefficients are as much as two standard errors from zero. Again, however, since these are time-averaged data, we should observe some positive autocorrelation at one lag. The fact that we do not suggests that the "true" underlying data probably reflect some negative autocorrelation-which would be expected if there were episodic "crises" in which interest rates rose and then fell sharply. The important point, however, is that the interest-rate series, like the price series, is very close to being a nonstationary series. The simplest way of dealing with two nonstationary series is to difference them to produce stationary series. We have done so for both the log of the price level and the consol rate, and then estimated the relationship between the two series. The results are as follows for 1729-1931 (with t-statistics in parentheses): (2)

dRt = .006 + .365 dlog Pr. (.33) (1.62) 2 R = .01, D.W. = 2.00.

At either the 5 percent level for a two-tailed test or the 10 percent level for

593

War, Prices, and Interest Rates

a one-tailed test, there is no significant relationship between the two series. The only paradox here is whether there is anything left to explain. 13.3 War and Interest Rates

General William Tecumseh Sherman informed us that war is hell. We have no fault with this description of the military attributes of war. But if one is interested in the economic attributes of war, it would be more appropriate to say that war is purgatory. Hell is permanent; war is temporary, and its salient economic effects arise from the temporary increase in government expenditures that occurs during war. The temporary rise in government expenditures during war is itself the product of the change in intertemporal demands that occurs at the outbreak of war. If a nation succeeds at war, the permanent incomes of its citizens will be higher than if it fails. Success demands that resources be available for use during the war: a rifle not produced until 1919 was of no use in fighting the battles of World War I. Thus during a war the demand for currently available goods rises relative to the demand for goods available in the (postwar) future. The change in intertemporal demands is converted into an intertemporal reallocation of resources via a rise in the real rate of interest. To see how temporary wartime expenditures produce a rise in the interest rate, we shall employ a simple two-period model of a closed economy in which the current period is war and the future period is postwar peace. Strictly for expositional convenience, we assume that all individuals are identical so that we can employ the device of a representative individual. The model is depicted in figure 13.3 where current goods are measured along the horizontal axis and future goods along the vertical axis. The economy's initial endowment is shown by the point (Eo, E 1) through which runs its productive transformation locus, LL. Absent the war and the representative individual will choose a productive (and consumptive) optimum at A where the indifference curve Uo is tangent to LL. The common values of the marginal rates of substitution and transformation equal - (1 + r), where r is the real interest rate. Current and future consumption are shown by (Yo, Yl) and investment (which equals savings) is shown by Eo - Yo. Consider now the onset of war, and assume that the war is financed wholly via current taxes. If these taxes equal Eo - Eo then the endowment available for nonwar activities becomes (Eo, E 1 ) and the new transformation locus for nonwar activities becomes LL', which is everywhere to the left of LL by the amount Eo - Eo, Le., by the amount of the current resources being used to fight the war. How will society respond to the lack of goods currently available for nonwar activities? Consider first point C, which involves holding current consumption unchanged and

594

Daniel K. Benjamin and Levis A. Kochin

L

y,

E,

L

Fig. 13.3

Current and future consumption in war and peace.

reducing investment to such a level that the burden of the war is borne solely out of future consumption. Since LL' is a horizontal transformation of LL, the slope of LL' at C must be steeper than the slope of LL at A. However if both current and future consumption are superior goods, the indifference curve passing through C must be flatter than the indifference curve passing through A, implying an excess demand for future goods and an excess supply of current goods. Hence the wartime optimum must be above C on LL'. Consider now point B, which would involve maintaining prewar levels of investment and future consumption, and thus absorbing the entirety of the rise in taxes out of an equal reduction in current consumption. Since LL' is simply a horizontal transformation of LL, the slope of LL' at B equals the slope of LL at A. However, given superiority of both current and future consumption, the slope of the indifference curve passing through B must be steeper than the slope of the indifference curve passing through A. Hence at B there would be an excess demand for current goods and an excess supply of future goods, implying a new optimum, such as at D, between points B and C. The new optimum thus involves reductions in both current consumption and investment. Since the optimum must lie below B, the marginal rate of transformation (and substitution) must be greater than before-the interest rate must rise.

595

War, Prices, and Interest Rates

Our assumption that current taxes are used to finance the war bears little relationship to the observed behavior of governments, which in fact finance wars largely through borrowing. However, the analysis developed for taxation applies exactly if domestic borrowing is the means of finance. It is ultimately neither taxation nor borrowing that "finances" the war: it is current goods that otherwise would be consumed or invested that finance the war. In the case of taxation the confiscation of current goods for war use produces an excess demand for current goods; as individuals attempt to borrow to finance current consumption, the interest rate rises until the excess demand is eliminated. In the case of deficit finance the government signals the excess demand by bidding up the interest rate via its borrowing activities. In either event, if Eo - Eo units of current goods are to be extracted from private use, their price must be bid up--the interest rate must rise. Several facets of the preceding analysis are worth emphasizing. First, it is the temporary nature of the war and the prospect of future peace that is central to the analysis. In the event of a permanent rise in "wasteful" government expenditures (such as defense expenditures), involving an equal reduction in both current and future goods available for private uses, a rise in the interest rate would be no more likely than a decline.4 Wars fundamentally involve a paucity of goods currently available for consumption and investment purposes relative to the amount of those goods that will be available in the future. The attempts by individuals to smooth their consumption over time creates an excess demand for current goods and the ensuing rise in the interest rate. Second, the relevant real-world interest rate is the one that applies between war and (postwar) peace. There is no reason for intrawar (e.g., very short-term) interest rates to rise, unless the probability is positive that the war will end before the short-term obligation comes due. To see this point, consider a simple world in which there are only two types of debt obligations: short-term and long-term. The short-term obligations are of annual duration and are issued (and repaid) in June of each year. The long-term obligations are consols, i.e., perpetuities with fixed coupon payments. Suppose that, absent war, both short- and long-term interest rates would be 5 percent per annum. Assume now that in June 1913 it becomes known with certainty that a war will begin in August 1914 and continue with uniform intensity until November 1918. Since there still will be peace in June 1914, the relative amounts of income available for nonwar uses in June of 1913 and 1914 are unchanged and the corresponding short rate of interest will be unchanged. In June 1914, shortterm rates will be substantially below 5 percent (and perhaps even negative), reflecting the· abundance of goods available then relative to (wartime) June 1915. Short-term obligations issued in June 1915 (and 1916 and 1917) will again bear 5 percent interest rates since the (depleted)

596

Daniel K. Benjamin and Levis A. Kochin

amounts of resources available for nonwar uses will be identical through June 1918. In the latter month, newly issued short-term obligations will bear sharply higher interest rates, reflecting the paucity of goods available then relative to (postwar) June 1919. Long-term rates will follow a slightly more complex pattern. The "announcement" in June 1913 of the future war will be accompanied by a slight decline in long-term rates since the depressed 1914-15 short rate implicit in the long rate is closer in time than is the elevated 1918-19 short rate. The long rate will continue to decline slowly until the onset of the war, at which point it will rise sharply (since it now only incorporates the elevated 1918-19 short rate). The long rate will rise slowly as the war progresses, reflecting the increasing importance of the advancing 191819 short rate.5 The long rate will peak just before the end of the war, droppping thereafter to the peacetime level of 5 percent. The third point to note is that opening the economy to allow for borrowing from abroad does not change the qualitative effects of our analysis unless the supply of funds from other nations is perfectly elastic. As long as the supply of foreign funds is positively sloped, the attempt by belligerent nations to borrow will generate an inflow of foreign capital (from neutrals) that will dampen but not eliminate the rise in the real interest rate. The dampening effect of borrowing from neutrals arises because such borrowing enables a belligerent nation to convert a purely temporary war into a partially permanent war. The final point worth emphasizing is that the wartime rise in interest rates is not a consequence of "Keynesian myopia" (the sufferers of which fail to perceive the future tax liabilities implied by current deficit finance). In fact, the Fisherian paradigm we have used to generate our implications corresponds in spirit most closely with "Ricardian omniscience," since our representative consumer clearly perceives the current and future implications of all actions. If the world is characterized by Keynesian myopia, the effects of war on the interest rate will be enhanced, since consumers, thinking their future incomes to be higher than they will in fact be, will attempt even more strenuously to maintain current consumption, thus producing a larger increase in the interest rate. One important implication is that attempts to distinguish between Keynesian and Ricardian views of the world must control for the effects of wars on interest rates. To test our theory, we have constructed a measure of "abnormal" defense expenditures; we refer to this series as WAR, subscripted to indicate whether it is contemporaneous (t) or precedes the current period by i periods (t - i). We began by summing together for each year British expenditures on army, navy, and ordnance and then deflating the sum by our measure of the price leve1.6 We refer to this deflated sum as real defense expenditures (D). We then estimated a log linear trend (log D =

597

War, Prices, and Interest Rates

+ ~t + E) over the sample period 1729-1931. The residuals from this regression (the deviations of real defense expenditures from their trend value) are the series we call WAR; its· values are shown graphically in figure 13.2 above and its autocorrelations in table 13.1 (col. 5).7 Our rationale for this measure is two-fold. First, if defense expenditures are really defensive, their long-run absolute level should be related to that which is being defended-the wealth of the country (see Thompson 1974). Using measured, current income as a proxy for wealth, the ratio of defense expenditures to income should thus tend to be a stationary (mean-reverting) series. We lack an annual time series on British national income prior to 1855, but know from other work (Benjamin and Kochin 1979, 1982) that the log of real income in Britain can be simply, albeit crudely, represented as a time trend (but see Nelson and Kang 1981, for precautions about placing too literal an interpretation on this representation). Hence our supposition that "normal" defense expenditures follow a log linear trend. Our second reason for using this measure is based on our notion that there are identifiable periods of abnormal and normal defense expenditures. When real expenditures are extraordinarily high or low, people expect them to return (eventually, if not immediately) to historically typical levels; it is this expectation that generates the forces on the interest rate that we have discussed. The use of deviations from the log linear trend of real defense expenditures captures this notion because the trend representation implicitly forces the deviations from trend to be mean reverting.8 Given our measure of abnormal defense expenditures, there remains the question of the correct method of estimating its effects on the interest rate. Our theory implies that the level of defense expenditures relative to their normal level should be positively related to the level of the interest rate. If the interest rate were a stationary variable, we could simply estimate R t = a + b WARt + Et • Given the nonstationarity of the interest rate, that course is inappropriate; statistical considerations dictate that we estimate the effect of war on changes in the interest rate. Fortunately, our theory implies that changes in the level of abnormal defense expen~i­ tures should be positively related to changes in the interest r(~.te. Estirri~t­ ing the relationship over the period 1729-1931 yields the following (with t-statistics in parentheses): a

(3)

.006 + 0.24 dWAR t • (.35) (3.51)

dR t

=

R2

= .06, D.W. = 2.11.

Two points are clear from this regression. First, our theory is consistent with the evidence. At a high level of confidence we can reject the

598

Daniel K. Benjamin and Levis A. Kochin

hypothesis that interest rates are independent of abnormal defense expenditures. Second, changes in the consol rate are largely determined by factors other than defense expenditures, at least as judged by the low R 2 of this regression. In part the low "explanatory. power" results from our use of the consol rate as a measure of "the" interest rate. Although the consol rate has the advantage of incorporating the implicit short rate between any future period of war and peace, it has the disadvantage of incorporating all other future short rates as well. In part our choice of the empirical counterpart to "the" interest rate suffers from the defect inherent in any interest rate measure: it is impossible ex post to measure the interest rate relevant between war and peace ex ante. Although equation (3) is a relatively poor predictor on average for the full sample period as a whole, it does surprisingly well when it matters most: periods of war. For example, from 1913 until the World War I peak in 1917, the consol rate rose 119 basis points; equation (3) predicts a change of 65 basis points. To take a more remote example, of the 104-basis-point rise in the consol rate from 1753 to 1761 (during the French and Indian War), equation (3) predicts a rise of 43 basis points. For the seven wars in our sample, equation (3) predicts between onethird and one-half of the movements (from either trough to peak or peak to trough) in the consol rate. The apparent predictive inadequacies of equation (3) are thus largely a result of its failure in peacetime-when we would not expect it to predict well in any event. 13.4 War and Prices

War and inflation have tended to occur together. During the two centuries covered by our data the highest price levels are observed during and immediately after the Napoleonic Wars and World War I. The highest inflation rates in our data are observed during and immediately after the same wars. The concurrence between war and inflation is no accident. The best-known cause of the link between war and inflation is the tendency for the governments of belligerent nations to issue fiat money during the course of the hostilities. The onset of a (less-than-fullyanticipated) war implies a rise in the present value of current and future government expenditures. If these expenditures are to be financed, the present value of current and future taxes must necessarily rise. If the marginal resource costs of taxation increase as the value of the tax collections increase relative to the value of what is being taxed, then efficient public finance requires that (1) both current and future taxes be raised to keep the marginal costs o~ taxation constant over time, and (2) taxes be raised "across the board" to keep the marginal costs of taxation equal across taxed entities. The issue of fiat money, and the ensuing

599

War, Prices, and Interest Rates

inflation it produces, provides the means of taxing (non-interest-bearing) money during wars. The incentive to use fiat-money issue as a means of taxation exists for all countries. But for countries on the gold standard there is an additional attraction to issuing fiat money. The gold reserves held by the government of a gold standard country can be thought of as an investment in brand-name capital, valuable to the extent that they induce confidence in the country's money. As discussed above, the onset of war produces an increase in the value of goods now relative to goods later. The ensuing rise in the rate of interest creates an incentive to convert capital goods into currently available goods. One means of doing so is by disinvesting in the stock of brand-name capital behind the belligerent's money supplyprinting fiat money, driving up the price level, and allowing gold reserves to flow abroad in search of lower-priced imports. Equivalently, one can think of the citizens of a belligerent nation as simply substituting paper for gold and shipping the latter to neutrals, thereby converting their confidence capital into imports of current goods and services that can be used to fight the war. There is a second reason for wartime inflations in belligerent nations, one that is less well recognized but one that also helps explain the substantial deviations from purchasing-power parity that are commonly observed in wartime.9 The onset of war implies that (1) current goods rise in value relative to future goods in belligerent nations and (2) currently available goods rise in value in belligerent nations relative to their value in neutral nations. The latter change in relative valuations implies that the real terms of trade should change so as to increase the real price of goods in belligerent nations relative to their real price in neutral nations. This provides the impetus for the large current-account deficit (excess of imports over exports of current goods) that enables belligerents to wage war. The rise in real interest rates implied by (1) provides the means by which the current-account deficit is financed by capital "exports," i.e., increased borrowing from abroad, for the relative rise in interest rates in belligerent nations attracts the foreign capital necessary to finance the current-account deficit. The rise in interest rates in belligerent nations is also the mechanism by which the relative prices in belligerent nations increase and the means by which the deviation from purchasing-power parity occurs. Absent the rise in belligerent interest rates, imports of goods could exceed exports only by the net outflow of gold. The rise in belligerent interest rates simultaneously permits (1) the direct link between a net inflow of goods and outflow of gold to be broken; (2) a rise in the relative price of goods in belligerent nations to occur; and (3) a deviation from purchasing-power parity that "favors" belligerent nations. The first of these events is a self-evident matter of accounting. The other two are more intricately related. Absent the rise in the interest rate

600

Daniel K. Benjamin and Levis A. Kochin

in belligerent nations, the relative price level in those countries would not deviate (far) from the price level in neutral countries, since the outflow of gold would equalize price levels quickly. A relative change in prices and the associated rise in imports of goods relative to exports of goods are made possible only by the rise in interest rates in belligerent nations, since it is this rise in interest rates that makes people willing to hold the belligerent's obligations. A concomitant effect is that the purchasingpower value of a belligerent currency rises relative to the value to be expected simply by looking at price levels because of the increased value of holdings of the belligerent's obligations. The positive association between war and inflation from 1729 to 1931 is graphically illustrated in the following simple regression (with t-statistics in parentheses): 10 (4)

~log

R2

+ .037 WARt-I. (.031) (4.15)

Pt = .002 =

.09, D.W. = 1.62.

Some may find our discussion of war and prices puzzling in a paper prepared for a conference on the gold standard. After all, the issue of fiat money at rates unrelated to gold holdings seems hardly descriptive of any gold standard rule. Yet war is not a recent invention. Looking ahead from any of the dates in our sample, few except the aged could have felt confident that war was ruled out in their lifetimes. Even in the aftermath of victory in the War to End All Wars the prospect of war remained. In 1919 Lloyd George told the service chiefs that they need not anticipate a major war within the next ten years. In 1925 the service chiefs asked again and were given the same answer: no major war within the next ten years. This answer was repeated in 1926 and 1927. Finally, in 1928, the service chiefs were told, on Churchill's prompting, that they need ask no more: the ten years' freedom from major war began automatically each morning. This instruction was revoked only in 1932. (Taylor 1965, p. 228) In our sample the yield on consols averaged 3.6 percent; thus on average, half of the present value was due to payments made beyond twenty years. The possibility of war and of a consequent departure from peacetime gold standard rules must surely have been a nontrivial consideration for the holders of consols. Moreover, much of the apparent price-level security of the gold standard derives from a tendency for past observers to focus on the success stories. It is true that Britain returned to gold at prewar parity after both the Napoleonic Wars and World War I, as did the Union after the Civil War and the United States after World War I. Yet when France returned to gold after World War I, she did so at a rate that was only one-fifth of the prewar level-a rate that surely would

601

War, Prices, and Interest Rates

have overjoyed the holders of Confederate currency in 1866 or Reichsmarks in 1923. The price-level security of a gold standard is in this respect no different from the price-level security of any monetary standard-it exists only so long as the government remains committed to a stable price level. 13.5 War, Prices, and Interest Rates

The results we have obtained thus far yield two conclusions: (1) the apparent association between the interest rate and the price level is largely spurious, and (2) prices and interest rates tend to move together due to the common influence of wars on both. The question we briefly address in this section is the obvious one: Is the combined effect of these two facts enough to eliminate the apparent association between the interest rate and the price level? As equation (4) and footnote 10 show, only the lagged value of abnormal defense expenditures (WARt-I) has an effect on the inflation rate. In equation (5) we hold the inflation rate constant. Thus it is necessary to decompose ~WARt into its components (WARt and WARt-I) so that their possibly differential effects on R, given that the inflation rate is held constant, can be taken into account. Doing so yields the following results (with t-statistics in parentheses): (5)

~Rt

= .005 + .29

~log

Pt + .28 WARt

(.29) (1.30)

(4.10)

- .21 WARt-I. (3.08) R2

= .10, D.W. = 2.18.

Once the effects of wars are accounted for, the weak relationship between prices and interest rates present in equation (2) disappears for all practical purposes. Gibson's paradox is the spurious product of war's effect on both prices and interest rates and not the result of any independent effects of prices on interest rates. ll 13.6 Conclusion

During the more than two centuries that Britain was (with two interruptions) on the gold standard, there is no evidence of persistent trends in either the price level or the consol interest rate. Both the price level and the yield on consols moved in what was essentially a random walk. The comovement of the price level and the level of the interest rate so apparent to the eye is largely visual spurious regression. In significant part the movements of both the interest rate and the price level have been produced by war. Once the influence of war is taken into account, there is

602

Daniel K. Benjamin and Levis A. Kochin

virtually no evidence of any linkage between the price level and the long-term interest rate.

Notes 1. Our interest-rate series is from Homer 1963, tables 13, 19, and 57. We use annuities until 1752 and consols thereafter except for 188{}-88 when the possibility of redemption at par made consol yields abnormally high. Following Homer's suggestion we use yields on 2.5 percent annuities for 188{}-88. The wholesale price index (Mitchell and Deane 1962, pp. 468-70, 474-75) links the Schumpeter-Gilboy (consumers' goods), Gayer, Rostow and Schwartz (domestic and imported commodities), and the Sauerbeck-Statist (overall) indexes. 2. See Hall 1980 for an alternative discussion of the effects of temporary defense expenditures on the interest rate. 3. We refer to the log of the price level as a Martingale rather than a random walk because a Martingale has serially uncorrelated changes but not necessarily random and independently distributed changes. Random changes would be random in either actual or absolute values. We find no evidence in the autocorrelation of price changes in Britain under the gold standard of systematic inflations or deflations, but there is systematic evidence of disturbed and quiet subperiods. The autocorrelations of the absolute value of changes in the log of the price level are significantly positively autocorrelated at lags out to six periods. This, together with the earlier results, says that when the price level changed under the gold standard there was no way of telling what direction the change would go in th~ next year. On the other hand, there was strong reason to anticipate that large changes would be followed by large changes and small changes by small changes. 4. Indeed, with homothetic preferences, the interest rate would remain unchanged. 5. The intrawar rises in the long rate will be further stimulated by the ongoing depreciation of the capital stock. 6. Reported nominal defense expenditures, from Mitchell and Deane (1962, pp. 38999), present several difficulties. First, during most of World War I and some of the Napoleonic Wars, the extraordinary military expenditures were "financed" with "votes of credit," with little or no detail shown in the records as to the disposition of the expenditure among army, navy, and ordnance. We added the votes of credit to the detail where shown and used the votes of credit alone where no detail at all is shown. The second problem is that the available defense-expenditure data are for fiscal rather than calendar years. Third, the dating of fiscal years changes four times during our sample. For 1727-54, the fiscal year ends 26 September; for 1755-99, it ends 10 October; for 1800-1854 it ends 5 January; in 1855, it ends 5 April; from then until the end of our sample, it ends 31 March. Finally, the series reports disbursements of funds rather than authorizations of expenditures. The principal economic effects of defense expenditures that we are concerned with would be expected to occur when the men and materials are ordered rather than when the cash disbursements happen to be made. We arbitrarily assumed that the lag between orders and disbursements is three months. Thus the disbursements reported for fiscal year 1740 (ending 26 September 1740) would reflect orders placed during the last months of 1739 and the first six months of 1740. On the further assumption that disbursements (and the accompanying orders) are uniform throughout a given fiscal year, the value of the orders placed during calendar year 1740 equal one half of the disbursements for fiscal year 1740 and one half of the disbursements for fiscal year 1741. We also assumed that (1) 26 September and 10 October are the same date (1 October); (2) 5 January is 1 January; and (3) 5 April is

603

War, Prices, and Interest Rates

31 March. Indexing calendar years by t and fiscal years by 7, we thus have the following algorithms for converting fiscal-year data into calendar-year data. For 1727-99, t = 1/2(7) + 1/2(7 + 1);

for 1800-1854, t = 3/4(7 + 1) + 1/4(7 + 2) ;

and for 1855-1931, t=

7

+ 1.

7. The serial correlations of LlWAR display-in addition to the Working autocorrelation at one lag-a significant tendency for negative autocorrelation at lags 4 to 10. There is a significant tendency for positive values of LlWAR to be followed by negative values in four to ten years. Wars (unlike "high" prices) tend to end. 8. An alternative to our method is to fit a time-series model to real defense expenditures and to use the difference between actual expenditures and forecasts of future expenditures as a measure of war. Two shortcomings of the alternative may be noted. First, what is the relevant future against which the present is to be compared? Second, the magnitude of defense spending in World War I is so large relative to the rest of the sample that the fitted model will be dominated by the few observations of those years, a problem that is less significant with the approach we chose. 9. See, for example, Friedman and Schwartz 1963, pp. 199-203. 10. Similar results are obtained by regressing the inflation rate on the current value of WAR. However, when both current and lagged values are included, only the coefficient of the lagged value is significant. These results suggest that it takes about a year for defense expenditures to have an impact on the inflation rate. 11. Even if wars are the complete explanation for Gibson's paradox, the price level may still appear to exert an independent influence on the interest rate in a regression such as equation (5). As noted earlier, the outbreak of a war involving Britain will drive up interest rates and prices in Britain. This will produce an outflow of gold in search of lower-priced goods from abroad and a search for borrowers abroad who are willing to lend at lower rates of interest. Both forces will tend to generate a rise in prices abroad as well as a rise in interest rates abroad. The same analysis holds in reverse for foreign wars: hostilities elsewhere will generate a rise in both the price level and the interest rate in Britain, holding the level of defense expenditures in Britain constant. We have fragmentary evidence of this force from America; during the Mexican-American War of 1846-47 and the Civil War, both the interest rate and the price level rose in Britain, even as the real level of British defense expenditures was declining. Controlling for foreign wars would of course require global information on real defense expenditures.

References Barro, Robert J. 1981. Output effects of government purchases. Journal of Political Economy 89 (Dec.): 1086-1122. Benjamin, Daniel K., and Levis A. Kochin. 1979. Searching for an explanation of unemployment in interwar Britain. Journal of Political Economy 87 (June): 441-78. - - - . 1982. Unemployment and unemployment benefits in twentiethcentury Britain. Journal of Political Economy 90 (Apr.): 410-36.

604

Daniel K. Benjamin and Levis A. Kochin

Cannan, Edwin. 1925. The paper pound of 1797-1821. London: P. S. King and Son. Fama, Eugene F. 1975. Short-term interest rates as predictors of inflation. American Economic Review 65 (June): 269-82. Feinstein, Charles. 1972. National income, expenditure, and output in the United Kingdom, 1855-1965. Cambridge: Cambridge University Press. Fisher, Irving. 1930. The theory of interest. New York: Macmillan. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Granger, C. W. J., and P. Newbold. 1974. Spurious regressions in econometrics. Journal of Econometrics 2 (July): 111-20. Hall, Robert E. 1980. Labor supply and aggregate fluctuations. Carnegie-Rochester Conference Series on Public Policy 12 (Spring): 7-33. Homer, Sidney. 1963. A history of interest rates. New Brunswick, N.J.: Rutgers University Press. Keynes, John M. 1930. A treatise on money. London: Macmillan. Klein, Benjamin. 1975. Our new monetary standard. Economic Inquiry 13 (Dec.): 461-84. Mitchell, Brian R., and Phyllis Deane. 1962. Abstract of British historical statistics. Cambridge: Cambridge University Press. Nelson, Charles R., and Heejoon Kang. 1981. Spurious periodicity in inappropriately detrended time series. Econometrica 49 (May): 74151. Shiller, Robert J., and Jeremy J. Siegel. 1977. The Gibson paradox and historical movements in real interest rates. Journal of Political Economy 85 (Oct.): 891-908. Taylor, A. J. P. 1965. English history, 1914-1945. Oxford: Clarendon Press. Thompson, Earl A. 1974. Taxation and national defense. Journal of Political Economy 82 (July/August): 755-82. Working, Holbrook. 1960. Note on the correlation of first differences of averages in a random chain. Econometrica 28 (Oct.): 916-18.

Comment

Phillip Cagan

Mr. Gibson's Paradox-Was It There? The Gibson paradox has played a long and notorious role in the history of monetary theory. It appears to defy the principles of classical monetary Phillip Cagan is professor of economics and chairman of the department at Columbia University, a research associate of the National Bureau of Economic Research, and adjunct and visiting scholar with the American Enterprise Institute.

605

War, Prices, and Interest Rates

theory. According to the quantity theory of money, long-run movements in the price level reflect corresponding movements in the stock of money. Since increases in the money supply also tend to ease financial markets, monetary increases that raise prices should also depress interest rates. The implication is that interest rates should be low when prices are rising and high when prices are falling. The Gibson paradox, that interest rates and prices move together in the same direction, appears to contradict the classical theory. I have argued elsewhere (1972) that insofar as monetary changes affect interest rates because of movements along the money-demand schedule, the so-called liquidity effect, such an effect will be transitory. If monetary changes have long-run effects on interest rates, they must reflect a saving of the revenue from money creation whereby the revenue is not treated as ordinary income and spent on consumption. That the revenue from money creation would be treated differently from ordinary income is open to question, however, so far as long-run movements are concerned. Therefore, conventional monetary theory does not necessarily imply that increases in the money stock reduce interest rates, and, while conventional theory cannot explain the paradox, they are not in opposition. Nevertheless, many antiquantity theorists delight in pointing to the paradox in support of their various anticlassical arguments. One of the earliest of these theorists, Thomas Tooke, who had a cost-of-production theory of the price level, argued that interest rates were a cost of production and had a positive effect on prices. Ricardo and Wicksell disposed of that explanation on the correct argument that interest rates influence relative prices, not the general price level, though Tooke's theory continues to be popular with the public and the press. Other explanations have never been in short supply. Gibson himself believed that the "cost of living" influenced saving, so that low prices increased the supply of loanable funds and lowered interest rates. That explanation is unsatisfactory, as was noted by Keynes (1930), who brought Mr. Gibson's articles to the notice of economists and enshrined his name on the paradox. Hicks (1950, p. 154n) proposed that short-term rates rose more sharply in the cyclical upswings when the secular growth in money is higher and presumably more variable; consequently, bond yields, which are an average of short-term rates, tend to remain higher when secular monetary growth is higher. Hawtrey (1913) had the traditional view that wages lag prices and do so by more, the faster prices rise; lower real wages raise profits and the real rate of interest. Macaulay (1938) claimed that rising prices are favorable to investment expenditures, an expansion of which increases borrowing and interest rates. More recently Shiller and Siegel (1977) have proposed that rising prices are unanticipated and shift wealth from lenders to borrowers, which reduces the net demand for financial assets and raises interest rates, and conversely for falling prices.

606

Daniel K. Benjamin and Levis A. Kochin

An influential explanation was proposed by Wicksell (1936) and independently by Keynes (1930). They posited secular fluctuations in investment demand. These fluctuations affected the supply of money-through the banking system for Wicksell and through the central bank's reserve ratio for Keynes. These fluctuations caused corresponding fluctuations in both interest rates and the money supply, and the latter in turn accounted for the behavior of prices. This explanation therefore relied on the quantity theory for the price effect. While velocity movements could conceivably finance part of investment fluctuations over the business cycle, an explanation of the long-run price movements in the Gibson paradox could not be attributed to velocity and required corresponding movements in the money supply. The fact that rising prices were accompanied by rising rather than higher interest rates could also be explained by supposing that the banking system or central bank responded to increases in investment by gradually expanding the money supply and raising interest rates. Each rise in interest rates was necessary to induce a further decline in reserve ratios and increase in the money supply. The fluctuations in investment demand were not explained but were taken as given, which was easy to accept in an age concerned with the instability of the private economy as symbolized by Kondratieff cycles, Schumpeter's creative destruction of capitalistic industry, and Keynes's animal spirits of businessmen. The Keynes-Wicksell theory is a plausible theory, and I have no objections to it except that it is inconsistent with the evidence, as I pointed out in my study of the U.S. money supply (1965). Their theory implies that secular movements in the money stock are due to the reserve ratio of banks or the central bank, whereas in fact such secular movements were due to high-powered money and growth of the gold stock in relation to total output. The gold stock is not related positively to prices or interest rates. If anything, the relation would be inverse. Higher prices and interest rates discourage gold production, and its positive association with prices reflects the quantity theory of money. The explanation for the Gibson paradox which is accepted by most monetary theorists today, as noted by Benjamin and Kochin, is Irving Fisher's-that nominal interest rates adjust to the rate of depreciation in the value of money. It has spawned a lively research industry to determine whether it fits the evidence. Benjamin and Kochin intend to spoil the fun by announcing, "Gentlemen, the object of your obsession does not exist!" which leaves nothing to explain except the obsession itself. They make the impressive argument that in the British data for the period 1729 to 1931, government defense expenditures account for the large fluctuations in prices and interest rates. These expenditures have the same effect as investment does for Keynes and Wicksell, but the source of the instability in the economy is cleverly shifted from the private

607

War, Prices, and Interest Rates

sector to the government. Benjamin and Kochin argue that defense expenditures will be partly financed by money creation and partly by borrowing and that the latter will raise interest rates. They go to great pains to argue that this is a "rational" response of the economy. According to the latest fashion, if it isn't rational it is not supposed to exist. But whether rational or not, I have no difficulty in accepting such an effect of defense expenditures as simply an historical fact. When these large fluctuations in interest and prices owing to defense expenditures are removed, not much worth mentioning remains. So far, so good. Then the authors put the Gibson paradox to the acid test by regressing the change in interest rates on the change in defense expenditures and the change in the log of prices. Amazingly, the paradox hangs on. The price variable is significant, but only marginally. In the version of the paper presented at the conference, their present equation (5) had only one WAR term and the t-statistic for dlog Pt was 1.97. In the present version of (5), WAR t - 1 has been added and the ton dlog Ptdrops to 1.3. Band K are relentless debunkers. At the conference they argued that the significant but anemic evidence of the paradox probably reflects foreign wars, since defense expenditures abroad would have repercussions on Britain similar to the domestic effect. In the present version of the paper they have found an equation form that reduces the significance of the paradox. I would be more sympathetic to the paradox. Their regression correlates concurrent changes in the interest rate and prices, putting a heavy empirical burden on the paradox and the Fisher explanation. This is so, because the concurrent changes may miss the long-run swings in these series and probably do not properly represent the expected changes in prices. I suspect one could shore up the significance of the Fisher effect with a lagged term or two. The authors justify the use of concurrent changes on the grounds that the level of interest rates and prices in the period tested display the property of random walks. This is not surprising since the cyclical movements are quite erratic, and the series display little overall trend from the beginning to the end of the period..The statistical problem with random walks is that they produce drift, and two such series can be correlated even if they are completely independent. However, the longer-run upward and downward movements that form the Gibson paradox number only three in the British data and essentially but two in the U.S. data from the mid-1800s to World War I. Even these few observations of the longer-run movements are partly obscured by the large short-run fluctuations due to wars and crises. Therefore, it is not surprising that when these fluctuations are not removed and the periods not specially selected, little evidence of the paradox or Fisher effect remains~ In a recent paper Lawrence Summers (1982) correlated the

608

Daniel K. Benjamin and Levis A. Kochin

lower frequency movements in U.S. interest rates and price changes, on the assumption that the Fisher effect pertained mainly to such longer-run movements, and found no consistent relation for a variety of periods. Contrary to the earlier more sympathetic search for the Fisher effect, the recent skeptical investigations conclude that there is no solid evidence of it. I count myself among the first to welcome this skeptical turnaround in econometrics. For a long time the journals have been filled, and unfortunately still are, with allegedly significant regressions that probably mean nothing. But now the new ARIMA techniques of purging the data of systematic movements to avoid spurious correlation are showing that it is difficult to demonstrate anything. That is probably as it should be. Yet when Benjamin and Kochin classify interest and prices under the gold standard as random walks and nothing more, I get uneasy. I am not sorry to see the Gibson paradox go, but I think the Fisher effect is getting the bum's rush. I do not wish to deprecate the long overdue revival of skeptical econometrics. Indeed, I welcome it and support it. But at the same time I would not forget about type-two error in the new zeal not to commit type one. Type-two error, as some may need reminding since it is so seldom mentioned in economics, is the rejection of hypotheses that are in fact true. I am not an expert, but I suspect that many of the new statistical tests that are rejecting very plausible relationships are weak tests, in the sense of having type-two errors. For one thing, regression equations impose a rigid timing relationship on data when in fact the relationship is probably variable. My eye may see random walks as trends, as Benjamin and Kochin warn, but by the same token my eye may be able to make allowances that are beyond the capability of dumb regression equations. My nostalgia for the old-fashioned qualitative analysis is not meant to be critical of Benjamin and Kochin, who have given us an excellent paper. I am simply reluctant to accept completely their well-reasoned conclusion. How far then might a case for the Fisher effect be carried? Let me start with the Cartesian proposition that the Fisher effect is not a figment of our imaginations. It really does exist! No one could be sure of this before the 1970s, but today there is no doubt. There is simply no way to account for the high nominal interest rates since the late 1960s, which have remained far above historical levels, without the Fisher effect. The only question is whether the effect would operate under the traditional gold standard. One might argue that it wouldn't. The gold standard, after all, gave a strong guarantee of a certain degree of price stability, seemingly eradicating expectations of persistent price changes. But prices were not in fact completely stable. U.S. wholesale prices, based on reference-cycle averages centered at reference peaks, fell from

609

War, Prices, and Interest Rates

the cycle in which the 1879 resumption occurred to the 1893-95 cycle over a decade later at a rate of 2.4 percent per year and rose thereafter to the 1907-10 cycle at a rate of 2.5 percent per year. It is a question relevant to the topic of this conference whether during these periods of the gold standard there was a change in expectations, first of declining prices and then of rising prices. It was not impossible, certainly, for expectations to change. No one at the time could be sure that the gold stock would grow just the right amount to produce perfect price stability. I think it is quite plausible that the price movements of the pre-World War I period, which though random from year to year persisted for about two decades first downward and then upward, would be recognized as trends whose duration was unknown. A gradual recognition would be rational if there was difficulty distinguishing permanent from transitory movements. It is likely that the corresponding movements in bond yields were not accidental but a response that would have come gradually, as it did.! I don't see any other explanation for these long-correlated movements. My eye rejects them as wholly accidental. There was no change in government expenditures of significance during these periods. Did the real rate of interest follow such a pattern? Doubtfully, since the possible causes of such a movement would not explain the gold and money-stock movements. No other equally attractive explanatio~ to Fisher's has been presented. The evidence is circumstantial and plausible but not, to be sure, definitive. Regressions will not support it unless we select the periods properly and abstract from the shorter-run movements and special periods. But so what? The Fisher effect under the gold standard is bound to be small and easily lost among stronger disturbances at work, so we must be especially on guard to avoid type-two errors. I would turn the traditional sequence of analysis around. Do not ask first whether the Gibson paradox was there, and then what caused it. Instead let us assert that the Fisher effect exists, therefore the Gibson paradox was there.

Notes 1. "Bond yields fell and rose during [the respective periods between reference cycle averages reported in the text] somewhat less: a fall of 1.4 percentage points in the first period (or 1.7 points allowing for the lagged upturn) and a rise of 0.1 in the second period (or 1.0 point from their trough in 1899-1902 to a prewar high for the 1913-18 reference cycle . . . ). Bond yields in money terms, therefore, seem to have accounted gradually and slowly for roughly half the average rate of initial appreciation and subsequent depreciation of money; and, the longer the movement of commodity prices in one direction, the larger was the adjustment" (Cagan 1965, p. 307).

610

Daniel K. Benjamin and Levis A. Kochin

References Cagan, Phillip. 1965. Determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. - - - . 1972. The channels of monetary effects on interest rates. New York: National Bureau of Economic Research. Hawtrey, R. G. 1913. Good and bad trade. London: Constable. Hicks, J. R. 1960. A contribution to the theory of the trade cycle. London: Oxford. Keynes, J. M. 1930. A treatise on money. New York: Harcourt, Brace. Macaulay, F. R. 1938. Some theoretical problems suggested by the movements of interest rates, bond yields, and stock prices in the United States since 1856. New York: National Bureau of Economic Research. Shiller, R. J., and Jeremy J. Siegel. 1977. The Gibson paradox and historical movements in real interest rates. Journal of Political Economy 85 (Oct.): 891-907. Summers, Lawrence H. 1982. The non-adjustment of nominal interest rates: A study of the Fisher effect. National Bureau working paper no. 836. Wicksell, Knut. 1936. Interest and prices. Translated from the German. London: Macmillan. (First published in 1898 as Geldzins und Giiterpreise.)

General Discussion KOCHIN started with a clarifying comment. He pointed out that both authors believe that the Fisher effect exists in the United States, Brazil, and Israel today. Nominal interest rates in Israel are different from those in Switzerland, and the difference is connected to the difference in inflation. But in Great Britain under the gold standard there was no rational basis to anticipate future inflation on the basis of past inflation, and thus no presumption that there should have existed a Fisher effect. In response to Barra, who recommended that Benjamin and Kochin use the levels of variables in at least some of their tests, Kochin cited the work of Granger and Newbold. Their work indicates that there is a 25- or 30-percent probability of observing a t-statistic suggesting a significant relationship when the variables are both random walks, if the levels of trended variables rather than first differences are used. MCCLOSKEY urged the authors to follow Barro's suggestion regarding causality statistics, but he argued against using those in Lewis F. Richardson's 1960 book, Statistics of Deadly Quarrels (Pittsburgh: Boxwood Press). He cited an even more up to date compilation by J. David Singer

611

War, Prices, and Interest Rates

and Melvin Small in a book entitled The Wages of War, 1816-1965: A Statistical Handbook (New York: John Wiley), published in 1972. McCloskey also urged the authors to undertake similar exercises for other countries and asked whether the authors' argument concerning temporary military expenditures also applies to other temporary expenditures. Will it apply to investment booms or to oil shocks? BENJAMIN agreed that it will, but emphasized the problem of measuring such temporary shocks in a systematic and credible way. FREEDMAN noted the authors' argument that the price level is a Martingale. Would this argument still apply if, when war begins, the rate of inflation accelerates suddenly, and everyone expects price levels to remain high and to continue to rise for awhile before coming back down to their normal levels in peacetime? If that is the case, one has to be very careful to take into account the existence of expected inflation during wartime. BENJAMIN suggested that the problem is the distinction between conditional and unconditional forecasts. His initial reaction had been much the same as Freedman's. The problem appears to be that once a war ended, residents could be reasonably confident, unless they lived in Germany, that the price level ultimately would fall. However, even in the British case, prices do not fall right after a war; instead, inflation continues, sometimes for very brief periods, sometimes for as long as twoand-a-half years. Thus, forecasters face two problems: When is the war going to end; and once it ends, is the price level really going to fall? Ultimately, it did fall in Britain, but that was not the case in many other countries. Residents could not predict with confidence what would happen in their country of residence. WEINTRAUB pointed out that everyone agrees that the Fisher effect has existed in recent years, after the demise of the gold standard. The question is whether it existed under the gold standard? Benjamin and Kochin conclude that it did not exist for the United Kingdom and Lawrence Summers found it did not exist for the United States. Irving Fisher, using less sophisticated techniques, also found it did not exist for the United States. It is not surprising that we fail to observe a Fisher effect under the gold standard, for prices were reasonably stable for long periods of time. Under a gold standard, we should fail to observe a Fisher effect since prices are not expected to change; however, that doesn't mean the Fisher effect doesn't exist. Interest rates could have been vastly different had prices been expected to change. MELTZER argued that an accurate reading of David Hume suggested that the lag between price changes and interest rates was highly variable-not constant, as Benjamin and Kochin had modeled it. Wars and other discontinuous events would show up quickly, whereas changes in the world money supply might show up more gradually.

612

Daniel K. Benjamin and Levis A. Kochin

MCGOULDRICK suggested that international comparisons embracing the 1850s and 1860s would strengthen the authors' case. In this earlier period, there was not only the Civil War in the United States but the Crimean War, two wars between Austria, Prussia, and Denmark, and finally the Franco-Prussian War in 1870. In comparison, after 1870 there were two decades of peace with no fighting except against the natives of the Third World. This development makes the rise in interest rates between 1896 and 1914, a period of general peace broken only by the Russo-Japanese War, somewhat puzzling. HARLEY expressed his disagreement with previous statements about what a reasonable expectation for price changes under a gold standard mught be. Reasonable expectations would have been based upon the economics of a nonrenewable resource, namely gold. Accordingly, the price of gold should have been expected to rise gradually over time, as stocks of other assets were augmented more rapidly than stocks of the exhaustible resource. Unanticipated gold discoveries would of course modify this story in important ways. KOCHIN drew attention to the distinction between ex post and ex ante interest rates and prices. While ex post one can discern trends in stock prices, surely few observers would dispute the efficiency of the stock market. That is, ex ante the best predictor of future stock prices is current stock prices. The same property holds for commodity prices under a gold standard. ZECHER inquired whether this notion should also be true for a period such as 1790-1815, when prices doubled? He recommended the use of indirect evidence to answer the question. For example, if one observes that people had formed political parties and organized in order to change the monetary system, as they did in the 1890s, this suggests they were expecting prices to fall, as prices in fact did. The free-silver movement provides a classic example of this phenomenon. WEINTRAUB argued that there is considerable evidence that during the last part of the nineteenth century, the public was frightened of what was perceived as the downward trend of prices. There is a considerable literature in which people describe their fears and thoughts. Similarly, after 1896 there was much discussion of upward trends in prices. Many observers would have bet at that time (1896-1916), early on rightly but later on wrongly, that over periods of five, ten, or more years the price level would have risen, as they would have bet between 1873 and 1893 that the price level would have fallen, again early on rightly but later on wrongly.

14

Some Evidence on the Real Price of Gold, Its Costs of Production, and Commodity Prices Hugh Rockoff

Under the classical gold standard the level of prices and economic activity depended to a considerable extent on the supply of new gold. This paper is about the stability and elasticity of that function. Periodically, the supply of gold was disturbed by a wide variety of factors: the discovery and subsequent exhaustion of gold fields, technological change in mining and extraction, government policies toward mining, and wars and revolutions. Below I survey these disturbances and answer some important questions about them. What kind of disturbance was the most important? How significant were the disturbances in terms of their effect on the world's stock of monetary gold? Perhaps of most interest is whether major disturbances were really lagged responses to changes in the real price of gold. In particular, to what extent can the great increase in the supply of gold at the turn of the century be more properly viewed as a movement along a long-run supply curve than as the product of an accidental, if fortunate, series of shifts in the supply curve? The notion that changes in the real price of gold influenced the subsequent supply of gold, and hence the general level of prices, is hardly new. Many references could be given, including Friedman and Schwartz (1963, p. 188) who noticed the broad trends in the relationship between increases in the real price of gold and increased supplies, Cagan (1965, pp. 60-67) who offered some qualitative evidence on the relationship, and Hugh Rockoff is professor of economics at Rutgers University, New Brunswick, New Jersey. The author would like to thank Hope Corman, John McDermott, Douglas Shaler, Marion Stewart, Richard Sylla, Eugene White, and Geoffrey Wood for their many helpful comments. He also learned a great deal from those who participated in seminars at Rutgers University and North Carolina State University. Rutgers University provided financial assistance. The responsibility for errors, of course, remains with the author.

613

614

Hugh Rockoff

Bordo (1981, p. 10) who showed that deviations from trend in the purchasing power of gold were significantly correlated with later deviations from trend in the world's supply of monetary gold. My intention here is to chart in detail the channels through which increases in the real price of gold influenced the supply. The exercise, I hope, will permit inferences to be made concerning the direction of causation with greater confidence that has hitherto been possible.! The motivation for exploring the supply of gold under the classical standard is clear. The gold standard is a device that depends on a privately produced supply of gold to replace deliberate governmental control of the stock of high-powered money. For the gold standard to provide superior monetary management, the disturbances produced by the factors mentioned above should be small and well timed in the sense that supply-augmenting disturbances should occur when commodity prices are falling or economic activity is at a low ebb. Before considering a return at the present time to the gold standard, we ought to know how significant disturbances were under the classical gold standard and whether conditions have changed in a way that makes such disturbances more or less likely. The paper is divided into six sections. Section 14.1 discusses the classical cost of production or "metallic" theory of the long-run price level under the gold standard. The section describes how the disturbances mentioned above would influence commodity prices under a gold standard and highlights the distinction between random disturbances and movements along a long-run supply curve. In the last part of the section I describe some ways in which the classical theory might be modified to incorporate the results of recent research on the economics of nonrenewable resources. Section 14.2 discusses actual fluctuations in the supply of gold under the classical gold standard and compares them with fluctuations in U.S. high-powered money in the post-World War II period. The surges in gold production during the middle and toward the end of the nineteenth century receive special attention. The latter surge, in addition, is also considered in detail in subsequent sections. According to section 14.2, it is fair to describe the fluctuations in the supply of gold under the classical standard as small and well timed. Section 14.3, which is concerned with the role of discoveries of new gold fields, begins the survey of disturbances to the supply of gold. The question explored is whether the great discoveries toward the turn of the century were endogenous, and some evidence is examined that suggests they were. Section 14.4 discusses the influence of technological changes and whether these changes were responses to changes in the real price of gold. The major concern in the section is with the origins and impact of the cyanide process for extracting gold from ore. This process is frequently cited as a disturbance that could have had great impact on

615

The Real Price of Gold

commodity prices. This section also deals with annual productivity change in gold production during the classical period to the extent possible, given limited data from which to ~onstruct quantitative estimates. Some judgments, however, can be made, and more recent evidence from South Africa is also presented. Section 14.5 then surveys disturbances emanating from governments. Both deliberate and inadvertent disturbances-the latter, by-products of wars and revolutions-are considered. It turns out that both types of disturbances were more important than has heretofore been recognized and that at least the deliberate policies, and possibly some of the seemingly inadvertent ones, were influenced by prevailing economic conditions. Section 14.6 summarizes the main conclusions. 14.1

The Cost of Production Theory of the Value of Money

The relationship between the cost of producing gold and the equilibrium price level under a gold standard was well understood by earlier generations of economists including Ricardo (1821, pp. 238--39), Mill (1871, pp. 499-506), Wicksell (1906, 2: pp. 146-53), Fisher (1922, pp. 99-104), and Marshall (1929, pp. 38-53, 282-84). The gold standard literature also contains rigorous modern treatments by Friedman (1953) and Barro (1979). What follows is a brief synthesis of this literature that highlights the significance of the issues discussed in this paper. Figure 14.1 displays the basic relationships. The vertical axis shows the real price of gold. This is the ratio of the nominal price of gold (Pg ) set by the government (the mint price) to the commodity price level (P). The horizontal axis shows the stock of monetary gold measured in physical units. At time 1 the real price of gold is determined by the intersection of the demand curve D and the short-run supply curve at time 1, SRSI. This equilibrium then determines the price level whether measured in ounces of gold (PIPg ) or, given the nominal price of gold, nominal units (P). The long-run supply of gold (LRS) , of course, then determines the analogous long-run magnitude. Before going further, several assumptions implicit in. the traditional analysis should be mentioned. The appropriate geographical unit is the set of countries that adheres to the gold standard. The horizontal axis then measures the stock of monetary gold across all those countries, and the vertical axis measures the average real price of gold, conversions from different monetary units being made at the exchange rates fixed by convertibility into gold. To explain the movement of gold among regions of one country on the gold standard, or the movement of gold between different countries on the standard, or changes in relative prices among different countries induced by monetary flows, the classical economists had recourse to some variant of Hume's price-specie-flow mechanism.

616

Hugh Rockoff

Pg p

Gold (oz)

Fig. 14.1

Supply of and demand for monetary gold.

That need not concern us here. It should also be noted that the figure depicts stock equilibria. The impact of the flow equilibria is crucial, but it is not depicted explicitly. Finally, the figure is drawn to focus attention on the determinants of the price level, pushing into the background concerns about the short-run relationship between prices and real output. The demand curve (D) in figure 14.1 is determined by three factors: the demand for money measured in nominal units, the ratio of the stock of gold measured in nominal units to the stock of money measured in nominal units, and the nominal price of gold. The demand for money in the classical theory was explained by the quantity theory of money. Thus, the D curve would shift to the right with increases in real income and to the left with increases in velocity. Again, because of the quantity theory, the demand for monetary gold slopes downward because as the price level measured in nominal units rises (and hence as the real value of gold falls given the nominal price of gold), the demand for nominal money balances rises. Given the nominal demand for money, the demand for monetary gold would tend to increase with an increase in the ratio of nominal gold stocks to nominal money stocks. Increases in the demand for monetary gold resulting from increases in the amount of money "backed" by a nominal unit of gold would raise the real value of gold and lower the nominal price level. The ratio of the nominal stock of gold to the nominal money stock was determined in turn by the behavior of the banking system with respect to reserves, the government's regulation of banks, and the public's preferences with respect to monetary assets.

617

The Real Price of Gold

Finally, changes in the nominal price of gold would be associated with inverse effects on the demand for monetary gold and direct effects on the nominal price level. Changes in ,the nominal price of gold, of course, were a form of monetary policy alien to the spirit of the classical gold standard and were rare during the classical period. The short-run supply curves of gold slope upward because of the existence of nonmonetary stocks of gold-the gold used for industrial and ornamental purposes. An increase in the real value of gold would discourage the use of gold for other purposes and increase the residual available for monetary purposes. Shifts in the demand for gold for nonmonetary purposes-due to changes in fashion, technological changes in gold-using industries, or changes in government policieswould produce opposite shifts in the short-run supply of monetary gold, with corresponding effects on the nominal price level. But in the absence of such nonmonetary demand shifts, the existence of nonmonetary stocks would cushion the economy against shifts in the demand for monetary gold. The foregoing analysis governed the attitude of economists toward the relative variability of the supply of and demand for money. Modern economists tend to regard the demand for money as stable and the supply as more variable. But the classical economists, given the tie of money to the stock of gold, given the existence of buffer stocks of nonmonetary gold, and given the adjustment response of the gold-mining industry (to be discussed shortly), tended to regard the supply of money as stable and , the demand for money as more variable. The equilibrium in figure 14.1 determined by the intersection of SRS1 and the D curve, however, would be temporary because the long-run supply curve of gold, LRS, is lower. With a real price of gold of (PgIP) 1 , gold mining would be highly profitable. New gold mines would be opened, lower-grade ores would be brought into production, and prospecting would be increased, with a possible outcome of new gold fields; other changes leading to more gold, discussed below, would occur. As a result, the short-run supply would shift to the right. Eventually, the short-run supply would shift as far as SRS2 and the long-run equilibrium value of gold and the price level would be reached at (PgIP)2. The period between the short-run equilibria would be one of commodity-price inflation. The gold-mining industry would perceive it as one of high but falling profits, as costs that were responsive to the general level of prices rose but the price of gold, fixed at Pg , did not. Similarly, a short-run equilibrium in which the real price of gold was below (PgIP)2 would give way to a period of decreased gold production and falling commodity prices. The long-run supply curve in figure 14.1 is drawn as perfectly elastic. In this special case the long-run price level is solely a function of the nominal price of gold, set by the government (but not regularly adjusted so long as

618

Hugh Rockoff

the government adheres to the spirit of the gold standard), and by the costs of producing gold. Changes in the demand for gold-due to changes in velocity, in real income, or even in the ratio of gold to money produced by the banking system-eould not affect the nominal price level in the long run. The determination of the price level in this case appears to be completely divorced from governmental de~ision-making processes. If there were some slope to the long-run supply curve, perhaps due to the exhaustion of high-grade ores, then variables influencing demand could influence the long-run equilibrium price level. To a considerable extent, however, the determination of the price level would have been taken out of the hands of government. The long-run classical equilibrium described in figure 14.1 could be disturbed by a variety of factors that are usefully divided into three categories. 1. Discoveries ofnew gold supplies. The discovery of an unusually rich source of ore, for example, might be regarded as a downwardshift of the long-run supply curve. But to the extent that new discoveries are regarded as an economic response to a particular value of gold, the discoveries could be treated, alternatively, as movements along a "long-run supply. "2 2. Technological progress in the production ofgold. The adoption of a new process, for example, the cyanide process for extracting gold from ore, would lower the long-run supply curve. In a growing economy one would want to look at the rate of technological progress in gold mining relative to the overall rate of technological progress. Again, if technological progress were endogenous, one could regard the resulting output changes as movements along LRS. 3. Governmental policies influencing the supply of gold. Gold mining, like other industries, maybe influenced in a host of ways by governmental policies designed to influence costs and output. Such policies, of course, have monetary consequences. Through this channel governments can reclaim some of the influence over economic activity that the gold standard appears on the surface to deny. The classical theorists, as suggested by the perfectly elastic LRS curve, did not dwell on the fact that gold was a nonrenewable resource. The reason is probably that the nineteenth century, taken as a whole, was one of increasing supplies, with important new finds in many parts of the world. In other words, over that century gold in fact had been a renewable resource. For the twentieth century, however, models that treat gold as a nonrenewable and durable resource may be appropriate. A full exploration of the ways in which such models might be used to refine the classical theory of the long-run price level under a gold standard is beyond the scope of this paper. But a brief look at two of the key propositions from that literature suggests some possible lines of inquiry.

619

The Real Price of Gold

First, Hotelling's rule-that the rent on the unexploited reserves of a nonrenewable resource owned by a competitive industry will rise over time at the real rate of interest-has been shown to apply to a durable resource such as gold as well as resources that are consumed as they are used (Stewart 1980; Levhari and Pindyck 1981). There seems to be no reason why the proposition should not apply to the monetary commodity as well as to durable resources produced for other purposes. The clear implication is that under a gold standard the behavior of the commodity price level will be described by Hotelling's rule. Hotelling's rule does not mean, however, that the real price of gold will always be rising and that under a gold standard with a fixed nominal price of gold, the nominal price level will always be falling. Decreases in extraction costs at the margin would permit Hotelling's rule to be satisfied even with a falling real price of gold. In fact, as Levhari and Pindyck stressed, the price profiles of many resources seem to have been U shaped, with a period in which the real price fell followed by a period in which it rose. Levhari and Pindyck present an ingenious explanation of a U-shaped profile, one that does not depend on new supplies of low-cost ores. Nevertheless, in the case of gold, it appears that the U-shaped patterns were due to the discovery of fresh supplies. Consider for example, the broadly U-shaped pattern that prevailed from the 1850s until the late 1890s. During the period following the discoveries in California and Australia, the real price of gold fell as it followed the decline in extraction costs. Later, as the rich alluvial deposits were exhausted and as the ore in the mines exhibited the common tendency of gold ore to decline in quality with depth, marginal extraction costs rose, and with them the real price of gold. The process then began again with the discoveries in South Africa and elsewhere. A second implication of the renewable-resource approach is that the classical gold standard could have generated an optimum quantity of money. This follows from the demonstration in the literature that if the industry that produces the durable resource is competitive-as gold mining undoubtedly was during the classical period-then the resource will be extracted at the optimal rate in the sense that the discounted sum of producer's and consumer's surplus will be maximized. This procedure would have been second best because a perfectly managed fiat standard could have produced a higher level of welfare in the community. Resources would be saved in the production of money. But the point is that under the gold standard, market forces-one of them being the profitmaximizing behavior of the gold miners-could generate the optimum. Under a fiat standard, as we well know, there are no corresponding assurances that the optimal monetary policy would be followed. An analogy may be drawn between these propositions concerning the gold standard and Friedman's (1969) discussion of the opt~mum quantity

620

Hugh Rockoff

of money under a fiat standard. Drawing out this analogy will serve to clarify and perhaps reinforce the main points. In that discussion Friedman argued that the optimal monetary policy would reduce the marginal nonpecuniary service flow from money to zero since the marginal social cost of producing fiat money was zero. Since each wealthholder would equate the real return on money (the marginal service flow less the rate of inflation) with the real rate of interest, the optimal policy for a fiat standard implied a price level falling with the real rate of interest. Under a gold standard, by way of contrast, the marginal social cost of producing money would be positive. With constant extraction costs at the optimal intensity of extraction, and neglecting the other costs of production, the marginal social cost would simply be the product of this extraction cost and the rate of interest (to convert to a flow). This social cost in conjunction with the wealthholder's equilibrium implies a nominal price level falling at the real rate of interest weighted by one minus the real cost of extraction. This result, arrived at by extending Friedman's logic, is precisely what Hotelling's rule tells us will result from market forces if gold is produced competitively. Or to look at the matter slightly differently, we can view Friedman's rule as a special case of the general rule for a metallic standard, the case in which the costs of extracting, refining, and minting the monetary commodity are zero. These results are merely a part of a full analysis. For example, if marginal extraction costs are not constant, the concept of the social cost of adding to the money stock would have to be modified. A full analysis would also examine the case in which gold was produced by a monopolist. The analogous case in the theory of fiat money in which the government is typically assumed to maximize the discounted revenues from creating money has been explored by Friedman (1971) and Calvo (1978) among others. While the case of a monopolistic producer of gold is not realistic for the nineteenth century, it may be relevant, at least as a polar case, for some future period in which we return to gold. Perhaps the examination of both the competitive and monopolistic cases will lead to a theory of the optimal monetary constitution that examines the costs and benefits of alternative monetary arrangements under alternative assumptions about how those arrangements are managed-a theory that contrasts a revenue-maximizing producer of fiat money, for example, with a gold standard in which gold is produced competitively. These speculations, however, must be left for future research.3 14.2 The Growth of the World's Stock of Monetary Gold

Before proceeding to the sources of disturbances to the supply of gold, it is worth examining their cumulative impact. Were the disturbances of

621

The Real Price of Gold

such magnitude and timing that the growth of the world's monetary gold stock was slower, more stable, and more elastic in the sense of better calculated to maintain price stability than what has been observed of comparable variables under favorable modern circumstances? Table 14.1 provides a tentative answer. It compares annual rates of growth of the world's monetary gold stock, 1807-1929, with rates of growth of the U.S. monetary base, 1949-79. This is a fair comparison for our purposes. The latter period, although punctuated by two wars, was free from the much greater disturbances produced by the world wars and the Great Depression. The U.S. monetary authorities, moreover, appear to have been as concerned with long-run price stability as managers of a fiat standard are likely to be in modern circumstances. Before discussing the two periods, however, a word on the sources of the data is in order. The world's stock of monetary gold is from Kitchin (1930). He constructed his estimates by assuming a base stock and then adding to it each year the difference between an estimate of world production and an estimate of the gold used for industrial and certain other nonmonetary purposes. The derivation may have contributed to stability of the final numbers. Kitchin's estimates were criticized by several scholars whose work is summarized by Hardy (1936, pp. 205-7). Table 14.1

A Comparison of Changes in the' World's Stock of Monetary Gold, 1807-1929, with Changes in the U.S. Monetary Base, 1949-79 Average Annual Percentage Rate of Change (1) World's Stock of Monetary Gold 1807-1839 0.63 1839-1849 1.16 1849-1859 6.39 2.64 1859-1869 1869-1879 1.63 1879-1889 1.07 1889-1899 2.98 1899-1909 3.79 1909-1919 3.17 1919-1929 1.95 U .S. Monetary Base 1.57 1949-1959 4.39 1959-1969 7.29 1969-1979

Standard Deviation of Annual Percentage Changes (2)

Coefficient of Variation (2) -7- (1) (3)

n.a. 1.27 1.95 0.64 0.48 0.42 0.93 0.68 1.12 0.68

n.a. 109.48 30.52 24.24 29.45 39.25 31.21 17.94 35.33 34.87

2.22 1.98 1.28

141.40 45.10 17.56

Sources: World stock of monetary gold from Kitchin 1930, table B, col. 1, pp. 82-84. U.S. monetary base from Tatom 1980, table 3, June dates, p. 23.

622

Hugh Rockoff

The criticisms rely on an alternative method apparently first used by Edie (1929) who constructed estimates of the world's stock of monetary gold for 1912 and 1928 from estimates of monetary holdings of central banks and gold in circulation. Edie's figures differed from Kitchin's, showing that Kitchin had underestimated world gold stocks by about 13 percent in 1913 and had overestimated them by about 3 percent in 1928, a difference that materially affected the growth rate between those years. But, as Kitchin (1931, pp. 69-71) argues, Edie's estimates were not themselves free of the sort of fill-in-the-blanks that make Kitchin's estimates uncertain. In any case, Kitchin's estimates seem to be the best long-run series now available. Clearly, revised figures would be useful. The U.S. monetary base-the St. Louis Federal Reserve's estimatetakes into account concurrent changes in reserve requirements that offset or reinforce the effects of actual changes in the monetary base on the money supply. Tatom (1980) describes the rationale and procedures for estimating the series. It appears to be the correct measure for comparison, if one has in mind a return to the gold standard in which the monetary base behaves as it did under the classical gold standard, and governments take no actions to alter the ratio of money stock to the monetary base. On the whole, the growth rate of the world's stock of monetary gold appears to have been slower and more stable in the classical period than in the modern period. The growth rate of the U.S. monetary base during the past decade was more rapid than in any of the decades listed for the gold standard. During the 1960s the growth rate of the monetary base was more rapid than in any decade except the 1850s when the great flows from California and Australia were entering the world economy. The comparison includes, it should be noted, the growth of the world's gold stock in the decade following the turn of the century when the flows of gold from South Africa, Western Australia, and other regions were making themselves felt. The year-to-year standard deviations of growth rates also appear to have been greater in modern times than under the gold standard, again with the exception of the 1850s. If the coefficient of variation is used as a measure of year-to-year stability, the overall comparison also favors the gold standard, although in this case there is an important exception-the 1970s were a period of high but stable growth by this measure. The growth of the gold stock was also elastic in the sense that low rates of growth were followed by surges that brought the average up to a long-run level consistent with a stable real value of gold. There were two major surges. The world's gold stock grew slowly in the period 1800 to 1850 and the real price of gold (based on U.K. Prices) rose at an annual rate of 1.3 percent per year (Jastram 1977, pp. 34-37). Growth of the gold stock then accelerated in the 1850s and 1860s, producing a fall in the real

623

The Real Price of Gold

price of gold at an annual rate of - 1.1 percent per year from 1850 to 1870. Growth of the gold stock decelerated in the 1870s and 1880s, producing an increase in the real price of gold at an annual rate of 1.4 percent from 1870 to 1890. The second surge in world gold production followed, bringing with it a fall in the real price of gold at an annual rate of - 0.4 percent from 1890 to 1910. Over the long run, from 1800 to 1910, the real price of gold was extraordinarily stable rising at 0.6 percent per year. The question is whether these surges should be regarded as fortunate random events that helped preserve price stability (and possibly the gold standard itself) or as movements along an elastic long-run supply curve. 14.3 Discoveries of New Gold The proximate causes of the major surges in the supply of gold during the classical period were a small number of great discoveries. These were the discoveries in Siberia (1814,1829), California (1848), eastern Australia (1851), western Australia (1889), and South Africa (1886). Table 14.2 arranges the production data in a way that illustrates the point. It shows the "concentration ratio," the share of the four top countries, in total output, by decade from 1801-10 to 1921-30. It also shows the share of the four countries containing the new gold fields referred to above. The dominance of these countries once their major fields were opened is clear. The data underlying table 14.2 (Ridgway 1929) appear to be the most comprehensive accounting available, but subject to a wide margin of uncertainty, particularly for the period before 1850 and for less developed countries. Here, too, a good deal of fill-in-the-blanks was required in order to arrive at comprehensive estimates. Ridgway, for example, shows exactly the same output for Africa in the three decades from 1801 to 1830 to six significant figures. The largest source of error, perhaps, was the undercounting of gold that was removed and sold secretly to avoid government taxes or other restrictions (deLaunay 1908, pp. 157-66). Table 14.2 shows that the long-run supply of gold can be described in terms of three phases. During the first phase, 1801-48, the supply of gold was dominated by South America (not shown) and the Soviet Union. Over the period 1801-40 it averaged about 0.5 million ounces per year. Then, with the discoveries in California and Australia, a shift took place to a new plateau. During the second phase, 1851-90, production averaged about 5.9 million ounces per year. Finally in the 1880s and 1890s, further discoveries, particularly in Australia and South Africa, produced a second shift in the supply of gold. From 1901 to 1930 world production averaged about 19.0 million ounces per year. Another way of viewing the dominant role of the great discoveries is by

585 382 469 657 1,712 6,456 6,110 5,658 5,239 10,161 18,381 20,639 18,012b

Annual World Production (thousands of fine ounces) (1) 74.9 66.0 68.5 73.3 83.4 93.7 90.8 87.6 77.5 76.5 74.0 74.7 75.6

Top Four (2)

0.1 2.2 19.1 26.0 42.2 50.4

-

South Africa a (4)

37.9 30.8 25.8 22.6 21.3 18.2 8.5 3.6

-

Australia (5) 0.9 2.7 23.2 34.5 42.3 12.8 14.2 21.5 20.7 11.6 6.8 4.7 3.5

Soviet Union (6)

3.2 6.2 26.4 39.3 72.5 92.0 82.5 81.2 75.7 76.5 74.0 74.7 70.3

(3)+(4)+ (5) + (6) (7)

aIncludes British South Africa and the Transvaal before 1901. Includes Bechuanaland and Swaziland after 1901. bRidgway's data only go through 1927, so his figures for the first seven years were multiplied by 10/7 to make this figure a decadal total comparable to the others.

2.3 3.5 3.2 4.8 30.2 41.3 37.5 33.8 30.2 24.5 23.0 19.3 12.8

United States (3)

Percentage of Annual World Production

Concentration of Gold Production under the Classical Gold Standard

Source: Ridgway 1929, table 58, an unnumbered appendix.

1801-1810 1811-1820 1821-1830 1831-1840 1841-1850 1851-1860 1861-1870 1871-1880 1881-1890 1891-1900 1901-1910 1911-1920 1921-1930

Table 14.2

625

The Real Price of Gold

asking how long a country was able to remain among the top four once it entered the list. Leadership, it turns out, typically was held for a long time. In the early decades of the century, Latin American countries, notably Colombia, Chile, and Brazil, were the leading producers. Colombia led the list in the first and second decades of the nineteenth century and finally dropped from the top four in the 1850s. What is now the Soviet Union led in the 1830s and 1840s and did not drop from the list (temporarily) until the 1920s. The United States entered the list at the top in the 1850s as a result of the discoveries in California. The United States then held first place until overtaken by South Africa in the first decade of the twentieth century. Table 14.2 also suggests that the supply of gold was potentially vulnerable to political shocks. In every decade the four leading producers mined two-thirds or more of total gold, and in slightly more than half the decades, over three-quarters of total gold. In the period from the California discoveries until the 1890s, the United States typically produced over 30 percent of the world's new gold. Thus, it was possible for policies or struggles for power that influenced supply in one country, to influence the world's supply. In short, while minor discoveries, changes in costs, and technological advances all impinged on the growth rate of the stock of monetary gold-these influences will be discussed in subsequent sections-the dominant role belongs to the great discoveries. The question naturally arises whether in the long run discoveries, particularly those that occurred when the real price of gold was high, should be regarded as endogenous. Some evidence is provided by the record of the local circumstances leading up to the discoveries. In one important case, California, the discovery seems to have been the accidental by-product of the expansion of agriculture into the interior of the state. The Spaniards, who had long resided in California, had taken gold from their Latin American possessions and even from California itself, but had not prospected in the interior. Instead, gold seems to have been discovered purely by accident near Sacramento, according to the standard account, in a stream where a dam was being built (Paul 1947, pp. 36-37). Even here, however, the discovery was not without antecedents. Gold had been discovered in the Los Angeles area in 1842 and worked for some years (Morrell 1940, p. 77). Had the real price of gold been lower in the 1840s and this find not worked successfully, the gold consciousness of the men who made the great discovery in 1848 might have been lower. It is also possible that had miners not arrived on the scene who knew how to work alluvial deposits, the initial gold rush would have petered out (Dane 1935, p. xiv). But most accounts suggest that in the absence of these factors, the California discoveries would have been delayed but not foreclosed.

626

Hugh Rockoff

In other cases, prospecting, partly in response to changes in the real value of gold, played an important role in the great discoveries. The California demonstration that unworked rich alluvial deposits could be found in newly settled territories led directly to an increase in exploration and to the quick discovery of gold in Australia and New Zealand. Blainey (1970) showed that later discoveries in Australia, including the rich western fields, were made in response to economic conditions. He emphasizes the level of unemployment rather than the real price of gold; in his view the discoveries were typically made by men who had moved from other employments into prospecting in periods of unemployment. Under a gold standard, of course, the real value of gold tended to rise during depressions when prices were low and unemployment high, so these discoveries could still be an element of the equilibrating mechanism. The California discoveries also led to prospecting in South Africa where some gold was found in the 1850s, but there were no major discoveries until 1874. That find led to an intense period of exploration culminating in the discovery of the main reef of the Rand in 1886 (Gray 1937, p. 30). Increased production in South Africa contributed about 35.3 percent of the increase in world production between 1890 and 1905. It may be conjectured that if the real value of gold had been declining during the long period of exploration, prospecting would have been curtailed, and the main reefs might have been missed, at least for some years. The U.S. finds that contributed about 21 percent of the increase in world gold production between 1890 and 1905 were the outcome of intense pro·specting. About 39 percent of that contribution was due to increased production in Colorado, much of it from the great Cripple Creek lode. The lode, found by a persistent prospector in a region that had been actively worked for thirty years, had an ore deposit that was somewhat different from what miners were familiar with and, possibly for this reason, had been overlooked (Morrell 1940, pp. 164-65). To be sure, silver prices were low, and that gave special urgency to the search for gold in Colorado, but the search might have concentrated on other minerals had the real price of gold been lower. The second largest contribution to increased U.S. output during this period-26 percent-eame from Alaska. Here, also, the discovery at Cape Nome, which led to the last of the great gold rushes, was made by a group of prospectors exploring that remote region (Morrell 1940, p. 402). Again, it may be conjectured that without the incentive provided by a high real price of gold, these major discoveries might have been missed. These cases appear to be typical. Crant; (1908) examined the origins of a large number of American gold and silver mines, distinguishing between those that were found by prospecting and those found accidentally.

627

The Real Price of Gold

Inclusion on his prospecting list depended on whether the mining journals reported the relevant information; that list appears to contain virtually every important field. Table 14.3 is constructed from Crane's data. Several conclusions emerge. First, the vast majority of finds were located by prospecting. Second, while the level of finds did not fluctuate synchronously with the real price of gold, there is a distinct fillip in the 1890s when the real price of gold was high. The evidence is not conclusive, but it suggests that many discoveries, particularly those that influenced world gold production between 1890 and 1905, should be regarded as movements along a long-run supply curve rather than as shifts in the curve. The suggestion, however, raises an important question. Were conditions in the nineteenth century unique so that a flow of new finds was possible, or are similar results to be anticipated in any period in which gold is the monetary standard? In some ways the nineteenth century appears to have been unique. European exploitation of natural resources was expanding into regions in which individual prospectors were encouraged to search for rich alluvial supplies. In Siberia as well as in the United States, Australia, and South Africa a core of dedicated prospectors made most of the finds. The conditions that produced that core of prospectors and the virgin territories in which they worked cannot be recreated. Some set of institutional arrangements would need to be substituted for the possibility of rich, easily worked, alluvial deposits. 14.4 Technological Change in Gold Mining

Various technological innovations affecting industry generally during the classical period were quickly adapted to gold mining, at least in the United States. Several large California mines are said to have been the

Table 14.3

Gold Mines in the United States by Mode of Discovery, 1821-1905

Accidental Prospecting 1821-1830 1831-1840 1841-1850 1851-1860 1861-1870 1871-1880 1881-1890 1891-1900 1901-1905 TOTAL

Total

1 1

3 6

4 7

2

13 19 20 21

15 21 20 22

2

o 1

o o o 7

Source: Crane 1908, appendix table 1, pp. 652-59.

4

4

15

15

4

4

105

112

628

Hugh Rockoff

first to install a long-distance telephone line (Kelley 1959, p. 50). Shortly after its invention, dynamite was first used on California's Mother Lode in 1868; electric power was first used there in 1896 (Logan 1934, p. 10). These adaptations did not disturb the commodity-price equilibrium under a gold standard; they served to maintain a rate of productivity growth in gold mining comparable to productivity growth in the economy as a whole. Some innovations were specific to gold mining, including hydraulic mining, dredging, and the chlorination and cyanide processes for extracting gold from ore. The first two appear to have evolved gradually from known technologies in response to the exhaustion of rich alluvial deposits. Large-scale hydraulic mining began in California in the early 1850s. Once the richest deposits were exhausted, miners sought methods to permit them to wash a much larger volume of gravel per hour. It was soon discovered that by directing a flow of water against a bank containing gold-bearing material and then draining the runoff through long sluices charged with mercury, relatively low-grade ores could be worked profitably. The scale of the largest operations increased steadily. By the late 1870s, the North Bloomfield mine, one of the largest, had over a hundred miles of ditches and flumes to carry its water supply and six tunnels totaling four miles in length to drain the runoff. Streams of water hundreds of feet long fired from cannons called "little giants" were used to tear away the mountains, and sluices thousands of feet long were used to collect the gold (Kelley 1959, pp. 47-53). Despite the scale of the hydraulic mines, however, no new principle seems to have been involved. The use of mercury to form an amalgam with gold had been known since ancient times, and the process of washing the gravel from the hydraulic mines was an adaptation of the principle the forty-niners used when they panned for gold. Indeed, a similar evolution from panning to hydraulic mining occurred in Brazil during the eighteenth century, although not pursued to the same extent as was later achieved. The main effect of the development of hydraulic mining was a partial offset to the exhaustion of the richest placer deposits. A similar account applies to gold dredging, but in this case the adaptation of technologies from other industries may have been important as well (Weatherbe 1907). The one inn
629

The Real Price of Gold

Africa in 1890. The cyanide process revolutionized the metallurgy of gold. World consumption of cyanide was probably less than 50-tons per year in 1889 and rose to 10,000-tons per year by 1905, with about one-third of the latter total consumed by South Africa (ClenneI1910, p. 31). The part to be assigned the introduction of this process in explaining the concurrent rise in world gold production from 5.8 million ounces in 1890 to 18.5 million ounces in 1905 is not clear. Kitchin (1931, pp. 57-58), perhaps the leading authority of his time, attributed tremendous importance to cyanide, claiming that it had increased yield per ton of existing producers by 50 percent and had permitted the treatment of much ore that would not otherwise have been profitable to mine. His view, however, may give too much weight to the early experience in South Africa where cyanide was crucial. Gold in the Rand is found in a finely divided state throughout the ore, so the ore was especially well suited to cyaniding. MacArthur's (1905, p. 314) memory of his first visit to Rand is probably accurate. A boom there had petered out as the ores that could be worked profitably with existing techniques were exhausted. One authority on South African gold (Letcher 1974, book 1, p. 99) refers to the role of the innovation as "salvation by cyanide." In other regions, however, cyanide appears to have contributed much less to the second surge in gold production. Increased production in the United States was second to the rise of the Rand in explaining the increased production at the turn of the century. California, long the major source of gold in the United States, increased production about 43 percent between 1890 and 1905, acounting for about 12 percent of the increase in U.S. output and about 2.5 percent of the increase in world output. But cyanide does not appear to have been a significant factor in California. About one-half of the increase in California's output was due to the expansion of dredging. The ores of the Mother Lode, moreover, were not well suited to cyaniding (Logan 1934, p. 10). Colorado, the largest contributor to the increase in the U.S. gold production, contributed about 8.6 percent of the world increase between 1890 and 1905. Much of the gold came as noted above from the rich Cripple Creek lode where chlorination, the older process, and cyaniding were highly competitive (Crane 1908, pp. 491-92).4 Colorado attained peak output in 1900. In certain other areas of the world that had long been producers, notably Europe and South America, there were no increases in production during the period in which cyanide was introduced. Further research may show that some of the other new fields that contributed to the increase in world gold production between 1890 and 1905 were made profitable by cyanide-fields in Australia, which contributed 18.9 percent of the increase; Mexico, which contributed 5.9 percent; and Canada, which contributed 4.9 percent. These three fields

630

Hugh Rockoff

were next in importance to the South African and U.S. fields. Tentatively, however, it seems that the cyanide process must be rated an innovation that had a major impact on world stocks of monetary gold primarily because of the joint discovery of the main reefs of the Rand. Strong circumstantial evidence indicates that the discovery of the cyanide process, like the discovery of some of the major gold fields discussed above, was the product of the high real price of gold prevailing in the mid-1880s. In other words, the increases in production attributable to cyanide must also be deemed movements along a long-run supply curve. Consider first that MacArthur and the Forrests, the innovators, were engaged in a commercial venture with the object of finding cheaper means of extracting gold from ore. According to MacArthur's own account (1905, p. 312), their syndicate was originally formed with the intention of perfecting the chlorination process. They simultaneously carried out research on a list of other solvents, one of which was potassium cyanide. Their research was clearly profit oriented; their discovery was not the accidental by-product of other research, for example, basic research pn the characteristics of cyanide. Basic research did account for the presence of cyanide on their research agenda and influenced their experiments, but the basic research had been done much earlier. Consider next that a number of other metallurgists were simultaneously studying potassium cyanide as a gold solvent with an eye to improving the extraction process. The degree of originality claimed for the MacArthur-Forrest process was challenged in the courts by metallurgists asserting their o\vn claims; the process thus gave rise to considerable litigation. One writer summed up the litigation this way: "An examination of the patents granted for the cyanide processes and improvements shows that it is an easy matter to obtain a patent, but a difficult matter to retain it if someone else wants to make use of anything claimed in the patent" (Wilson 1902, p. 15). The most negative assessment of the originality of MacArthur and the Forrests that I have found is by Gaze (1898, p. 5), who maintains that several metallurgists were close to perfecting the process (himself included) and that the process was well known in the jewelry trade. But even Clennell (1910, p. 23), who believes that MacArthur and the Forrests were well in front of the competition, acknowledges that there was one patent, issued to the American Jerome Simpson in 1865, for a process that might have been modified into a commercial success had it been tried. Evidently, the extensive experimentation with cyanide in the late 1880s would sooner or later have yielded a commercially feasible process. It is hard to escape the conclusion that this activity was the product of a high real price of gold. A similar conclusion applies to the chlorination process that was used in many regions before cyanide. Chlorination was discovered indepen-

631

The Real Price of Gold

dently by metallurgists in the United Kingdom in 1848 and Germany in 1849 (Rose 1915, pp. 300-301) when the price of gold was nearing its midcentury peak. The German discoverer had in mind the commercial exploitation of certain gold-containing residues from the production of arsenic. Thus, the chlorination innovation appears to have been motivated by the high real price of gold. To go beyond qualitative evidence on particular aspects of the production of gold to year-by-year measures of total-factor productivity would require data directly measuring stocks of capital and labor or indirectly measuring costs. Unfortunately, the data for most areas, for the United States in particular, do not appear to be adequate. It is possible, to be sure, to collect observations on what the industry referred to as working costs. Janin (1912, 1915), for example, reports data on mining costs, usually expressed as dollars per ton of ore mined. But these figures generally omit capital and exploration costs and sometimes other overhead costs as well. They also show enormous variation from place to place, as is to be expected. High-grade ores are worth exploiting even if mining costs per ton are high. The variance and the absence of continuous observations on a particular site make it impossible, in most cases, to construct a usable time series. For an amusing view of the difficulties involved in measuring mining costs, it is instructive to peruse the replies to a U.S. Senate Committee which asked various experts to give their opinions on the cost of mining silver in the early 1890s. The data offered were sparse, and opinions ran the gamut from the vew that silver mining was highly profitable to the view that, if exploration costs were included, silver was mined at a loss (U.S. Congress, Senate, Committee on Mines and Mining 1892). For South Africa, however, there may be sufficient data on resources employed or costs to compute annual measures of productivity change. Some crude measures, tons of ore milled per worker and tons of ore hoisted per worker in large mines, are shown in table 14.4. The measures suffer the problem common to all measures of labor productivity; they confound changes in the ratio of capital to labor with technological change and do not deal with the additional problem of aggregating white and nonwhite labor. The first and third columns in table 14.4 aggregate white and nonwhite labor on a one-for-one basis; the second and fourth columns weight white labor by the prevailing ratio of white to nonwhite wages. The latter procedure undoubtedly overstates productivity differences between the two classes of labor because the racial situation distorts the wage data. Thus the two estimates probably bound the true measure. Despite the weaknesses in the approach, the data suggest some plausible conclusions. The shorter-run movements shown in the upper part of the table appear to be influenced a good deal by discoveries, particularly the

632

Hugh Rockoff

Table 14.4

Productivity Change in South Africa's Gold Mines, 1916-60

Tonnage Milled per Worker (1) 191~20 192~30 193~40 194~50

195~60

191~60 191~50 192~60

Tonnage Milled per Equivalent Nonwhite Worker at Prevailing Wages (annual percentage (2)

Tons Hoisted in Large Mines per Worker rates of change) (3)

Tons Hoisted in Large Mines per Equivalent Nonwhite Worker at Prevailing Wages (4)

0.96 1.37 1.54 0.51 -0.10

-0.73 3.92 0.42 -1.16 -0.74

-0.12 2.17 1.42 0.40 0.51

-1.21 4.77 0.84 -1.39 -0.05

0.86 1.10 0.83

0.36 0.64 0.61

0.88 0.97 1.13

0.59 0.75 1.04

Source: Katzen 1964; tonnage milled, number of workers, and wages, pp. 18, 19; tons hoisted per workers in large mines, p. 26.

opening of the "New Rand" region which became significant in World War I. Over the long run some growth in labor productivity apparently occurred, although at a slower rate than for economy-wide measures. Kuznets's (1971) survey of growth rates showed output per worker growing at 1.81 percent per year in the United States, 1920-60, which is higher than any of the comparable rates in table 14.4. He also showed that output per capita in a number of industrial countries was growing more rapidly than labor productivity in the gold mines. If one recomputes the first two columns of table 14.4 using gold mined, rather than tons milled, the long-run growth rates are still lower due to the exhaustion of more accessible ores. Similar calculations for the nineteenth century, based on a wide geographic area in which new mines were included, if feasible, would probably show higher rates of productivity change that were closer to economy-wide measures. Such a result would have been conducive to price stability under a gold standard. In effect, the supply of gold was shifting out as rapidly as the demand, abstracting from other influences, thus maintaining commodity-price stability. 14.5 Politics and the Supply of Gold

In this section I examine ways in which governments influenced gold mining, including both deliberate acts of monetary policy and other acts that inadvertently altered the supply of gold. It will become obvious as I

633

The Real Price of Gold

proceed that the real price of gold also influenced the supply indirectly through its effects on government policies toward mining. Neither set of influences was a major source of variability in the supply of gold during the classical period, but both are important because they occurred when the role of government in gold standard countries was more limited than it is today. The history of political influences on the supply of gold thus provides an important set of analogies for understanding the significance of a return to the gold standard in modern circumstances. 14.5.1

Deliberate Intervention

In the period prior to the discoveries of gold in California and Australia, production was dominated by supplies from South America and Russia. In both regions heavy taxes were imposed on gold mines, but tax collection varied with the profitability of the mines. Thus in periods of low commodity prices, when gold-mining profits were high and profits in other industries low, the equilibrating mechanism was hampered as governments in those regions attempted to collect shares of the profits of the mines. The regimes there found mercantilist policies congenial, so a gold standard based on supplies from these regions was unattractive. This state of affairs 'vas well understood by Jevons. Writing in the wake of the discoveries in California and Australia, he foresaw a period in which the supply of gold would be more abundant, more stable, and more responsive to the real value of gold. The abundance and, in part, the stability of supply were to be accounted not only by the richness of the finds, but also by the location of the finds in regions subject to the rule of laissez-faire. Accordingly, in the future the supply of gold would be more responsive to changes in its real value. To quote Jevons (1884, p. 74): Before the recent discoveries, no gold mines of value have been in possession of any Anglo-Saxon nation. They have been chiefly in the hold of the Spanish and Russian Governments, subject to arbitrary restrictions and taxes. In English or American hands the production of gold becomes a matter of free industry and skill. It must follow that the produce will conform more closely to commercial principles; a rise or fall in the value of gold will be followed more exactly by an extension or cessation of the production. At the same time, the greater area of production, offering scope for more various competition and equalization of local fluctuations, and the greater and more various modes of consumption, will all tend to render the demand and supply of gold more equable and its value more constant. While Jevons was right about the direction of the change in the role of politics, the U.S. and Austral~an experience were not free of interactions between politics and gold mining. Perhaps the most interesting U.S. example was the restriction of hydraulic mining in California and its

634

Hugh Rockoff

subsequent removal. In section 14.4, I mentioned the large-scale operation of hydraulic mines in California during the 1870s. Privately some of these mines were highly profitable, but they created a significant externality. The debris they produced moved downstream, ruining farmland and clogging navigable rivers. The protracted legal and legislative battle between the farmers and the miners was documented by Kelley (1959). The farmers prevailed. In a series of injunctions issued between 1882 and 1884 the courts closed the major hydraulic mines. The battle went on for several more years, with some operations continuing on a clandestine basis and others openly in regions where the damage to farmland was less severe. The effect on the output of gold from California was pronounced, although an exact estimate cannot be made. One problem is that the best source of data (Hill 1929, pp. 20, 21) on the output of the California mines splices two series in 1883 in the midst of the hydraulic-mining controversy. Nevertheless, a plausible estimate of the effect of the closing of the hydraulic mines can be made by subtracting average production in 1885-87, the first three years in which the full effect was felt, from the average in 1879-81, the last three years of unrestricted hydraulic mining. The shortfall amounts to 289,000 ounces per year, about 5.7 percent of world production in 1885-a small but not trivial effect. This analysis, incidentally, adds one more irony to the long list of those in monetary history. A small part of the price decline of the 1890s of which farmers complained so bitterly might have been produced by those in California who forced the closing of the hydraulic mines. To be sure, the closing of the hydraulic mines was not a conscious act of monetary policy. Farmers trying to protect the value of their property were responsible for closing the mines. Monetary considerations played a role, however, in the early 1890s when there was an attempt to revive hydraulic mining; the Carminetti Act, a federal law signed in 1893, established a California Debris Commission to authorize the opening of mines, provided they had made adequate preparations in the form of dams or settling basins to prevent debris from entering the water supply. In part, the act simply aimed to start up a profitable industry in a period of depression, a course that might have been equally appealing with respect to a nonmonetary industry, although the fixed nominal price for gold made hydraulic mining a better candidate for aid. But in part, the support for the bill reflected a desire to expand the money supply (Kelley 1959, pp. 271-72). The act's costly requirements for restraining debris, however, prevented a renewal of hydraulic mining on the scale of the 1870s. A crude estimate may be calculated of the increased production owing to the Carminetti Act analogous to the estimate of the shortfall computed above. Subtract average output in 1890-93, the three years preceding the act, from average output in 1895-97, the first three years following the

635

The Real Price of Gold

effective removal of the restriction. The increase amounts to 172,000 ounces per year, about 1.8 percent of world production in 1895-again, a small but not trivial effect. Thus, this increase in output, and certain increases attributable to dredging where environmental concerns were also important, can be linked to the high real value of gold or at least to the associated economic conditions. The law was amended again in the 1930s to provide for a federal dam to hold back debris and thus permit greater hydraulic mining, a response comparable to that in the 1890s in similar economic conditions. The hydraulic-mining controv~rsy is the most dramatic U.S. example, but it is not the only example nor probably not the most important one. Meade (1909, pp. 112-13) regarded the extension of the railroad network into gold-mining areas the single most important factor that lowered private-gold-production costs by reducing freight costs and making the introduction of heavy machinery possible. As a result, both the United States and Mexico substantially increased their production at the turn of the century. Given the long history of government involvement with railroads in this country and Mexico, an element of subsidy likely was present and, given the timing of the construction, overall economic conditions likely influenced the decision to expand the supply of gold. In eastern Australia, the first region opened, production was initially based on placer mining. When production declined, the government imported and worked diamond drills to facilitate finding and working deep lodes (Jevons 1884, pp. 117-18). In western Australia, the government built stamp mills to which small mines brought their ore. It also introduced a water system to facilitate production at the Kalgoorlie mines (Mead 1909, pp. 115-16). In both areas, the government probably responded to output declines resulting from the exhaustion of high-grade ores rather than to larger monetary considerations. In eastern Australia, the actions were destabilizing in the sense that they tended to increase world production in a period of rising prices, although the gentle upward trend in prices might have been desirable. In South Africa, the government influenced the production of gold in a host of ways. Trying to predict in the midthirties the long-run supply from South Africa, Hardy (1936, p. 67) argued that any tendency for the supply to fall could be offset by actions that reduced the burden of direct and indirect taxation. In his words: In short, the present level of costs is shot through with items which really constitute a distribution of the net income of the industry to elements in the community which are able to grasp it rather than the payment of costs that really have to be met in order to maintain a high level of output. Account must be taken of the gradual elimination of these elements . . . as the industry approaches the time when further maintenance of production would involve a cost disproportionate to the return.

636

Hugh Rockoff

The costs included direct taxation, which was higher on the gold mines than on other industries in South Africa (Busschau 1936, pp. 165-78), railway rates, and even some elements of wages. It was clear, then, that the supply of South African gold could not be projected without taking into account the political response of the South African government to changes in the value and output of gold. Hirsch (1968) documented many of these interventions for later years. The 1930s, although they lie outside the classical period, provide further evidence of the role of politics. The real value of gold rose during the early ·1930s, and world output increased. The largest increases between 1929 and 1934 were recorded by the Soviet Union (about 3.1 million ounces), Canada (1.0 million ounces), and the United States (0.7 million ounces). South Africa, by contrast, raised its output only by .07 million ounces. The increase in the United States was attributable to the working of low-grade ores including the revival of placer mining-much of it panning in California-undoubtedly a response to the rising real value of gold. The exploitation of new finds in Ontario, also partly in response to the rising real value of gold (Hardy 1936, pp. 58-60), accounted for the increase in Canadian production. In the case of the largest increase, that of the Soviet Union, the role of politics was pronounced. By one account, and given the nature of things the account would be a difficult one to verify, the increase in Soviet output ultimately flowed from Stalin's judgment that Siberia could be developed more rapidly by emphasizing gold. Stalin, supposedly familiar with the history of gold in California, including the vivid writings of Bret Harte, thought that the California gold rush had led to· rapid economic development and had provided the North with an important war chest for use in the Civil War. As a result, he gave a high priority to the revival and expansion of gold production in Siberia (Littlepage 1937, pp. 26-33). Indeed in order to increase prospecting, the Soviets introduced their own form of the gold rush in 1933. Private prospectors were rewarded with a lump-sum payment and the right to work their claims for one year. All payments were in "gold rubles," a currency that could be used in special stores and worth considerably more than the paper ruble in which other wages were paid. The campaign was apparently successful (Littlepage 1937, pp. 125-31). In the midst of the Depression, Keynes (1936) argued that communist efficiency in producing gold might save capitalism by encouraging expansionary monetary policies. He might have been surprised, however, at the extent to which the increase in production relied on capitalist techniques. 14.5.2 War, Revolutions, and Strikes During the decade in which U.S. and Australian supplies dominated the growth of the gold stock, interruptions due to wars and revolutions

637

The Real Price of Gold

did not occur. In earlier and later periods, however, such interruptions were common. The Latin American wars of independence in 1815 were associated with a sharp drop in world gold production. World output fell from 5.9 million ounces in the first decade of the nineteenth century to 3.8 million ounces in the second. All of this fall can be accounted for by the decline in South American and Mexican production; production in other areas was stable or increasing. It is true, as Von Humbolt (1900, pp. 32-33) argues, that particularly in Brazil, before political upheavals shook the region, placer-mining output was declining in South America, and some of the decline may have been in reported rather than actual output. But it seems likely that the upheavals, at least in the Spanishspeaking areas, if not in Brazil, did leave a substantial mark on the growth rate of world monetary gold stocks. The greatest producer of gold toward the end of the classical period was South Africa. Significant interruptions in the flow of gold occurred there, the most severe during the Boer War. Tensions of long standing in southern Africa between settlers of Dutch and British descent were heightened in 1886 when the discovery of gold in the Transvaal brought an influx of British and other foreigners into a region formerly controlled by Dutch descendants. The conflict began in the fall of 1899 and was not concluded until the spring of 1902. After some initial Boer successes, the British in the summer of 1900 seemed to win control of the region by capturing Johannesburg, the center of the gold-mining industry, and other towns. But the Boers then turned to guerilla warfare and only two years later were finally subdued. The impact of the Boer War on the output of gold was substantial. South African output fell from 3.6 million ounces in 1899 to a mere 0.3 million ounces in 1901. It was not until 1905 that output exceeded the level in 1898, the last full year before the conflict. The shortfall in South African output, however, did not have the severe consequences for world financial markets that one might have predicted, although Friedman and Schwartz (1963, p. 148) note that there was some stringency in money markets in the fall of 1899 that could be associated with the onset of the Boer War. There was a stock market panic, and some banks and financial institutions failed, but the period of stringency was brief. The reason the war had so little effect is that the South African shortfall occurred after several years of substantial increases in the output of gold. The decline in South Africa was cushioned by increases in other areas-Alaska, the Klondike, Colorado, and western Australia-as Clapham (1907, p. 378) noted. Total world output in 1900, although 18.2 percent lower than in 1899, was still the third highest output recorded up to that date, about equal to the average of the preceding five years. The supply of South African gold was also interrupted in 1922 by the strike of white miners on the Rand. The fall in the purchasing power of gold relative to prewar levels and the consequent cost squeeze on mar-

638

Hugh Rockoff

ginal mines were the basic factors underlying the strike. In that environment, efforts by the mines to alter work rules soon led to fears that the color bar was being abandoned. A violent strike broke out that the government quelled, but South African production fell by about 1.1 million ounces between 1921 and 1922. This decline also was offset, to some extent, by increased production in other areas; world production fell only by about 0.5 million ounces, corresponding to a 3.4 percent fall in the rate of production. These are the most important examples of interruptions in gold output, but there are many more. Mexico and Russia, for example, experienced significant output declines during revolutionary periods, and labor strikes in New Zealand hampered production after World War I. The main point, however, is that such interruptions in supply were rare during the heyday of the gold standard because of the political stability of the regions from which much of the gold came. It one were contemplating a return to the gold standard, one would want to know whether such conditions would again be the norm. 14.5.3

Governments and the Real Price of Gold

These examples suggest that a complete model of the gold standard would need to take into account the motives that prompt government intervention in industry generally and gold mining in particular. If government actions with regard to gold mining were undertaken with a view toward the monetary implications, then they might well be stabilizing. The measures designed to revive hydraulic mining in California, undertaken during a period of depressed prices and business activity, appear to fit the case. Such actions, moreover, could well be undertaken simultaneously throughout the gold bloc since most countries on the standard would likely be sharing similar experiences with respect to prices and real output. Intervention would also be stabilizing if the gold-mining industry were viewed simply as a particularly promising candidate for aid in a period of depressed trade and governments were shopping for likely candidates for aid. Since the demand for gold with respect to its nominal price is perfectly elastic under a gold standard, measures designed to aid the industry might appear more effective than similar efforts expended on behalf of other industries. On the other hand, if government actions were tied to circumstances in a particular gold field, there would be less chance for simultaneous responses. The efforts in Australia, for example, to stimulate deep-lode mining in the eastern fields after placer-mining output declined fit this case. But it is not at all certain that government aid to gold mining would be countercyclical or neutral. Under a gold standard, it should be remembered, gold-mining profits tend to behave countercyclically. If the industry were able to win aid, as many industries do, when its output was

639

The Real Price of Gold

depressed by rising costs, and if the rising costs were responding to a general rise in prices, then such aid would inhibit the equilibrating process. Such aid, moreover, might well occur simultaneously in the gold standard world. Consider, for example, an inflation. The cost-ofproduction theory of the value of gold and commodity prices envisions an inflation brought to a halt by a decrease in the output of gold caused by a rise in mining costs. But the process might be inhibited if gold-mining industries simultaneously won governmental relief that maintained production. Some recent events in South Africa and elsewhere might be interpreted in this light, but I have not found clear examples in the classical period. Interruptions of supply that were the by-product of political conflicts, as I noted above, were important in certain periods. It might be possible in some cases to link these interruptions to the monetary role of gold and to changes in the real price of gold. So a complete theory of gold supply might consider this link as well as more conventional ones. For example, the strategic role of gold might encourage military actions to control the supply, often cited as the ultimate cause of the Boer War. In most cases, however, such interruptions appear to have been random shocks. 14.6 Conclusions

The stability and elasticity of the long-run supply of gold played a crucial role in the classical analysis of the gold standard. In the simplest case-a stable and perfectly elastic supply curve-the cost of producing gold is the ultimate determinant of the price level. Temporary increases (or decreases) in the price level lead to decreases (or increases) in supply that ultimately return prices to their long-run equilibrium. The main weakness in the classical theory from a modern perspective is that it does not treat systematically the case of gold as a nonrenewable resource. Hence, conditions appear to be favorable for a marriage between the classical theory of the gold standard and the emerging theory of the economics of a nonrenewable and durable resource. Some examples of the progeny that might be expected from such a marriage are given in section 14.1. From a very long-run perspective, the supply curve appears to have been stable and elastic in the classical period. Rapid increases in supply occurred at points in time, at midcentury and in the late 1890s, when they served to raise the nominal price level after a period of deflation. This point is developed in section 14.2 in relation to modern changes in high-powered money. A closer look reveals, however, that the surges were due primarily to the development of new gold fields and to a lesser extent to technological changes. The question is whether these events should be viewed as fortunate accidents or as changes induced by pre-

640

Hugh Rockoff

vious changes in the real price of gold. Circumstantial evidence, particularly for the second surge in production, suggests that the discoveries were induced. Tentatively then, it seems appropriate to regard changes in supply produced by new discoveries at the end of the century as movements along a long-run curve rather than as a series of curves with arbitrary shifts between them. A similar conclusion seems in order for technological change. The most famous innovation, the introduction of the cyanide process, probably had a somewhat smaller impact on the supply of gold than has been suggested in the literature, but it appears to have been induced by the prevailing high real price of gold. Unfortunately, data are lacking to determine the exact course of technological change and its dependence on the real price of gold during the classical period. Some more recent data for South Africa are examined in section 14.4. Governmental policies represent a third channel through which changes in the real price of gold influenced supply. Environmental regulations, for example, were weakened in California to encourage production in a period of low commodity prices. Purely random shocks from political sources were rare during the heyday of the gold standard, in part due to the dominance of supply by politically stable countries committed, to some extent at least, to laissez-faire. During the early part of the nineteenth century and after the 1890s, however, political shocks became more important. It would be unwise, however, to infer from the stability and elasticity of the supply of gold in the nineteenth century that a gold standard would work similarly today. Certain conditions that existed then exist no longer. The regular discovery of new supplies depended on a core of dedicated prospectors working in largely unexplored areas. Neither that core of prospectors nor unexplored regions of the same type now exist. It is uncertain whether a similar core could be created by other means. The absence of political interference, moreover, hardly seems as likely in today's world as it was then. Finally, the long and uncertain lags that marked the supply response hardly seem acceptable in an impatient age.

Notes 1. I am indebted to Richard Sylla for drawing my attention to the significance of the causation issue. 2. The long-run supply curve referred to here is more general than the one ordinarily considered. It incorporates, for example, the effects of endogenous technological change. 3. After presenting this paper I learned that John McDermott had independently arrived at similar conclusions. He is currently working on a more sophisticated model of the behavior of a gold standard when gold is a nonrenewable resource.

641

The Real Price of Gold

4. In an earlier draft of the paper I tried to reinforce this point by referring to evidence showing that quoted production costs for the two processes were similar. Geoffrey Wood, however, called to my attention that average (as distinct from marginal) production costs, as these figures undoubtedly were, are of no special significance. What matters is that both processes could compete successfully for customers.

References Barro, Robert J. 1979. Money and the price level under the gold standard. Economic Journal 89: 13-33. Blainey, Geoffrey. 1970. A theory of mineral discovery: Australia in the nineteenth centure. Economic History Review 23: 298-313. Bordo, Michael David. 1981. The classical gold standard: Some lessons for today. Federal Reserve Bank of St. Louis Review 63: 1-17. Busschau, W. J. 1936. The theory of gold supply. Oxford: Oxford University Press. Cagan, Phillip. 1965. Determinants and effects of changes in the stock of money, 1875-1960. New York: Columbia University Press. Calvo, Guillermo A. 1978. Optimal seigniorage from money creation: An analysis in terms of the optimum balance of payments deficit problem. Journal of Monetary Economics 4: 503-17. Clapham, (Sir) John. 1970. The Bank ofEngland: A history. Cambridge: Cambridge University Press. Clennell, J. E. 1910. The cyanide handbook. New York: McGraw-Hill. Crane, Walter R. 1908. Gold and silver: Comprising an economic history of mining in the United States, the geographical and geological occurrence of precious metals, with their mineralogical associations. New York: John Wiley and Sons. Curle, James Herbert. 1905. Gold mines of the world. London: George Routledge and Sons. Dane, Ezra G. 1935. Introduction. In The life and adventures ofJames W. Marshall: The discoverer of gold in California by George R. Parsons. San Francisco: George Fields. de Launay, Louis. 1908. The world's gold, its geology, extraction, and political economy. New York: G. P. Putnam's Sons. Edie, Lionel Danforth. 1929. Capital, the money market, and gold. Chicago: University of Chicago Press. Fisher, Irving. 1922. The purchasing power of money. 2d ed. New York: Macmillan. Friedman, Milton. 1953. Commodity-reserve currency. In Essays in Positive Economics. Chicago: University of Chicago Press. (First published in Journal of Political Economy 59: 203-32.)

642

Hugh Rockoff

- - - . 1969. The optimum quantity of money. In The optimum quantity of money and other essays. Chicago: Aldine. - - - . 1971. Government revenue from inflation. Journal of Political Economy 79: 846-56. Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history ofthe United States, 1867-1960. Princeton: Princeton University Press. Gaze, William H. 1898. A handbook of practical cyanide operations. Melbourne: George Robertson and Company. Gray, James. 1937. Payable gold. South Africa: Central News Agency, Limited. Hardy, Charles Oscar. 1936. Is there enough gold? Washington, D.C.: Brookings Institute. Hill, James M. 1929. Historical summary of gold, silver, copper, lead, and zinc produced in California, 1848-1926. Report prepared for the U. S. Department of Commerce, Bureau of Mines. Economic Paper 3. Washington, D.C.: GPO. Hirsch, Fred. 1968. Influences on gold production. International Monetary Fund Staff Papers 15: 405-88. Janin, Charles. 1912. Operating costs of California mines. Mining and Scientific Press. 105: 520-23. - - - . 1915. Placer-mining methods and operating costs. Report prepared for the U.S. Bureau of Mines. Bulletin 121. Washington, D.C.: GPO. Jastram, Roy W. 1977. The golden constant: English and American experience, 1560-1976. New York: John Wiley and Sons. Jevons, W. Stanley. 1884. A serious fall in the value of gold ascertained, and its social effects set forth. In Investigations in currency andfinance, ed. H. S. Foxwell. London: Macmillan. Katzen, Leo. 1964. Gold and the South African economy. Capetown: A. A. Balkema. Kelley, Robert L. 1959. Gold vs. grain, the hydraulic mining controversy in California's Sacramento Valley: A chapter in the decline of the concept of laissezjaire. Glendale, Calif.: A. H. Clark Co. Keynes, J. M. 1936. The supply of gold. Economic Journal 46: 412-18. Kitchin, Joseph. 1930. The supply of gold compared with the prices of commodities. In Interim report of the Gold Delegation, Annex 11, 79-85. Prepared for the League of Nations. Geneva. - - - . 1931. Gold production. In The International gold problem. Oxford: Oxford University Press. Kuznets, Simon. 1971. Notes on the pattern of U.S. economic growth. In The reinterpretation of American economic history, ed. R. W. Fogel and S. Engerman. New York: Harper and Row. Letcher, Owen. 1974. Gold: Historical and economic aspects. New York: Arno Press.

643

The Real Price of Gold

Levhari, David, and Robert S. Pindyck. 1981. The pricing of durable exhaustible resources. Quarterly Journal of Economics 96: 365-77. Littlepage, John D. 1937. In search of Soviet gold. New York: Harcourt, Brace. Lock, Alfred G. 1876. Gold: Its occurrence and extraction. New York: Spon. Logan, Clarence August. 1934. Mother Lode belt of California. Prepared for the California State Division of Mines, bulletin 108. MacArthur, John S. 1905. Gold extraction by cyanide: A retrospect. Journal of the Society of Chemical Industry 24: 311-15. Marshall, Alfred. 1929. Money, credit, and commerce. London: Macmillan. Meade, Edward S. 1909. The story of gold. New York: Appleton. - - - . 1897. The production of gold since 1850. Journal of Political Economy 6: 1-26, 138. Mill, John Stuart. 1871. Principles ofpolitical economy, ed. W. J. Ashley. 7th ed. London: Longmans, Green and Company. Morrell, W. P. 1940. The gold rushes. London: Adam and Charles Black. Paul, Rodman Wilson. 1947. California gold: The beginning of mining in the Far West. Cambridge: Cambridge University Press. Ricardo, David. 1821. The principles of political economy and taxation. 3d ed. London: Dent. Ridgway, R. H. 1929. Summarized data of gold production. Prepared for the U.S. Bureau of Mines. Economic paper no. 6. Washington, D.C.: GPO. Rose, T. K. 1915. The metallurgy of gold. London: Charles Griffin and Company. Salmon, J. H. N. 1963. A history of gold mining in New Zealand. Wellington, New Zealand: R. E. Owen. Stewart, Marion B. 1980. Monopoly and intertemporal production of a durable extractable resource. Quarterly Journal of Economics 94: 99112. Tatom, John A. 1980. Issues in measuring an adjusted monetary base. Federal Reserve Bank of St. Louis Review 62: 11-20. U.S. Bureau of the Census. 1960. Historical statistics ofthe United States: Colonial times to 1957. Washington, D.C.: GPO. U.S. Congress. Senate. Committee on Mines'and Mining. 1892. Cost of production ofgold and silver bullion. Report no. 1310. 52nd Cong., 2d sess. Von Humbolt, Baron Alexander. 1900. The fluctuations of gold. Trans. William Maude. New York: Cambridge Encyclopedia Co. Weatherbe, D'Arcy. 1907. Dredging for gold in California. San Francisco: Mining and Scientific Press.

644

Hugh Rockoff

Wicksell, Knut. 1906. Lectures on political economy. Trans. E. Classen; ed. Lionel Robbins. London: Routledge and Kegan Paul. Wilson, E. B. 1902. Cyanide processes. 3d ed. New York: John Wiley and Sons.

Comment

Robert J. Barro

Rockoff's paper analyzes the reaction of gold production to movements in the relative price of gold. With gold's nominal price held fixed-as it was for most of the 'periods and countries studied-we are considering here the response of gold output to variations in the general price level. Rockoff discusses gold discoveries, advances in mining techniques (with special attention to the cyanide process), the willingness of producers to adopt existing capital-intensive methods of production, and the tendencies of governments either to invoke environmental-protection regulations that hampered mining or to relax these restrictions. The results suggest various mechanisms by which gold output would respond to market incentives. But it is difficult to use Rockoff's findings in a quantitative manner to assess responses either at the level of an individual mining operation or at the aggregate level. Mark Rush (1978) calculated the following regression on annual data for the 1897-1913 period (which he has been reluctant to report himself):

log(gt) = ... + .31 'log(I/Pr) (.12)

+ .82 ·IOg(gt-l),

0- = .062,

(.09) where gt is the world's total gold production for the year (as reported by Warren and Pearson 1933, p. 197), and Pr is an estimat~ of the GNP deflator for the United States. Note that the dollar value of gold was fixed over this period, so that the variable I/Pr measures the price of gold relative to other goods in the U.S. (Standard-errors-of-coefficient estimates are in parentheses-o- is the standard error of estimate.) The results show a short-run price elasticity for gold production of .3 and-if one takes seriously the coefficient of the lagged dependent variable-a long-run price elasticity of 1.7. There are some simultaneity problems that are not handled in Rush's regression-notably, an autonomous shift in gt would tend to raise Pt. However, this effect is from gt to the inflation rate rather than to the price level. The direction of bias would also seem to be downward for the estimated coefficient of 10g(I/Pr). Robert J. Barro is professor of economics at the University of Chicago and a research associate of the National Bureau of Economic Research.

645

The Real Price of Gold

Michael Bordo (1981) reports a statistically significant relation of the world's monetary gold stock to prior values of the purchasing power of gold. However, his results were not presented in a form that allows me to calculate price elasticities of output response. By looking at the monetary gold stock-which would ultimately be of concern from a monetary standpoint-Bordo's estimates would also combine responses of nonmonetary stocks of gold with changes in the output of new gold (whereas Rush looked only at the production of new gold). Another possibility for quantitative analysis involves the detailed data that are available on equity prices for South African gold-mining companies. These figures are presented in the book by S. Herbert Frankel (1967). Frankel uses these numbers only to calculate long-period rates of return to holdings of stock in gold-mining companies. But we could utilitze these data to assess the impacts on equity values of changes in the general price level, gold discoveries, technical advances in mining, shifts in government policies, wars, and so on. That is, we could isolate quantitative effects Jor many of the disturbances that Rockoff discusses. A distinction between expected and unexpected events would come into play in assessing the response of equity prices to the pertinent variables. In an earlier part of his paper, Rockoff contrasts the performance of the gold standard with that of the post-World War II U.S. monetary regime. He compares the means and standard deviations of annual growth rates for the world's monetary gold stock from 1839 to 1929 with that of the U.S. monetary base from 1949 to 1979. In terms of these monetary-base comparisons, the gold standard period looks good, even in terms of exhibiting lower standard deviations of growth rates on a year-to-year basis. Others who have made these types of comparisons-such as Bordo (1981)-have examined broader monetary aggregates, the general price level, or real variables like output. On this basis the gold standard seems to provide for long-run nominal stability, but not for stability of nominal or real variables on a year-to-year basis. Of course, the short-run behavior of broad monetary aggregates and prices under the gold standard depends, among other things, on shocks to the banking system, on changes in the velocity of broadly defined money, and on movements in the ratio of the monetary base to the reserve stock of monetary gold . We also have to worry about the distribution of the world's total stock of monetary gold among the various countries. Basically, endogenous movements in a country's stock of monetary gold may buffer that country's price level satisfactorily against various disturbances in the long run, but not in the short run. What should we conclude when we observe a greater year-to-year fluctuation in a country's broad monetary aggregate under the gold standard than under the post-World War II monetary regime? Perhaps

646

Hugh Rockoff

the comparison reflects other changes-such as the implementation of deposit insurance for the U.S. banking system-that stabilized the ratio of broad money to the monetary base. Then if similar structural changes had been implemented under the gold standard, we might have observed the relatively low year-to-year variability in the world's monetary gold stock-which is studied by Rockoff-carrying over to relative stability in broader monetary aggregates and the price level. A comparative study of short-run variability in broader monetary aggregates and prices may reflect alterations in banking institutions or other changes, rather than the shift away from the gold standard. References Bordo, M. 1981. The classical gold standard: Some lessons for today. Federal Reserve Bank of St. Louis Review 63 (May): 2-17. Frankel, S. Herbert. 1967. Investment and the return to equity capital in the South African gold mining industry. Cambridge: Harvard University Press. Rush, M. 1978. Money, output, and unemployment in the United States under the gold standard. Typescript. Warren, G., and F. Pearson. 1933. Prices. New York: John Wiley & Sons.

Reply

Hugh Rockoff

If Barro's suggestions for estimating various regressions may be regarded as agenda for future research, then I have no quarrel with them. More information is always welcome. Nevertheless, his exclusive emphasis on quantification forces me to defend the historical approach taken in my paper. The value of that approach is that it allows us to extract information from a knowledge of the particular circumstances that surround a particular data point, information that is lost when we quantify the data. Mark Rush's regression quoted by Barro, for example, is estimated over the years in which the discoveries in South Africa were made and the cyanide process was developed. The impact of the discoveries is clearly evident in the data that underlie Rush's regression. Should we regard them as events that shifted the level of output and possibly the supply elasticities, or should we regard them as a movement along the supply function? There are, of course, econometric techniques for attacking these issues that rely solely on the price and quantity data. But we can gain additional insight, I believe, by exploring the actual historical circumstances in which the discoveries· were made. In the case of cyanide, we can ask

647

The Real Price of Gold

whether the discovery was the accidental by-product of research directed toward other purposes, or whether it was the result of applied research aimed at finding better extraction methods. In the paper I argue that the latter model applies, so the case for treating the additional output associated with this development as induced is considerably stronger. In short, while I agree with Barro that a quantitative model of the supply function would be useful, I want to emphasize that the historical approach also has a role to play. Barro cites Frankel's (1967) study of rates of return to investments in South Africa's gold mines as an additional source of quantitative evidence. I agree that these data might provide additional insights, but two qualifications should be kept in mind. First, Frankel's data, of course, begin with the founding of the industry in South Africa, so they apply only to part of the period I examined, excluding many of the important nineteenth-century developments. Second, they are based on stockmarket valuations of mining stocks. For this reason proper use of the data requires exact knowledge of when it was first perceived that a technological change, or similar event, would increase the income of the mines. Beyond that date, the impact of the innovation will be capitalized in share prices and only normal returns will be observed. Share prices, moreover, are influenced by a host of factors such as the possibility of a return to the gold standard-an effect stressed by Frankel-that are difficult to evaluate. In other words the data, valuable as they may be, do not provide a short cut that avoids the need for detailed historical research to determine quantitative estimates of the effects of various disturbances. In the last part of his remarks Barro draws attention to the relative year-to-year stability of broad monetary aggregates, prices, and real output in the post-World War II period when compared with the gold standard period. If we leave aside the problem that the stability in the world's stock of monetary gold that I found may not imply stability in the monetary base for anyone country, a genuine paradox emerges. Barro's suggestion is that changes in the monetary system, particularly deposit insurance, may have stabilized the ratio of broad money to the monetary base, and that in turn this produced the greater stability exhibited by prices and real output after World War II. The implication is that if similar changes had been introduced during the gold standard period, the year-to-year stability of the stock of monetary gold might have carried through to prices and real output. Other sorts of changes after World War II, however, might have occurred that reduced year-to-year variability in the rates of growth of prices and real output relative to their trends. Cagan (1979), for example, found that the responsiveness of prices to variations in the business cycle declined after World War II, perhaps as a result of the higher average rate of inflation. If that effect or some other nonmonetary change accounts for the relative stability of prices and real

648

Hugh Rockoff

output in recent times, then the gold standard could not have produced such stability even with a modified financial superstructure. References Cagan, Phillip. 1979. Changes in the cyclical behavior of prices. In Persistent inflation: Historical and policy essays. New York: Columbia University Press. Frankel, S.Herbert. 1967. Investment and the return to equity capital in the South African gold mining industry, 1887-1965: An international comparison. Cambridge: Harvard University Press.

General Discussion ROCKOFF was generally sympathetic with Barro's comments. He suggested that there exists a great deal of quantitative evidence on South African gold production, which would be a fruitful area for further quantitative work on the determinants of gold supplies. He agreed that whatever the long-run responsiveness of the supply of gold, the short-run supply curve was rather inelastic. Hence, gold supplies provided little year-to-year stabilization of the price level. MCCAULEY asked how recent evidence suggesting that despite the rapid rise of the real price of gold in the last ten years, gold production has declined, squares with the evidence presented by Rockoff. KOCHIN replied that there is a specific phenomenon connected.with the economics of deep mining. If, e.g., there is a permanent increase in the real price of gold but rich veins of gold are surrounded by lower-quality ore, an attempt to mine the high-quality ore first, bypassing the lower quality ore, will impose greater costs later on. Consequently, if it is believed that the real price of gold has risen permanently, then firms will start mining the less rich ore that sits around the rich vein. The result is that in the short run, though the number of tons of ore mined increases slightly, the average gold content of the ore in ounces per ton falls and the number of ounces of gold produced decreases. In the long run, however, the supply of gold will respond positively as expected. BORDO expanded on Kochin's point. He described some empirical work he did for the staff appendix to the Gold Commission Report on the relationship between market-economy gold output and the real price of gold. Using annual data for the post-World War II period, he found a negative short-run supply elasticity. However, when he examined a longer series of data going back to the nineteenth century, he found a positive short-run response. This response is probably explained by the

649

The Real Price of Gold

increased exploitation of intensive bodies of ore in South Africa in the last fifty years as well as a policy of the South African government to subsidize the mining of lower quality ore. FLEISIG suggested a theoretical framework into which the parts of Rickoff's paper fit quite naturally-the standard model of finite-resource extraction. ROCKOFF responded by pointing out that his paper does discuss this class of models at the end of the section on the traditional commodity theory. Some very interesting propositions also emerge from the theory of durable commodities concerning the optimal intertemporal price path in particular. CAGAN argued that alternative gold standard rules may render the policy implications of Rockoffs paper irrelevant. He mentioned in particular Irving Fisher's idea of a compensated dollar, according to which the gold content of the dollar is changed gradually over a long period of time in order to stabilize the price level independently of the supply of gold. While acknowledging some problems with this approach, Cagan suggested that it was certainly worthy of discussion. Under certain conditions Fisher's rule provides a way to remove any long-run drift in the price level. In effect, the regime amounts to stabilizing the price level rather than the price of gold. Of course, this raises the question of whether there is any point to being on the gold standard.

15

The Image of the Gold Standard Leland B. Yeager

15.1

The Gold Standard from the Viewpoint of Prospective Adherents

Bills were pending in the parliaments of Austria and Hungary in 1892 to put those countries on the gold standard. In Austria, Deputy Anton Menger, a lawyer and brother of the economist Carl Menger, was chosen to sum up the pro-gold-standard position at certain stages of the debates. Anton Menger tried to refute the objection of the critics that gold, like the paper gulden, could be unstable in value-perhaps even more unstable. His economist brother, although also in favor of the gold standard, recognized that it had some "undeniable disadvantages," including the instability in the value of gold and especially its rise in value in recent decades. The paper gulden, he recognized, had been satisfactorily stable in value for domestic business (Menger 1936, p. 123).1 By the value of gold and value of the gulden, Carl Menger, as well as the actual critics of the gold standard, clearly meant purchasing power. Anton Menger evidently did not understand this. He ~uffered not merely from a money illusion but from a pound-sterling illusion specifically. Anyone can see how stable gold has been, he told critics, if they would take the trouble to look at a statistical table of its price on the London market from 1878 to 1891. He conceded that some fluctuations had occurred. Do you know how large these were? The greatest change amounted to 0.13 per cent. What a difference? In one year our money notes have changed in money-value by about ten per cent, while gold, over the course of a period of 12 to 13 years, has changed only by 0.13 per cent. Leland B. Yeager is professor of economics at the University of Virginia.

651

652

Leland B. Yeager

Under such circumstances there is no doubt at all that there can be no question at all of comparing banknotes with gold in regard to the function of measure of value. (Austria, Parliament, Chamber of Deputies 1892, p. 7182)2 Evidently, then, the economic understanding of his brother Carl had not automatically rubbed off onto Deputy Anton Menger. I thought I might say something tonight about the image of the gold standard in the late nineteenth century. I'll look at it from the standpoint of legislators, economists/ and pamphleteers in countries that were contemplating a move onto gold, namely, Austria-Hungary and Russia. These countries had fiat paper moneys, having been inflated off of their traditional silver standards several decades earlier by paper-money issues to cover the expenses of wars and revolutions. 15.2 Arguments Pro and Con Automatic regulation of the money supply was one of the advantages most strongly argued in Austria in favor of the gold standard. In countries with a sound monetary system, Carl Menger explained, money flows out if commodity prices rise and in if prices fall appreciably in relation to their normal level. It was a defect of Austria's isolated monetary system that flows of money and gold could not fill temporary gaps in the balance of payments, so that balance had to be maintained entirely in goods and securities. Domestic business suffered because its changing needs for money confronted an inadequately elastic paper-money supply. "We lack the mechanically operating, regulating influence of the inflow of money onto our markets; for this reason, we have apathetic, insensitive commodity markets; prices in Austria are not calculated precisely, as in England or Holland; we have apathetic markets, which paralyzes the spirit of enterprise" (Menger 1936, p. 294).4 Carl Menger recognized that the typical European currency was not a gold currency but a "gold-plated" one. It had a core of paper, surrounded by layers of minor coins and silver and then with an outer layer of gold plating. This arrangement was good enough for him, provided the plating was strong enough to resist the acid test of financial crisis. What he wanted was stable exchange rates (p. 247). In Hungary, Istvan Tisza expressed similar ideas: A paper-money country experiences no significant inflows and outflows of money. "In gold-standard countries, on the other hand, the size of the need for gold is decisive." Rich or populous countries need more media of exchange, poor or less populous ones need less. The relation between quantity and need must be such that in both places they are equal; and money even goes from the richest country to

653

The Image of the Gold Standard

the poorest if there is relatively greater need for it there, for then it can be turned to better account there; and the elements of the balance of payments will necessarily change as the relation between the quantity of and need for gold requires.... The tendency of balances of payments will always be . . . to equalize the relation between demand and supply in all countries and give all as much gold as they need in relation to the needs of others. (Tisza 1890, pp. 92-93) Its Austrian supporters saw the gold standard less as a transmitter of foreign disturbances than as a means of cushioning domestic disturbances by linkage with the presumably more stable world economy. Franz Perl wrote in 1887 that in the isolation in which our currency places us, we are left to our own resources whenever credit is shaken; that international flow of money which stands helpfully at the side of other money markets in times of need is lacking to us; our securities, which only in rare cases have a real abode abroad, return to us at the least sign of mistrust; we lack that equilibrating help. (Perl 1887, p. 64, citing Alfred von Lindheim) Deputy Anton Menger also believed that business crises were "very considerably intensified" by the inability of money to flow into and out of Austria, a monetary island. The value of Austrian money rested only on a very dangerous basis, its scarcity, for a country should have enough money and not too little (Austria, Parliament, Chamber of Deputies 1892, pp. 7182-83). Another deputy (Eim) pointed out that the value of a paper money depended on the need for money and on the amount in circulation. The latter could be controlled, but the need for money could hardly be calculated. Thus, the value of paper money "is subject to continual changes, which depend on the most various circumstances, often on chance, indeed even on speculation (p. 6989). The economist Julius Landesberger likewise saw it as a grave defect of a system of purely fiduciary money that it could not work well "unless it were continuously possible to ascertain most reliably the need of the whole economy for means of circulation at all times and to regulate the monetary circulation correspondingly. To this, however, the resources of science are not adequate today" (Landesberger 1892, p. 68).5 Russian supporters of gold also argued that that standard made a country's money supply appropriately elastic. Under a paper system, by contrast, the money supply supposedly did not respond appropriately and automatically to the changing need for means of circulation; yet it was impossible to calculate and deliberately meet that need. In a goldstandard country, though, a deficiency of the domestic money supply would remedy itself through a balance-of-payments surplus and an inflow

654

Leland B. Yeager

of gold, and a superabundance of money would remedy itself through a deficit and an outflow. Each country would automatically come to hold the quantity of metallic money appropriate to its wealth and transactions, without anyone's having to try to estimate the required quantity.6 Opponents of the gold standard sometimes argued that the sacrifices required to get onto gold would prove to have been in vain in case Austria-Hungary should get into another war. The progold reply was that the country should have hard money in peacetime to save the possibility of paper-money issues-the state's "note credit"-for wartime. With the country having a depreciated paper money even in time of peace, said Perl (1887, p. 29), every economist and patriot must shudder to think of what would happen in time of war or fear of war. In Russia, also, gold standard opponents argued that monetary reform would not be worth the trouble, since a new war would only make the paper money irredeemable again. The reply was that irredeemable paper money should be abolished now so that new issues could be put into circulation if the occasion arose. The currency reform could be a "reconstruction of war material" (Schultze-Gavernitz 1899, p. 462). Starting from the gold standard, the government would have better wartime financial alternatives than if it started from irredeemability. Early in 1879, when the world-market price of silver had sunk so low that the Austrian gulden was again worth as much as its supposed silver content, or even slightly more, the coinage of silver on private account threatened to inflate the money supply and price level. The Austrian and Hungarian governments responded by closing their mints to the free coinage of silver. That action had been taken in a legally very informal way, however, leaving the possibility that the silver standard would come alive again. For some years the gulden floated at a value above that of its supposed silver content. By 1890, a different aspect of the loose link remaining between the gulden and silver-one working through speculation about domestic redemption and coinage policy and American silver-purchase policycame to the fore, providing one of the strongest motives to reform. The Austrian financial press and Parliament seemed preoccupied with the progress of the Sherman Silver Purchase Bill in the U.S. Congress, and unusual day-to-day jumps in the price of silver and the gulden's exchange rate were generally attributed to news from Washington . . Finance Minister Steinbach warned Parliament on 14 May 1892 that f@rces supporting and opposing free coinage of silver in the United States were almost evenly balanced; powerful influences on the Austrian currency could come from that direction. "The rate fluctuations of the year 1890, which you all remember, gentlemen, have brought us a small foretaste of what would happen if silver coinage were made free today in

655

The Image of the Gold Standard

the United States of North America" (Austria, Parliament, Chamber of Deputies 1892, p. 5930). Another aspect of legal untidiness was the existence of four distinct types or concepts of gulden: (1) the ordinary fiat gulden ("gulden of Austrian currency"), in which currency and bank deposits were denominated and in which most prices and debts were expressed; (2) the silver gulden, in which some bonds and preferred stocks were still denominated and which could again become separated from the ordinary gulden if silver rose sufficiently in market value; (3) a gold gulden worth two-anda-half French francs, in which some bonds and customs payments were expressed; and (4) another gold gulden, worth 1.2 percent less, which had some slight application ingovetnment accounting; it was the gold equivalent of the standard silver gulden at the 15.5:1 bimetallic ratio of the Latin Monetary Union. As Josef Kreibig later observed, "if there was one drastic proof of the necessity of a reform, it was this peculiar splitting of the monetary unit" (Kreibig 1899, pp. 61-62). 15.3 Climates of Opinion

Dominant Hungarian interests switched in favor of gold around 188990. Earlier they had opposed it out of fear that it meant appreciation of the paper gulden to equality with the two-and-a-half-franc gold gulden, hampering agricultural exports. But as the Hungarians came·to realize that the gulden would not be pegged upward at that rate and that the gold standard might mean resistance to further appreciation, or even a partial reversal of recent appreciation, the sentiment of the country's exportand import-competing interests shifted.7 It seems that the experts, so considered by the Establishment, were almost all in favor of the gold standard. Being an expert (and so being invited to testify before the commissions mentioned in footnote 3) apparently presupposed, almost by definition, advocacy of the gold standard. None of the major Austrian political parties, as a party, opposed the gold standard, although many individual deputies did. Even proponents of the gold standard recognized that a large opposition existedand that opponents might possibly outnumber proponents-but outside the most influential circles. The masses had supposedly become accustomed to the existing currency situation and were apathetic about reform. Among the articulate, though, advocacy of gold dominated. A propaper pamphleteer suggested a version of the fable of the emperor's clothes: even people who did not understand the supposed disadvantages of paper money and the supposed advantages of gold nevertheless joined the progold chorus in order not to seem unenlightened (Gruber 1892, pp. 114-15).

656

Leland B. Yeager

Even so, opposition consumed most of the time in the parliamentary debates. This was understandable: the government's position took the form of definite bills, and only so much could be said in their favor without repetition, while opposition views were aired in great variety. Only a minority of the opponents forthrightly favored retaining fiat paper money. Most of them wanted bimetallism, or thought that the time was not ripe for the gold standard, or believed that action should await some sort of international agreement, or wanted a gold standard different from what the government's bills would introduce, or engaged in nit-picking about such issues as the emperor's titles on the new gold coins. The only amendment adopted was one expanding his titles from Imperator et Rex to a long list including King of Bohemia, King of Gallicia, and ending with Apostolic King of Hungary. Some pamphleteers did state the case for retaining a fiat paper money with floating exchange rates---a case centered around the greater importance of domestic than of exchange-rate stability and the importance of a measure of insulation from foreign deflation and crises. Josef Neupauer predicted that "a slow and steady increase in the means of circulation will without doubt encourage the spirit of enterprise, and all the more remain without influence on the price of the Austrian money as indeed the population grows and the whole economy develops." He proposed that the new money necessary to accompany real economic growth be put into circulation through purchase of securities on the Bourse. He even hazarded a guess about the proper rate of annual increase in the money supply-4 percent (Neupauer 1892, p. 26 and passim). Dominant trends of opinion were apparently quite different in Russia. The discussions of the Imperial Free Economic Society in St. Petersburg in March-April 1896 (cited in note 3) serve as evidence that advocacy of the gold standard was not part of the conventional wisdom among economists and leading thinkers. Even advocates of the gold standard acknowledged that apathy toward the reform was quite general. Schulze-Gavernitz referred to those discussions to justify his assertion that "the State carried out the currency reform against public opinion, with few exceptions, against the press, against the tough resistance of the public" (Schulze-Gavernitz 1899, pp. 461-62, 470-71). Finance Minister Sergei Witte also testified to the climate of opinion. As he said, nearly the whole of thinking Russia was initially opposed to his reform. Even he, while new in office, contemplated abandoning his predecessors' work of moving toward the gold standard.8 The opposition to the gold standard was so strong in Russia that in order to enact it, the Tsar had to bypass the usual legislative procedure, which involved various committees. Supporting his finance minister, the Tsar enacted it piecemeal by autocratic decrees.9

657

15.4

The Image of the Gold Standard

Noneconomic Motives

Before turning to the noneconomic reasons for adoption of the gold standard, particularly in Austria, I want to emphasize that the economic reasons did not include poor performance of the fiat paper currency (see Yeager 1969, pp. 61-89). Exchange-rate fluctuations were not e.xtreme by present-day standards, and the paper currency was not suffering price inflation. (In fact, the price trend had been downward since about 1871, though less steeply downward than in the gold standard world.) Yet Deputy Anton Menger complained. He said that importers and exporters were able to perceive seasonal tendencies in the exchange rate-very feeble tendencies, so far as the figures show-and profit from them by shrewdly timing their purchases and sales of foreign exchange. This sounds like stabilizing speculation to us-hardly grounds for complaint. Yet Menger implied, without articulating his complaint explicitly, that the gains of the shrewd traders were necessarily coming at the expense of the country as a whole (Austria, Parliament, Chamber of Deputies 1892, p. 7473). The gold standard would put a stop to that. Apart from the economics of the matter, the fluctuating exchange rate was widely viewed as a symbol of disorder and backwardness, whereas being on the gold standard-the most modern monetary system-was the mark of a civilized country. Vienna's leading newspaper deplored the monarchy's confused monetary system-with silver as the basic metal, with irredeemable paper notes in circulation, and with the gulden's value exposed to the vicissitudes of wild international speculation-"while all civilized states have long since assured themselves of a stable measure of value, a money as steady in value as possible" (Neue Freie Presse, 7 September 1890).10 Considerations of prestige were at work. In the Hungarian Currency Inquiry of 1892, Koloman Szell, a former finance minister and future prime minister, declaimed about "the stigma of a paper economy, unworthy of a civilized nation" (quoted in Gruber 1892, p. 117). The Currency Committee of the Austrian Parliament observed in 1892 that "considerations of state [had] influenced the decision of the government" to proceed with gold-standard legislation. Twenty years before, Austria had not been alone in using paper money; since then the United States, Italy, and even little Rumania had gone onto the gold standard. Russia, the only other major power still with a paper standard, was already making preparations for going onto gold. "Every year it detracts more from the State prestige of Austria that it still belongs to the countries with an unregulated currency" (Austria, Parliament, Chamber of Deputies 1892, Rei/age no. 491, p. 8). Deputy Dr. Foregger reminded his colleagues that the "scrap-of-paper

658

Leland B. Yeager

economy" degraded Austria economically to a second-rate power. "We demonstrate that our Empire does not ·have the strength to introduce among us, too, the means of payment, hard money, that holds sway in the civilized world. We thereby incessantly damage our credit, our economic flexibility and competitiveness." Lack of foreign confidence extended beyond the economic sphere into all other sides of our international relations; it lessens respect for us, esteem for us; it lessens our power position. We must therefore make all efforts to bring the strength of our Monarchy into full effect again by regulating our monetary system.... We cannot have a separate, an insular, currency continue: if we want to take part in the competition of civilized nations, we too must accept the international means of payment, and the international measure of value is just nowadays gold. (Pp. 7132-33) The "scrap of paper" to which Dr. Foregger alluded was itself a source of dismay. The state currency notes (as distinguished from the notes of the Austro-Hungarian Bank) were thought of as an actual debt to be paid off sooner or later. This view found support not only in linkage of the legally permissible quantities of state notes and treasury bills (Salinenscheine) under a ceiling on their combined amount but also in the inscription on the notes themselves, which acknowledged each note as "a part of the common floating debt of the Austro-Hungarian Monarchy" ("common" here meaning shared by the two governments). The term "floating debt" sounds more ominous in German than in Englishschwebende Schuld--eonveying the impression of a "hovering guilt" still to be expiated. One of the purposes of the monetary reform bills of 1892, the Austrian government said, was to abolish these state notes, which had been issued under the compulsion of "shattering political events" (Austria, Parliament, Chamber of Deputies, Beilage no. 436). The reference was to monetary inflation during the Austro-Prussian War of 1866. The yearning to banish an ever-present reminder of the humiliation of Koniggratz was an old one. On 1 November 1884 the Neue Freie Presse said that "redeeming the floating debt" was "an old duty of honor of Austria." On 1 January 1892 the newspaper lamented "the dismal legacy of revolution and wars, the irredeemable notes, these hateful stains on the name of Austria.... The paper gulden is ... [a] sad monument that has been erected in our budget to remind us of the sufferings of the past." Even the analytical Carl Menger "most decidedly" rejected "the opinion of those who deny Austria-Hungary the right to reshape her currency on the pattern of that of the civilized nations. It should not be interpreted as immodesty if we too wish to be counted among the 'nations les plus avancees dans la civilisation,' among the nations that are already 'ready for gold,' and not among the peoples 'of the other currency area,' which

659

The Image of the Gold Standard

should content themselves with silver currency" (Menger 1936, pp. 17273). Among its advocates in Russia, the gold standard "had become, in the midnineties, more than ever a matter of national respectability and economic advantage.... For Russia (as for any civilized country at that time) it was a prerequisite for sound credit and economic progress in general. Above all it would encourage more foreign investment in Russian industry" (Von Laue 1963, p. 139). A. N. Gurjev was one of the economists who held such a view. For him, restoration of the ruble to a metallic basis had political and cultural as well as economic significance: Membership in worldwide civilization is unthinkable without membership in the worldwide monetary economy. . . . A country with an isolated monetary economy cannot enter into stable cultural intercourse if it is separated from civilized peoples by the whole complex of economic evils connected with the disorder of the monetary system (Gurjev 1896, p. 163). Finally, we have the judgment of an eminent Austrian economist of a later generation. Modern economists will be quite unable to understand, said Joseph Schumpeter, why countries such as Austria-Hungary, Russia, and Italy imposed hardships on themselves to adopt gold parities for their currencies. No important economic interests clamored for that policy. Noneconomic considerations were decisive. Gold symbolized sound practice and honor and decency. "Perhaps this explanation raises more problems than it solves. That it is true is certain" (Schumpeter 1954, p. 770). 15.5 Some Reflections about This Conference Now I'll make a sharp change of course. Toward the end of my talk I'll return to the theme of the image of the gold standard. When Professor Michael Bordo invited me to speak this evening, he suggested that I might want to reflect on what we had learned from the conference papers and from the discussion. I was particularly glad to hear what Donald McCloskey and Richard Zecher have to say about purchasing-power parity. (I hasten to add that accepting purchasing-power parity does not necessarily imply accepting the extreme version of the monetary approach to the balance of payments, which identifies a balance-ofpayments deficit with a process of working off an excess supply of money and identifies a surplus with a process of satisfying an excess demand for money.) Purchasing-power parity is one of the few dependable generalizations that we have in economics. It is a generalization about the range within

660

Leland B. Yeager

which a floating exchange rate tends to fluctuate and about what relation has to hold between prices in different countries if a fixed exchange rate is to remain workable. Purchasing-power parity should not be understood to deny that other factors influence exchange rates or international price relations, including specific historical factors such as the "country risk" of investment in Italian securities that Michele Fratianni and Franco Spinelli spoke of. Yet there seems to be a passion among more than a few economists to deny or question our most dependable generalizations, perhaps by interpreting them in so exaggerated and rigid a way that they are not strictly true. Other examples are the treatment often accorded to the quantity theory of money and the marginal-productivity theory of the demand for labor and other factors. Overeager critics need to be warned against what Thomas Sowell (1980, pp. 291-92, 324) calls the precisional fallacy. I was glad to hear McCloskey and Zecher ask, when told about the failure of purchasing-power parity: Failure relative to what? Failure by what standards? Failure relative to what alternative theory? (Here I am embroidering a bit.) In dealing with their respective subject matters, what theories are more satisfactory than the quantity theory or purchasing-power parity or marginal-productivity theory? Do we really want to say that the price level (in a country with a floating exchange rate) depends on all sorts of influences, among which the quantity of money plays no distinctive role? Do we want to say that the floating exchange rate between two currencies depends on supply and demand, which reflect the influences of all sorts of actual and potential transactions, and that nothing more definite can be said? Would we want to say that no meaningful generalization can be offered about the relation between price levels in different countries under a regime of fixed exchange rates? Agreed, all sorts of influences, some more and some less bound to specific historical circumstances, do affect a country's price level, the relation among different countries' price levels, and the ranges within which free exchange rates tend to fluctuate. But we are asking about theories-generalizations. What better generalizations are available to replace the quantity theory and the purchasing-power-parity doctrine? Unless one can answer that, talk about their "failure" is premature. McCloskey and Zecher took some swipes at what they call closedeconomy theorizing. I hope they do not mean that the acceptability of purchasing-power parity actually precludes any such theorizing. For at one stage of presenting monetary theory, it is useful and legitimate to assume a closed economy or what for the purpose at hand amounts to the same thing, namely, an economy with an independent currency and a floating exchange rate. We see why and how the supply of and demand for money affect nominal income and the price level.ll The discussion prepares us for understanding the interrelations among money supply, money demand, and the balance of payments under fixed exchange rates.

661

The "Image of the Gold Standard

I hope that accepting purchasing-power parity does not require us to assume that prices in each particular country under fixed exchange rates are simply dictated to it by the outside world. How, then, would the world price level be determined? How would gold discoveries under the gold standard affect world prices? It would seem odd to maintain that the money supply follows and responds to the externally dictated price level in each particular country, although the money supply leads and determines the price level in the world as a whole. To repeat, closed-economy theorizing, or something closely akin to it, does have a legitimate place in explaining relations between money supplies and price levels. But open-economy theorizing is also necessary. Some of the conference papers and commentaries have made me wonder whether there isn't too much polarization knocking around-too much discussion of possibly parallel or alternative economic processes or aspects of processes as if descriptions of them were the mutually contradictory assertions of rival theories. An example is the supposed rivalry between Hume's theory of price-level shifts, which would hold true in a case of slight substitutability between domestic and import and export goods, and the theory of Adam Smith and modern devotees of the monetary approach to the balance of payments regarding how the relation between money supply and money demand affects spending and the balance of payments directly, even apart from relative price shifts. The monetary approach, sensibly interpreted, does not say (as Mordechai Kreinin and Lawrence Officer, quoted in Michele Fratianni and Franco Spinelli's paper, interpret it as saying) that the famous elasticities of the elasticity approach to balance-of-payments analysis are irrelevant. Rather, the extreme monetary approach, together with its related assertion that the law of one price holds strictly, carries certain implications about the elasticities: they are extremely high; goods are highly substitutable. In reality, of course, the law of one price does not hold strictly, and the elasticities are not extremely high; but contemplating the unreal polar case can still be instructive. I also suspect polarization between supposedly rival theories in some of the "testing" we have seen for observance of the rules of the goldstandard game. As Fratianni and Spinelli say, an identification problem is . involved: Are changes in the domestic component of the monetary base a response to or a cause of changes in the international component? Does causation run/rom or to domestic assets? Shouldn't distinctions be made between long-run and short-run behaviors with regard to rules of the game? I'd like to echo Heywood Fleisig's call for clearness about what disturbance is being supposed and about what is being taken as exogenous and endogenous. I conjecture that differences among what may appear to be different theories-eoncerning, for example, Hume's price-specie mechanism, direct-spending effects, the law of one price, the direction of causality

662

Leland B. Yeager

between changes in domestic and international components of a country's monetary base, the direction of causality between price levels and exchange rates, and different mechanisms of balance-of-payments adjustment-do not so much indicate contradictions (to be settled by empirical research) as, rather, reflect different assumptions and different scenarios. The remedy for such apparent contradictions is to layout the questions at issue precisely: What happens if such-and-such circumstances prevail and if such-and-such a disturbance occurs? When we bring historical facts and figures into a discussion about supposedly rival theories, we should try to be clear about what questions the facts and figures are meant to shed light on. I suspect that some of the papers and comments at this conference have failed to distinguish sharply enough between (1) using economic theory as a tool of historical research-as an aid to gathering, sorting, organizing, and interpreting historical facts-and (2) appealing to history to discriminate between correct and incorrect economic theories. The facts underlying or informing economic theory "ought" to be more basic, dependable, and enduring and more firmly rooted in human nature and the human condition than the contingent facts of specific historical conditions and episodes. I am suspicious of looking to such contingent facts to settle basic theoretical issues.

15.6 The Seductive Appeal of the Gold Standard

Now, in preparation for coming to a conclusion, I want to return to my earlier themes. These themes concern the appeal or the desirability of the gold standard. As we know, some prominent economists and politicians nowadays are recommending a return to the gold standard-or the adoption of whatever it is that they are marketing under that label. My response is not that we must not turn back· the clock. That hackneyed slogan betrays a provincialism about one's own time, a shallow meliorism, a moral futurism. Nor is my message that we can't turn back tl)e clock. Rather, my message is a reminder of what it is that we would have to turn back to. It is a reminder of the entire situation in which the gold standard flourished. More exactly, perhaps, the gold-standard world is an idealized past state of affairs. The few, very few, decades during which the international gold standard flourished offered almost uniquely favorable conditions. Mint pars among gold standard currencies, instead of being arbitrarily chosen, expressed an equilibrium that had evolved gradually between themselves and national price levels. Mildly rising world prices after 1896 facilitated relative adjustments of prices and wages, while the uptrend did not last long enough-until war destroyed the system-to dissipate its possible

663

The Image of the Gold Standard

benefits by becoming embodied in expectations. Relative calm in social and political affairs and the absence of excessively ambitious government programs and excessive taxation all favored confidence in monetary stability. The age of the gold standard was an age of peace, relatively.12 Hugh Rockoff suggested that the tolerably good performance of the gold standard before World War I hinged on favorable conditions that no longer prevail: a corps of dedicated gold prospectors working in unexplored areas; absence of political interference (a laissez-faire atmosphere); patience with the long and uncertain lags in the response of the gold supply to the changing demand for money. By and large, people (in countries that happened to be on the gold standard, anyway) were freer from government control than in any age before or since-freer to transact business, to make investments, to transfer funds, to travel. There is a certain charm in the reminiscences of an old German banker of how, during his student days at Heidelberg, he and some friends, one of whom had just come into an inheritance, left on the impulse of the moment for a tour of Italy, where the banker in the first town they stopped at considered it an honor to cash in gold coin the large check written by the young stranger. There is similar charm in Jules Verne's story of Phineas Fogg, who left on short notice for his eighty-day tour of the world, paying his expenses from a carpetbag full of Bank of England notes, accepted everywhere. The civility and internationality prevalent during the age of the gold standard have such charm for us nowadays that it seems almost sacrilege to ask whether these benefits resulted from the gold standard or, instead, coexisted with it by mere coincidence. The gold standard, in short, evokes the "good old days." This association is well illustrated by two quotations, the first from Benjamin M. Anderson, a lifelong champion of gold, and the second from John Maynard Keynes, his generation's leading critic of that standard. Those who have an adult's recollection and an adult's understanding of the world which preceded the first World War look back upon it with a great nostalgia. There was a sense of security then which has never since existed. Progress was generally taken for granted.... We had had a prolonged period in which decade after decade had seen increasing political freedom, the progressive spread of democratic institutions, the steady lifting of the standard of life for the masses of men.... In financial matters the good faith of governments was taken for granted.... No country took pride in debasing its currency as a clever financial expedient. (Anderson 1949, pp. 3-4, 6) What an extraordinary episode in the economic progress of man that age was which came to an end in August, 1914! ... [A]ny man of capacity or character at all exceeding the average [could escape from the working class] into the middle and upper classes, for whom life

664

Leland B. Yeager

offered, at a low cost and with the least trouble, conveniences, comforts and amenities beyond the compass of the richest and most powerful monarchs of other ages. The inhabitant of London could order by telephone, sipping his morning tea in bed, the various products of the whole earth, in such quantity as he might see fit, and reasonably expect their early delivery upon his doorstep; he could at the same moment and by the same means adventure his wealth in the natural resources and new enterprises of any quarter of the world, and share, without exertion or even trouble, in their prospective fruits and advantages.... He could secure forthwith ... cheap and comfortable means of transit to any country or climate without passport or other formality, could despatch his servant to the neighboring office of a bank for such supply of the precious metals as might seem convenient, and could then proceed abroad to foreign quarters, without knowledge of their religion, language, or customs, bearing coined wealth upon his person, and would consider himself greatly aggrieved and much surprised at the least interference. But, most important of all, he regarded this state of affairs as normal, certain, and permanent, except in the direction of further improvement, and any deviation from it as aberrant, scandalous, and avoidable. (Keynes 1920, pp. 10-12) Reminiscences like these reinforce my impression that the outbreak of World War I was a momentous turning point and a great tragedy in the history of the world-a tragedy all the more poignant because the war broke out so accidentally. The building in Sarajevo near which the assassin was standing when he fired the fateful shots bears a plaque saying that here, on 28 June 1914, Gavrilo Princip carried out an act expressing resistance to tyranny and the will to freedom. The inscription says nothing about the initiation of a chain of events that may, even yet, carry to the destruction of Western civilization. It says nothing about the start of our present age of wars, of globally expansionist tyrannies, and of the perversion of democratic government into an instrument whereby each interest group seeks to plunder society in general, to the unintended net loss of practically all. If I were asked for my recommendation, therefore, I would not merely recommend going back to the gold standard. By itself, apart from restoration of its preconditions, that would hardly be a constructive step. My nostalgia is for the whole pre-1914 climate, not for one specific facet of it. I recommend repealing World War I, root and branch. If only we could! 15.7 Attitudes Necessary for Sound Money Repealing World War I would have to include restoring certain attitudes that seem to have been more prevalent in public affairs before 1914 than they are now. Those attitudes favored limitations on the scope of government activity and restraint on seeking special advantage through

665

The Image of the Gold Standard

the instrumentality of government. Broadly speaking, these were liberal attitudes in the nineteenth-century sense. These attitudes have now been undermined in ways analyzed, in part, by Ortega y Gasset in The Revolt of the Masses. Nowadays, we have tyranny in the nondemocratic countries and, in the democratic countries, democracy perverted in such a way that political decisions are made out of short-run expediency and without due regard for long-run consequences. But in the gold-standard era, as Lars lonung says, "the democratic system had not beep. fully developed." (Peter Lindert detects signs of the perversion of democracy in the United Kingdom, however, even before World War I.) Without a return to liberal attitudes and self-restraints, a restored gold standard would not work well and would hardly endure. After all, the gold standard is simply a particular set of rules for policy regarding the monetary system; and these rules are no more inherently self-enforcing than any other set of monetary rules. Michele Fratianni has been telling us of the readiness of Italian politicians to throw out the gold standard, and Peter Lindert has noted the propensity of the gold standard and key-currency systems to collapse when shocked. (Even today, before we have gone back to a supposed gold standard, there is plenty of reason for suspecting that what ~f\me of its supporters are advocating is not a real but a pseudo gold standard, to echo a distinction made by Milton Friedman [1961, pp. 66-79].) Maybe some hope is to be found in constitutional restraints on government taxing and spending, maybe in the depoliticization of money. It would be outside my assignment to discuss these possibilities tonight. My purpose, rather, has been to set our examination of the classical gold standard into the context of the conditions and attitudes that apparently prevailed at the time. Given the required attitudes and the related restraints on government, the gold standard is not the only set of monetary arrangements that would function tolerably well. Economists can easily imagine, and have proposed, monetary arrangements that would function better. The required attitudes were illustrated in Austria even while the country was still on fiat paper money. The government and the financial press repeatedly agonized even over budget deficits that would seem delightfully small to us today. Although the price level was generally steady or even trending mildly downward (except during wars and immediately afterward), the government and the press worried about the value of money as reflected in the exchange rate. (Nowadays, attention would more suitably focus on a price index.) The Neue Freie Presse took exchange rates of 120 guldens or higher for ten pounds sterling as a particularly ominous warning. I will conclude with three quotations from that newspaper.

666

Leland B. Yeager

London: 120! A cannon shot cannot shock us more than this figure; and it also forms an urgent warning for the many finance ministers of the Monarchy to maintain moderation, to retrench, to resume the policy of soundness.... When the exchange rates, this manometer of credit, rise, then it is better to reef in the sails a bit. Is it really our fate eternally to bear the mark of shame of a disordered currency? Will there never come a chancellor of the treasury who will have the will and also the power to restore the most important basis of the economy? (14 October 1883) The price of foreign bills is the loudest and gravest accusation against the government. . . . [When the opposition parties] want to depict the sad condition of the state with one stroke, then they need only unfold the Cursblatt [sic] and say: Things have gone pretty far in Austria when one franc equals half of our gulden on the world market. ... what the ghosts were for poor Macbeth, the foreign exchanges are for [Finance Minister] Dunajewski; indeed, we are convinced that he often wakes up at night, terrified, and suddenly perceives a figure before him that mockingly hisses at him: London 126.50! (26 April 1885) To introduce my final quotation, I should explain that the Austrian police from time to time confiscated issues of publications containing articles considered too critical of the government. The Neue Freie Presse occasionally carried a notice on its front page saying that its preceding issue had been confiscated. (To compensate its subscribers, the newspaper would either reprint the confiscated issue without the offending material or else make the next issue especially large.) In one of its editorials denouncing the confiscations, the paper complained about discrimination, as well: Unlike itself, the official Coursblatt of the Vienna Bourse had never been confiscated. Yet its latest issue quoted London exchange at 120.95. "And if we were to write our fingers sore, we could not portray the situation more precisely. Cqnfiscate the Cursblatt [sic], Mr. Attorney General" (26 February 1882).

Notes 1. Cf. Menger 1936, pp. 147, 196, 233-34. 2. The quotation is from the 14 July 1892 transcript. Menger had previously used the same argument on 25 May, p. 6192. 3. These include some of the expert witnesses testifying before commissions convoked in Vienna and Budapest in March 1892 and in St. Petersburg in March-April 1896. See Austria, Wahrungs-Enquete Commission 1892; Imperatorskoe Voljnoe .Ekonomicheskoe Obshchestvo 1896. 4. Cf. Menger 1936, pp. 138-39, 226--27. 5. Landesberger thus seemed to imply that if the supply of a fiat money could be regulated appropriately, exchange-rate fluctuations would not count decisively against that

667

The Image of the Gold Standard

system. Some people, he noted, even considered the fluctuations a desirable insulator against price deflation in gold countries. 6. These arguments are reported in Vlasenko 1963, pp. 85-86; Raffalovich 1896, p. 369; Trakhtenberg 1962, pp. 174ft; and Finance Minister Witte's bill to authorize contracts in gold currency, quoted in Saenger 1927, p. 16. 7. Vienna Board of Trade 1887, p. 388; Kamitz 1949, pp. 147-48; Aktionar, 22 June 1890, first supplement, dispatch from Prague; Silin 1913, pp. 394, 395, 399, quoted and paraphrased at length in Trakhtenberg 1962, pp. 265-66. Tisza (1890, esp. pp. 93-95) explained the incorrectness of the earlier fears and argued that the gold standard would serve Hungarian interests. 8. Witte 1921, pp. 59-60; cf. Von Laue 1963; Crisp 1967, p. 211; Migulin 1899-1904, pp. 130-31. 9. Witte 1921, pp. 59,61; Von Laue 1963, pp. 141-44; Migulin 1899-1904, pp. 284-86; Trakhtenberg 1962, p. 267; Russia, Finance Ministry 1902, 2: pp. 422-25. 10. Earlier (1 November 1884) the same newspaper had exclaimed, "What enthusiasm it would stir up if at last the warmly longed-for moment had arrived to raise Austria onto the height of the civilized states!" 11. This process was so clearly described by Knut Wicksell back in 1898 that I propose calling it the "Wicksell process" (to be distinguished from his "cumulative process," which is rather different). See Wicksell 1965, pp. 39-41. 12. The Neue Freie Presse (Vienna) and Aktionar (Berlin), both evident organs of liberal bourgeois thought, repeatedly stressed that peace was good for business.

References Aktioniir. Berlin. Anderson, Benjamin M. 1949. Economics and the public welfare. New York: Van Nostrand. Austria. -Parliament. Chamber of Deputies. 1892. Stenographisches Protokoll, XIth session. - - - . Wiihrungs-Enquete Commission. 1892. Stenographische Protokolle uber die vom 8. bis 17. Miirz abgehaltenen Sitzungen. Vienna: K. k. Hof- und Staatsdruckerei. Crisp, Olga. 1967. Russia, 1860-1914. In Banking in the early stages of industrialization by Rondo Cameron et al. New York: Oxford University Press. Friedman, Milton. 1961. Real and pseudo gold standards. Journal ofLaw and Economics 4 (Oct.): 66-79. Gruber, Robert. 1892. Nationales oder internationales Geld? Die Quintessenz der Wiihrungsfrage. Vienna: Lesk und Schwidernoch. Gurjev, Aleksandr Nikolaevich. 1896. Reforma denezhnago obrashchenija. St. Petersburg: Kirshbaum. Imperatorskoe Voljnoe .Ekonomicheskoe Obshchestvo. 1896. Reforma denezhnago obrashchenija. Reprinted from the society's Trudy 3. St. Petersburg: Demakov.

668

Leland B. Yeager

Kamitz, Reinhard. 1949. Die osterreichische Geld- und Wahrungspolitik von 1848 bis 1948. In Hundert Jahre osterreichischer Wirtschaftsentwicklung, 1848-1948, ed. Hans Mayer. Vienna: Springer. Keynes, John Maynard. 1920. The economic consequences of the peace. New York: Harcourt, Brace & World. Kreibig, Josef Clemens. 1899. Unser Wiihrungs- und Munzwesen wiihrend der letzten funfzig Jahre. Vienna. (First published in Osterreichische-Ungarische Revue 24, nos. 1-6.) Landesberger, Julius. 1892. Uber die Goldpriimien-Politik der Zettelbanken. Vienna: Manz. Menger, Carl. 1936. Collected works of Carl Menger. Vol. 4, Schriften iiber Geldtheorie und Wiihrungspolitik. London: London School of Economics and Political Science. (First published in 1893.) Migulin, P. P. 1899-1904. Russkij gosudarstvennyj kredit, 1769-1899. Vol. 3-2. Kharkov: Pechatnoe Delo. Neue Freie Presse. Vienna. Neupauer, Josef Ritter v. 1892. Die Schiiden und Gefahren der Valutaregulirung fur die Staatsfinanzen, die Volkswirthschaft und die Kriegsbereitschaft. Vienna: Lesk und Schwidernoch. Ortega y Gasset, Jose. 1930. La rebeli6n de las masas. Madrid: Revista de Occidente. Perl, Franz. 1887. Zur Frage der Valutaregulirung in Oesterreich-Ungarn. Zurich: Verlags-Magazin. Raffalovich, Arthur. 1896. Historie du rouble-credit. Journal de la Societe de statistique de Paris 37 (Oct.): 369. Russia. Finance Ministry. 1902. Ministerstvo finansov, 1802-1902. St. Petersburg: .Ekspeditsija Zagotovlenija Gosudarstvennykh Bumag. Saenger, Max. 1927. Die Wittesche Wiihrungsreform. Vienna and Leipzig: Deuticke. (First published in Zeitschrift fur Volkswirtschaft und Sozialpolitik, n.s. 5, nos. 10-12.) Schulze-Gavernitz, Gerhard von. 1899. Volkswirtschaftliche Studien aus Russland. Leipzig: Duncker & Humblot. Schumpeter, Joseph. 1954. History of economic analysis. New York: Oxford University Press. Silin, N. 1913. Avstro-vengerskij bank. Moscow. Sowell, Thomas. 1980. Knowledge and decisions. New York: Basic Books. Tisza, Istvan. 1890. Valutank rendezesef6l. Budapesti Szemle 62 (no. 160). Trakhtenberg, Iosif Adoljfovich. 1962. Denezhnoe obrashchenie i kredit pri Kapitalizme. Moscow: Izdateljstvo Akademii Nauk SSSR. Vienna Board of Trade (Handels- und Gewerbekammer in Wien). 1887. Bericht uber den Handel, die Industrie und die Verkehrsverhiiltnisse in Nieder-Oesterreich wiihrend des Jahres 1886. Vienna.

669

The Image of the Gold Standard

Vlasenko, Vasilij Evtikhievich. 1963. Teorii deneg v Rossii, Konets XIXdooktjabrskij period XX v. Kiev: Izdateljstvo Kievskogo Universiteta. Von Laue, Theodore H. 1963. Sergei Witte and the industrialization of Russia. New York: Columbia University Press. Wicksell, Knut. 1965. Interest and prices. Trans. R. F. Kahn. Reprint. New York: Augustus M. Kelley. Witte, Sergej Juljevich. 1921. The memoirs of Count Witte, ed. and trans. Abraham Yarmolinsky. Garden City: Doubleday, Page. Yeager, Leland. 1969. Fluctuating exchange rates in the nineteenth century: The experiences of Austria and Russia. In Monetary problems of the international economy, ed. R. A. Mundell and A. K. Swoboda. Chicago: University of Chicago Press.

Participants

Moses Abramovitz Department of Economics Stanford University Stanford, California 94305 Robert J. Barro Department of Economics Univeristy of Chicago Chicago, Illinois 60637 Daniel K. Benjamin U.S. Department of Labor 200 Constitution Avenue Washington, D.C. 20210 Michael D. Bordo Department of Economics University of South Carolina Columbia, South Carolina 29208 Karl Brunner Graduate School of Management University of Rochester Rochester, New York 14627 Phillip Cagan Department of Economics Columbia University New York, New York 10027

671

Carolyn Clark Department of Economics & Business Adminstration Washington State University Pullman, Washington, 99136 Lindley Clark Wall Street Journal 22 Cortlandt Street New York, New York 10007 Michael Connolly Department of Economics University of South Carolina Columbia, South Carolina 29208 Rudiger Dornbusch Department of Economics Massachusetts Institute of Technology Cambridge, Massachusetts 02139 John Dutton Department of Economics & Business North Carolina State University Raleigh, North Carolina 27650

672

Participants

Stephen T. Easton Department of Economics Simon Fraser University Burnaby, British Columbia Canada V5A 156

Milton Friedman Hoover Institution Stanford University Stanford, California 94305

Barry Eichengreen Department of Economics Harvard University Cambridge, Massachusetts 02138

Charles A. E. Goodhart Bank of England Threadneedle Street London EC2R 8AH England

Stanley L. Engerman Department of Economics University of Rochester Rochester, New York 14627

Paul Gregory Department of Economics University of Houston Houston, Texas 77004

Martin M. G. Fase De Nederlandsche Bank N.B. Postbus 98 100 AB Amsterdam The Netherlands Heywood Fleisig World Bank Economic Analysis and Projections Department 1818 H Street, N.W. Washington, D.C. 20433 Michele Fratianni School of Business Indiana University Bloomington, Indiana 47405

C. Knick Harley Department of Economics University of Western Ontario London, Ontario Canada N6A 5C2 Wallace E. Huffman Department of Economics Iowa State University Ames, Iowa 50010 Roy W. Jastram School of Business Administration University of California Berkeley, California 94720 Lars Jonung Department of Economics University of Lund S-220 05 Lund Sweden

Charles Freedman Department of Monetary and Financial Analysis Bank of Canada 245 Sparks Street Ottawa, Ontario Canada KIA OG9

Robert E. Keleher Research Department Federal Reserve Bank of Atlanta Atlanta, Georgia 30303

Jacob A. Frenkel Department of Economics University of Chicago Chicago, Illinois 60637

Levis A. Kochin Department of Economics University of Washington Seattle, Washington 98195

673

Participants

Leslie Lenkowsky Smith Richardson Foundation, Inc. 210 East 86th Street, 6th Floor New York, New York 10028 Peter H. Lindert Department of Economics University of California Davis, California 95616

Donald E. Moggridge Department of Economics University of Toronto West Hill, Ontario Canada M1C 1A4 Robert A. Mundell Department of Economics Columbia University New York, New York 10027

Robert E. Lipsey National Bureau of Economic Research 269 Mercer Street New York, New York 10003

John Pippenger Department of Economics University of California Santa Barbara, California 93106

James R. Lothian Citicorp Investment Bank 399 Park Avenue New York, New York 10043

Georg Rich Research Department Schweizerische Nationalbank 8022 Zurich Switzerland

Robert N. McCauley Research Department Federal Reserve Bank of New York 33 Liberty Street New York, New York 10045

Hugh Rockoff Department of Economics Rutgers College New Brunswick, New Jersey 08903

Donald N. McCloskey Department of Economics University of Iowa Iowa City, Iowa 52240

Anna J. Schwartz National Bureau of Economic Research 269 Mercer Street New York, New York 10003

Paul McGouldrick Department of Economics State University of New York Binghamton, New York 13901

Ronald A. Shearer Department of Economics University of British Columbia Vancouver, British Columbia Canada V6T 1Y2

Allan H. Meltzer Graduate School of Industrial Administration Carnegie-Mellon University Pittsburgh, Pennsylvania 15231

Richard E. Sylla Department of Economics & Business North Carolina State University Raleigh, North Carolina 27650

674

Participants

Peter Temin Department of Economics Massachusetts Institute of Technology Cambridge, Massachusetts 02139 Brinley Thomas Department of Economics University of California Berkeley, California 94720 Richard H. Timberlake, Jr. Department of Banking and Finance University of Georgia Athens, Georgia 30602 *Robert E. Weintraub Joint Economic Committee 6313 Dirkson Senate Office Building Washington, D.C. 20510

Lawrence H.White Department of Economics New York University New York, New York 10003 Geoffrey E. Wood Centre for Banking & International Finance City University Business School Frobisher Crescent Barbican London EC24 8HB England Leland B. Yeager Department of Economics University of Virginia Charlottesville, Virginia 22901 J. Richard Zecher Chase Manhattan Bank 1 Chase Manhattan Plaza New York, New York 10081

*Robert E. Weintraub died 12 September 1983.

Author Index

Angell, J. W., 65-67 Arak, Marcelle, 524 Ashton, T. S., 266 Ayres, Leonard, 465 Bagehot, Walter, 26, 45-47, 95 Bairoch, Paul, 367, 400 Barro, Robert, 615 Beach, W., 59-60, 553 Beckhart, B., 548 Bengtsson, Tommy, 387 Benjamin, Daniel K., 597 Beveridge, Steven, 561 Blainey, Geoffrey, 626 Bloomfield, Arthur I., 7, 88-89, 93-94, 96-97, 174-75, 195-97, 217, 315, 38284, 419, 553 Bonomo, Victor, 552 Bopp, Karl, 343 Borchardt, Knut, 329 Bordo, Michael, 3, 13, 174,613-14,645 Borts, George, 116, 549 Box, George, 181 Brainard, William, 331 Brisman, Sven, 380 Brown, William A., 73 Bryce, R. B., 566 Burns, Arthur F., 465 Cagan, Phillip, 330, 376, 479, 498, 604, 613, 647 Cairncross, A. K., 93, 115,549 Cairnes, John E., 43-44

675

Calliari, Sergio, 438 Calvo, Guillermo, 620 Cantillon, Richard, 31-33 Carr, R. M., 549 Cassel, Gustav, 75-76, 393-96 Chambers, Edward J., 552 Chandler, Lester V., 351 Chisholm, Derek, 195,277 Clapham, Sir John, 174, 177,471-72 Clarke, Steven, 351 Collins, Michael, 471 Courchene, T., 279, 293 Curtis, C. A., 282, 287, 555 Dam, Kenneth, 2 Dane, Ezra G., 625 Darby, Michael, 459, 490 Davidson, David, 368, 369 Deutsch, J. J., 282, 287 Dornbusch, Rudiger, 8 Dutton, John, 176 Edie, Lionel D., 622 Eliot, G. A., 279, 293, 295 Fisher, Irving, 14,27,51-53,455,587, 590, 615 Fleetwood, Erin, 379 Flink, Salomon, 314, 335, 344 Flodstrom, Isidor, 365 Ford, A. G., 83-84, 176, 197, 553 Frankel, S. Herbert, 645, 647 Fratianni, Michele, 410

676

Author Index

Frenkel, Jacob, 8, 90, 137 Friedman, Milton, 5, 86-88, 126, 140, 157-61, 268, 345, 351, 376, 418, 423, 458, 472, 474-75, 498, 553, 613, 615, 619-20,637 Gallman, Robert E., 465 Gandolfi, Arthur, 140 Genberg, A. Hans, 140 Girton, Lance, 427 Goodhart, Charles A. E., 94-96, 176, 178-80, 196, 203-5, 222-23, 553 Goschen, George J., 44-45 Graham, Frank D., 63-64 Granger, Clive, 519, 588 Gray, James, 626 Graybill, F. A., 135 Gregory, T. E., 72, 250-51 Grieves, R., 445 Habakkuk, H. J.~ 116 Hardy, Charles 0., 621, 635 Hartland, Penelope, 555 Hawtrey, Ralph G., 70-72, 553, 605 Hay, Keith, 552-53 Hecksher, Eli, 368, 381 Hicks, Sir John, 268, 605 Hirsch, Fred, 636 Hoffmann, Walther, 333, 342 Holborn, Hajo, 329 Horsefield, J. K., 269 Hughes, J. R. T., 266, 268 Hume, David, 33-35 Hyndman, H. M., 234 Ingram, James C., 549, 577 Jastram, Roy W., 622 Jenkins, G. M., 181 Jevons, William Stanley, 13, 633, 635 Johansson, Osten, 367 Johnson, Harry, 90 Johnson, J. F., 548 Jonung, Lars, 365, 368, 377, 380, 393-96 Jorberg, Lennart, 367, 387-88 Kang, Heejoon, 597 Kelley, Robert L., 628, 634 Keynes, John Maynard, 27, 74, 78-81, 114-15, 157, 233, 278, 477-78, 587, 605 Kindleberger, Charles, 351, 498 King, W. T. C., 269

Kitchin, Joseph, 622 Knox, F. A., 282, 287 Kochin, Levis, 597 Kravis, Irving B., 135, 154 Kreinen, Mordecai, 405-6 Krugman, Paul R., 133, 136 Kuhn, Thomas, 30 Laidler, David, 438 Lebergott, Stanley, 329 Letcher, Owen, 629 Levhari, David, 619 Lindert, Peter, 94-95, 173, 384 Lipsey, Robert E., 135, 154 Lothian, James, 140, 477 Lucas, Robert E., 513, 524, 525, 541 Macaulay, F. R., 605 McCloskey, Donald, 4-5,90-91,97-98, 116, 387, 524, 529, 553 McGouldrick, Paul, 323, 343 McIvor, R. C.., 278 MacLeod, Henry Dunning, 246-47 Mann, Golo, 329 Marshall, Alfred, 13, 48-51, 615 Matthews, R. C. 0.,466-67 Meese, Richard, 425 Meier, Gerald M., 549 Menger, Carl, 651-52 Michaely, M., 6 Mill, John Stuart, 40-43, 235, 243, 615 Mints, Lloyd, 175 Mintz, lIse, 561 Mitchell, Wesley, 465 Moggridge, D. E., 331 Mood, Alexander M., 135 Morgan, E. Victor, 177 Morgenstern, Oskar, 91-92, 196, 332, 339, 342, 512, 542, 552 Morrell, W. P., 625, 626 Mundell, Robert, 90 Nelson, Charles, 561, 597 Newbold, P., 588 Neyman, Jerzy, 134 Nurkse, Ragnar, 69, 175, 195,382,419 Officer, Lawrence, 405-6 Ohlin, BertH, 114 Paul, Rodman W., 625 Pearson, E. S., 134

677

Author Index

Pesmazoglu, J. S., 96 Pindyck, Robert, 619 Pippenger, John, 175

Report of the Commission on the Role of Gold, 182 Ricardo, David, 35-38, 615 Rich, George, 553 Ridgway, R. H., 623 Rogoff, Kenneth, 425 Roll, Richard, 128 Roper, Don, 427 Rosenberg, Carl M., 368-69 Rosenbluth, Gideon, 566 Rush, Mark, 644 Saidi, Nasser, 459 Samuelson, Paul, 114 Sandberg, Lars, 401 Sargent, Thomas, 180, 519 Sayers, Richard S., 7, 94-95, 174-75, 177-78, 197, 332, 343, 475 Schumpeter, Joseph, 659 Schwartz, Anna, 5, 86-88, 126, 140, 15861, 268, 345, 351, 376, 418, 423, 472, 474-75, 498, 553, 613, 637 Shearer, Ronald, 548 Sheehan, R. G., 445 Shiller, R., 587, 590, 605 Siegel, Jeremy, 587, 590, 605 Sims, Christopher, 480, 519 Smit, J. C., 73 Spinelli, Franco, 410, 437-38 Stewart, Marion, 619 Stockman, Alan, 459, 490 Stovel, J. A., 549 Summers, Lawrence, 607

Supino, Camillo, 415 Tanner, J. Ernest, 552 Tatom, John A., 622 Taussig, F. W., 55-59, 123,552-53 Temin, Peter, 470 Thomas, Brinley, 7, 93-94, 269, 515, 539, 542 Thornton, Henry, 38-40 Tinbergen, Jan, 96 Tobin, James, 331 Triffin, Robert, 31, 88, 169, 196 U.S. National Monetary Commission, 178, 331, 334-36, 342 Verga, Giovanni, 438 Viner, Jacob, 33, 45, 6(k)1, 76, 116,277, 343, 547-50, 555 Warburton, Clark, 466, 469 Ward-Perkins, C. N., 266, 268 Whale, P. B., 81-82, 333, 335 White, Harry D., 61-63 White, T., 287 Wicker, Elmus R., 351 Wicksell Knut, 54-55, 123, 368, 393-96, 605,615 Williams, John, 64-65 Williamson, Jeffery G., 89, 93 Working, Holbrook, 59~91 Yeager, Leland B., 657 Zecher, J. Richard, 4-5, 9~91, 97-98, 116, 387, 524, 529, 553 Zellner, Arnold, 480

Subject Index

Arbitrage: commodity, 33-34, 36, 45, 51, 90, 122-23, 129, 130-33, 146-47, 159, 450,460,529,541; interest-rate, 91-94; rationality of, 127-28; Swedish experience, 387-91. See also Interest-rate parity; Law of one price; Purchasingpower parity Argentina, 64-65, 85-86, 448 Atlantic economy, linkages, 94, 114-16, 167-68, 513-14, 516, 520, 545-46 Balance of payments -adjustment mechanisms, 36-39, 41-42, 49-50, 52-53, 62-63, 83-91, 230-31; asset market approach, 355, 358-59; monetary approach, 90-91, 122; pricespecie-flow approach, 10-11,24,33,54, 86-87, 174,243; theories of, 576-77, 580-81 -in Canada, 60-61, 550-51, 577 --center vs. periphery, 83-84, 116, 168 -shocks: real vs. nominal, 42-43, 44647; temporary vs. permanent, 42, 446-47 -under fiat money, 63-65 -under gold standard, 37, 41-42 Bank Charter Act of 1844,45-46,68-69, 179, 222, 238-41, 265-66, 271; rules vs. discretion, 26, 267; suspension of, 25359,265,269,470 Bank of England: Bank rate, 95, 177-78, 186-91, 200, 205, 209-10, 215-16, 22930; bankers' balances, 204, counter-

678

cyclical policy, 176, 191, 197, 201-2, 232,269,343; crises of 1847, 8,234-37, 245-52, 272-73; gold devices, use of, 177-78,223; internal vs. external goals, 77-82, 178-80, 222, 231, 244, 250-59; notes, 238, 246-47, 251, 255-57, 26468, 271; on gold standard, 6-8, 26-27, 45-47, 73-82, 94-98, 168, 173-80, 18698,205-16,222-23, 494; policy tools, 177-78; profit-making, 173, 176, 210, 224-25, 266, 272; proportion, 98, 168, 179, 204, 208; reserve-deposit ratio, 206-8, 212-14, 244; responsibility of, 45-47, 95, 173; variable spectral patterns, 217-21 Bretton Woods system, 15-16, 196, 447, 459 Business cycles, transmission of --cross-country gold flows, 458-60 --direct absorption effects, 458-60 --effect of real shocks, 460 -financial panics, 40, 479, 497-500 -money shocks and, 458-60, 479 -price and interest-rate arbitrage, 458-60 -U.K.-U.S. interdependence, 460-93, 507-9; antebelleum, 464-71; in greenback period, 471-72; under gold standard, 473-76, 493; interwar, 476-78 Canada -model of financial system, 303-6 -pre-World War I gold standard, 60-61; balance of payments under, 554-66;

679

Subject Index

countercyclical reserve ratios under, 566-71; countercyclical monetary base, 554-59, 577; price-specie flows under, 565-65; procyclical money stock, 55154, 559-61, 577; 1914 suspension of standard, 278-79 -1920 return to gold, 82, 291-93; role of monetary base reduction, 290-91, 306 -1928 suspension of standard, 293-300; role of gold drains, 296-96, 307 Capital flows -long-term, 75; barriers to, 353-54; interest rates, 44-45, 82, 91; and international borrowing, 55-59, 92; long swings in, 93-94, 117-18; relation to gold standard, 393, 403; to Sweden, 379-80, 390, 393, 401; transfer mecha- . nism, 82, 84-85, 114-16,227-28 -short-term, 25, 44-45, 60-63, 91, 174, 243-44, 54, 577 Central banks, role of under gold standard, 4, 26-27, 82, 277-78, 354-55, 457-58; Reichsbank, 315-26; Riksbank, 382-85. See also Bank of England Country risk, 355-56, 427-28, 431-32, 444 Discount rate, 65; in Germany, 312, 331-34 Defense expenditures, 596-97, 610-11 Econometric estimation problems, 18086, 197, 228-29, 341-42, 479-93, 51011, 545-46, 608-9 Exchange rates, 32, 34, 89, 174; BerlinLondon, 341-42; Canadian dollarpound, 282-83; case of flexible, 79; lira-French franc, 423-28, 431; lirapound, 426-27; nominal vs. real, 427, 444-45, 450; spot vs. forward, 448 France: gold standard experience, 61-63; interest rates, 208-17, 225,427-28 Genberg-Zecher criterion, 136, 140 Germany: pre-World War I stable growth, 326-30, 352; real, not monetary, cycles, 332; stable reserve-deposit ratios, 340-41 Gibson's paradox, 587-88, 604-9; Fisher explanation of, 589-92; genuine, 604-9; spurious, accounting for war, 593-602. See also Interest rates; Price level

Gold -commodity money: theory of, 11,4041, 43-44, 51-52, 54, 87; natural distribution of, 41, 75 -demand for, 394-95 -flows: internal vs. external, 26, 32, 40, 46-47, 174, 180, 254-55, 422; vs. income change, 84; from Canada, 29599; stable in Germany, 312 -mint pars, 16, 18 -points, 122, 125, 178,278,423 -real price: determination of, 350, 612, 615-20; Hotelling's rule, 356, 619-20 Gold standard: as British, 75, 97, 169, 173,458; classical views of, 31-47; country policies under, 16, 18, 71; cycle transmission under, 452-58, 460-79; dating of, 514; external vs. internal stability under, 31,77-82,259-60; gold production, 11-13, 613-15; Harvard school views, 55-67; international adjustment mechanisms, 24-27, 52-54, 63-67, 524-32; interwar critics, 67-82; managed, 94-96, 384; necessary conditions for, 17-19,78,345-46,662-66; popular perceptions of, 652-59; postWWII interpreters, 82-98; price level under, 11-13, 30, 76, 174, 260, 394-95; self-destruction if successful, 447-48; stylized facts of, 68-73, 80-84, 88. See also Price-specie-flow mechanism; Purchasing-power parity; Rules of the game Gold supply: adequacy of, 18; growth rate of world monetary gold, 621; history of, 12-13, 620-23; new discoveries, causes of, 395, 623-27; mine productivity, 631-32; politics, effects of, 632-39; technological changes in, 627-32 Granger causality tests, 10; cross-country real output, 519-23, 533-34, 539-40; Italian output, prices, exchange rates, 429-31; U.K.-U.S. real and monetary variables, 479-93, 509-11; before and after 1879,517-20 Gresham's law, Italy's bimetallic standard, 409-10, 415 Interest rates: in Canada, 567-71, 579-80, 582-83; effects of war on, 593-98; as function of price levels, 587-90; market vs. natural, 393-94; parity, 2-3, 91-94;

680

Subject Index

Interest rates (cont.) cross-country spreads, 339-43; time series properties of, 593--94 International adjustment mechanisms. See Balance of payments Key currencies, 94; held by Riksbank, 384-86,447 Latin Monetary Union, 409-10; Italy, effects on, 432 Law of one price, 25, 32-34, 39-40, 118; and arbitrage, 33, 388, 393; traded vs. nontraded goods, 33, 114, 129, 163 Lender-of-Iast-resort, 47, 224, 260, 267, 272,312; Canadian government as, 279 Monetary approach to the balance of payments, 5-6, 69-70, 354, 358, 661-62; base composition in, 358, 387; equilibrium model, 126-69; expenditure changes in, 406-7, 451; fixed vs. flexible exchange rates, 15; interest-rate parity results, 405-8, 423--29; Italian case, 42{}-22; purchasing-power-parity results, 122, 405-8, 423--29; role of national money, 126-27, 407; Swedish case, 385-91; tests of, 407. See also Price-specie-flow mechanism; Purchasing-power parity Monetary reform proposals, 37-38, 4849, 53-55, 81 Money supply: and balance-of-payments adjustment, 84, 86, 354; in Canada, 282-84; convertibility into gold, 23, 76, 173; determinants of, in Sweden, 36567,376-85; determinants of, in U.K., 497; determinants of, in Canada, 55966; under flexible exchange rates, 86; and gold reserves, 16-17, 62, 95-96, 168, 392 Peel's Act. See Bank Charter Act of 1844 Price level: effects of war on, 598-600; endogenous or exogenous, 127-28; 514, 524, 529-32; forecasts of, 611-12; relation to gold production costs, 614-20; secular changes in, Cassel vs. Wicksell, 393-96; spurious relation to interest rates under gold standard, 587-90, 611; time series properties of, 59{}-92; in wartime, 599-601

Price-specie-flow mechanism, 24, 33, 54, 86,87, 126-27, 174,243; according to Angell, 66-67; Canadian evidence for and against, 547-54; evidence for and against, 54-60, 63, 81-82; GrangerSims causality tests of, 406, 429-30, 445; irrelevant, 125,231; Italian case, 405-8; relative prices and, 406, 451; role of money in, 321,407,576-77; Swedish case, 387-89; U.S. case in 1879, 126, 159-60 Purchasing-power parity, 4-5, 127-30, 151-52: alleged failure, 133--36, 139, 159; commodity arbitrage, 128-30; consequences of, 659-61; economic interdependence, 121-23; efficient market hypothesis, 128-29; and exchange rate, 163--64, 449-50; Italian case, 423-29, 431-32,444-46; monetary policy, 14346, 162-63; tests of, 137-43, 152-56, 164-65, 449; Swedish case, 387-91, 400, 429-31,448; U.S. inflation case, 14{}43, 16{}-62 Real output: cross-country Granger-Sims tests, 519-23; data problems, 497, 5079,542; demand vs. supply shocks, 52223, 539-40; finite price elasticity of demand for, 529-32; inflation tradeoff, 7 countries, 523-29, 541; models of, 51416; money shocks and, 479; weak international linkages of, 516-18 Reichsbank: bills of exchange as base money, 33{}-32, 336; compared to Bank of England, 312-15; countercyclical money liabilities, 319-26; interest rate dampening by, 333; motives of, 35{}-52; procyclical bill portfolio, 315-19; profitmaking role, 343--44, 358; role of, 35253, 355; targets of, 33{}-36, 357 Riksbank: foreign exchange operations, 382....85; goals of, 38{}-81; gold reserves, 372-76; growth of note issues, 370-72, 377-79, 392; influence of U.K. on, 38{}81; key currencies held, 384-86 Rules of the game --defined, 174-76, 195-96 -and discretion, 96-97 -evidence for and against, 81-82, 88, 191-92, 216-17, 223 -how observed: by Canada, 553-54, by Italy, 419-22, by Reichsbank, 311-12,

681

Subject Index

315-26, 333; by Riksbank, 382-84 -importance of, 544 -inconsequential, 98 Sweden: current account deficits under gold standard, 379; determinants of money supply, 365-67, 376--85;eco-

nomic development, 361-63, 401-2; role of gold standard, 399-403, 450; spread of banking: 363-65 Terms of. trade, 32, 43, 82, 164, 354; and real transfers, 114; vs. income expenditure, 118