The Monetary Transmission Process Recent Developments and Lessons for Europe
Edited by The Deutsche Bundesbank
The Mo...
25 downloads
1619 Views
4MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
The Monetary Transmission Process Recent Developments and Lessons for Europe
Edited by The Deutsche Bundesbank
The Monetary Transmission Process
This page intentionally left blank
The Monetary Transmission Process Recent Developments and Lessons for Europe Edited by
The Deutsche Bundesbank
Editorial matter and selection © Deutsche Bundesbank 2001 Chapters 1–8 (and Comments/Discussions) © Palgrave Publishers Ltd 2001 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1P 0LP. Any person who does any unauthorised act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The authors have asserted their rights to be identified as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2001 by PALGRAVE Houndmills, Basingstoke, Hampshire RG21 6XS and 175 Fifth Avenue, New York, N. Y. 10010 Companies and representatives throughout the world PALGRAVE is the new global academic imprint of St. Martin’s Press LLC Scholarly and Reference Division and Palgrave Publishers Ltd (formerly Macmillan Press Ltd). ISBN 0–333–77244–X This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data The monetary transmission process : recent developments and lessons for Europe / edited by The Deutsche Bundesbank. p. cm.
Includes bibliographical references and index.
ISBN 0–333–77244–X (cloth)
1. Monetary policy—Europe. 2. Finance—Europe. I. Deutsche
Bundesbank.
HG925 .M6646 2000
332.4'94—dc21
00–042080
10 10
9 09
8 08
7 07
6 06
5 05
4 04
3 03
Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire
2 02
1 01
Contents
Notes on the Contributors
vii
Introduction Heinz Herrmann
1
1
2
Analysis of the Monetary Transmission Mechanism: Methodological Issues Bennett T. McCallum
11
Discussion Lawrence J. Christiano
44
Discussion Harald Uhlig
51
Price Stability as a Target for Monetary Policy: De®ning and Maintaining Price Stability Lars E. O. Svensson
60
Discussion Mervyn King
103
Discussion ~ Jose Vinals
107
3
The Transmission Process Allan H. Meltzer
112
4
Asymmetric Interest Rate Policy in Europe: Causes and Consequences Axel A. Weber
131
Discussion Carlo A. Favero
159
Discussion Philippe Moutot
164
Legal Structure, Financial Structure and the Monetary Policy Transmission Mechanism Stephen G. Cecchetti
170
Discussion Manfred J. M. Neumann
195
5
v
vi Contents
6
7
8
Discussion Ignazio Angeloni
201
Differences Between Financial Systems in European Countries: Consequences for EMU Reinhard H. Schmidt
208
Discussion Charles A. E. Goodhart
241
Comment Alain Vienney
247
European Labour Markets and the Euro: How Much Flexibility Do We Really Need? Michael C. Burda
252
Comment Hendrik J. Brouwer
276
Comment LuõÂs Campos e Cunha
279
The Monetary Transmission Process: Concluding Remarks Otmar Issing Francesco Giavazzi Claes Berg Ignazio Visco Manfred J. M. Neumann
283 294 298 303 311
Name Index
315
Subject Index
319
Notes on the Contributors
lgnazio Angeloni, European Central Bank Claes Berg, Sveriges Riksbank Hendrik J. Brouwer, De Nederlandsche Bank NV Michael C. Burda, Humboldt University, Berlin LuõÂs Campos e Cunha, Banco de Portugal Stephen G. Cecchetti, Federal Reserve Bank of New York Lawrence J. Christiano, Northwestern University Carlo A. Favero, Bocconi University Francesco Giavazzi, Bocconi University Charles A. E. Goodhart, London School of Economics and Political Science Heinz Herrmann, Deutsche Bundesbank Otmar Issing, European Central Bank Mervyn King, Bank of England Bennett T. McCalIum, Carnegie Mellon University Allan H. Meltzer, Carnegie Mellon University Philippe Moutot, European Central Bank Manfred J. M. Neumann, University of Bonn Reinhard H. Schmidt, Johann Wolfgang Goethe University, Frankfurt Lars E. O. Svensson, Stockholm University Harald Uhlig, Tilburg University Alain Vienney, Banque de France ~ Jose Vinals, Banco de Espana lgnazio Visco, Organisation for Economic Cooperation and Development Axel A. Weber, Johann Wolfgang Goethe University, Frankfurt
vii
Introduction
Heinz Herrmann
The debate on the monetary transmission process has not only continued in the past few years but has also intensi®ed. There are a number of reasons for this. In academic circles there are still various different approaches that seek to explain how monetary policy in¯uences an economy. More and more areas have come to be included in the analysis which attracted less attention in the past. Thus, nowadays ± on the basis of an improved microeconomic theory ± more attention is being paid to the role of ®nancial intermediaries and ®nancial systems in general than was previously the case. In economic policy, central banks are confronted with the task of safeguarding a level of price stability which, in most cases, is much greater than that achieved in previous decades. How this target can best be achieved is a question which depends in part on our state of knowledge of the transmission process. In Europe, the debate has been given an additional stimulus by the transition to European monetary union. How will this changeover to a single currency in¯uence the effects of monetary policy in the eleven member countries? Will the single monetary policy have similar transmission effects in the different national economies or will marked differences continue to exist? These are questions to which we are seeking to ®nd answers at the start of monetary union. The Bundesbank therefore decided to invite economists from universities, central banks and other institutions to a conference dedicated to these issues on 26 and 27 March 1999. This volume contains the papers given and the comments on them, with the introductory statements to a panel discussion which concluded the conference (pp. 283±314). Diverse models and methods have been developed in recent years aimed at gaining a better understanding of the monetary transmission process. The impression has arisen that the conclusions were not independent of the methods of analysis used. For example, a study by the BIS (1995) found that, depending on the type of model used, very different answers were provided to the question of whether and how the transmission processes differ between the countries examined. 1
2
Heinz Herrmann
Against this background Bennett McCallum in Chapter 1 considers methodological aspects in `Analysis of the Monetary Transmission Mechanism: Methodological Issues', designed to help ®nd appropriate procedures and models which allow meaningful statements to be made about how monetary policy works through to the economy and about what constitutes a meaningful monetary policy. McCallum argues that we should concentrate on the systematic components of monetary policy and not on the impact of monetary shocks. It follows that it is important to base the analysis on structural models (rather than on vector-autoregression (VAR) models). To be able to assess models, it is important to examine to what extent they are able to re¯ect properties of the analysed economic variables as they act in reality. In this connection McCallum stresses the importance of vector autocorrelation functions as diagnostic tools, whereas he is critical of the much-used impulse response functions for the respective policy shocks. He presents his preferred approach using a sticky-price model. To demonstrate the effects of monetary policy, he applies different interest rate rules, which may be regarded as variants of a Taylor rule. Finally he discusses two other analytical approaches, the narrative approach of Romer and Romer (1989) and a VAR procedure of Sims (1998) in which the actual monetary policy stance in a historical period is replaced by an alternative policy rule. But he is rather sceptical about both approaches. In his commentary, Lawrence Christiano considers both the methodological proposals made by McCallum as well as the model used by McCallum by way of illustration. He ®rst defends the approach of closely analysing the impact of monetary shocks as an important method of selecting suitable models and also mentions various new methods of testing the models' consistency with reality. In this context, he also explains another method which tests whether a model is capable of reproducing key features of particularly interesting historical periods. He stresses the advantages of a limited participation model of money over a sticky-price model which become evident in the process. Harald Uhlig states that analysing both monetary shocks and the systematic component of monetary policy are important and that the two complement each other. He then asks some critical questions about the model presented by McCallum. For example, he is of the opinion that the exogeneity of investments in the model is unsatisfactory and that some differences to which McCallum drew attention regarding the advantages of certain monetary policy rules ultimately seem to have little relevance. Finally, he makes some proposals as to how the model can be used in the future for evaluating different monetary policy stances, given its microeconomically-based structure. The question of how monetary policy in¯uences real variables and prices leads directly to considerations about how central banks should shape their monetary policy. In Chapter 2, `Price Stability as a Target for Monetary Policy: De®ning and Maintaining Price Stability', Lars Svensson addresses several
Introduction 3
problems: How is the general term `price stability' to be de®ned? How should central banks formulate the rules which they follow? What are the lessons for the European Central Bank (ECB)? Regarding the ®rst point, Svensson discusses, inter alia, the advantages and disadvantages of in¯ation targets versus price-level targets, the question of an appropriate price index for formulating an in¯ation target and selecting the appropriate level for an in¯ation target. Concerning the formulation of monetary policy rules, Svensson draws attention to the advantages of forecast targeting over an instrument rule and an intermediate targeting strategy. He then extends his basic model to include non-linearities and stochastic coef®cients. He considers a forecast targeting strategy which takes account of the probability distribution of the forecast to be an adequate response to such model extensions. He advises the ECB to use a strategy of forecast targeting, to explicitly formulate the in¯ation target, to publicise the in¯ation forecasts and to refrain from assigning a special role to the money stock. In his commentary, Mervyn King echoes Svensson's preferences for forecast targeting, but warns against overemphasising the differences between the strategies as applied in practice. He states it is more important that the transmission mechanism is understood. He stresses the importance of looking at the entire gamut of in¯ation forecasts when making monetary policy decisions. Unlike Svensson, he considers it is justi®ed to assign a special role to monetary growth in the monetary policy strategy. Jose Vin Ä als likewise sees considerable advantages in a monetary policy strategy characterised as forecast targeting by Svensson. His view is that such a strategy is consistent with major principles of a good monetary policy practice, among which he includes transparency and an orientation to the medium term. He expresses the view that it is also much easier to communicate to the general public than an optimal instrument rule. Svensson has based his statements on monetary policy rules on a rather generally formulated system of equations for describing the transmission process in which a control variable, which is normally identical with a shortterm interest rate, stands in relation to predetermined variables (and possibly forward-looking variables). In Chapter 3, `The Transmission Process', Allan Meltzer looks more closely at this question, asking whether the importance of money market rates or central bank rates is really that pivotal and exclusive in the transmission process as is assumed in such models. Using examples from US history, he seeks to show that focusing on a short-term interest rate is not justi®ed. Rather, there had been periods in which monetary policy showed effects which were not consistent with this usual conception of the monetary transmission process. In his analysis he concentrates on historical phases of de¯ation in which the development of the (real) monetary base explained economic developments better than (real) short-term interest rates. Meltzer mentions that two of these periods are suitable for showing that no liquidity trap will emerge even when interest rates are low, a statement which takes on
4
Heinz Herrmann
added importance in the current phase of low interest rates. To support his thesis of an effective wealth effect, he presents consumption functions in which the real money stock has a clear explanatory value. He therefore advocates keeping an eye on monetary growth when making monetary policy decisions. Empirical studies of the monetary transmission process which, on the one hand, analyse the effects of the central banks' interest rate policy on the economy and, on the other, ask which behaviour pattern is to be recommended for monetary policy, are confronted with a severe methodological problem: it is not easy to decide whether the link between central bank rates and/or money market rates and other economic variables such as in¯ation is the result of a central bank's reaction function or whether it re¯ects the implications of interest rate policy on the corresponding economic variables. This prompts Axel Weber in Chapter 4, `Asymmetric Interest Rate Policy in Europe: Causes and Consequences', to examine the link between short-term interest rates and also in¯ation rates in the European countries (as well as in the United States and Japan) and also the link between national interest rates in Europe during the period of the European Monetary System (EMS) and to demonstrate possible ways of better identifying the nature of the links. Drawing on past papers (such as King and Watson, 1997), he uses bivariate VAR models which take into account either interest rates and in¯ation, or the German interest rate and that of another European country. One of his ®ndings is that, in contrast to the United States, it is rather doubtful whether German monetary policy can be accurately described by an interest rate-smoothing rule if one assumes a long-run Fisher effect and long and variable lags in monetary policy. Regarding the interest rate pattern in Europe, he suggests that many central banks in Europe followed German interest rate policy, but that there existed no simple long-run homogeneous pattern of co-movement. In his commentary, Carlo Favero agrees with the main thrust of Weber's argument that considering monetary policy rules in isolation is a major methodological problem. One should study such rules in the framework of macromodels instead. However, he argues that bivariate models speci®ed in ®rst differences, as proposed by Weber, are only a ®rst step in the right direction. It would be desirable to extend such a framework to achieve a closer relation between theoretical and applied models. Philippe Moutot shares the view that Weber raises an important question which should trigger a useful research programme. Furthermore, he makes some more general remarks with respect to monetary policy rules. He suggests that this chapter makes it clear that it is dif®cult to identify the rules followed by the central banks under consideration. Against this background, he warns against the hope that simple rules can replace discretion. In his view the relevant issue is how to combine the need for some limited discretion with the need for transparency vis-a-vis the public, and describes how the ECB has proceeded in this respect.
Introduction 5
Differences in the monetary transmission process in European countries may be caused by a variety of reasons. With increasing attention being paid to the role assumed by information problems in the transmission mechanism ± as is the case with the credit view, for example ± the focus has shifted to the different organisation of ®nancial systems in the member countries. Both Stephen Cecchetti and Reinhard Schmidt examine what in¯uence differences in the ®nancial systems of the European countries may have on the transmission of monetary policy. Stephen Cecchetti in Chapter 5, `Legal Structure, Financial Structure and the Monetary Policy Transmission Mechanism', examines the links between the structure of ®nancial systems, the conditions for ownership rights in a given country and the effectiveness of monetary policy. He bases his considerations on the lending view, as presented, for example, by Kashyap and Stein (1994). This implies that the central bank's interest rate policy should have a stronger impact in countries in which enterprises mainly rely on bank borrowing and in which the banking system is characterised by a multitude of small and more fragile banks than in countries with robust capital market ®nancing and with dominant big banks. Against this background, Cecchetti ®rst identi®es marked differences in the ®nancial systems of the EMU member countries. He then refers to comparative studies on the strength of the transmission process in those countries which, in accordance with the lending view theory, likewise show marked differences. In a next step Cecchetti looks for the reasons that explain the different ®nancial structures. In this process he focuses on the role of the respective legal systems, which result in different rights of creditors and shareholders in the individual countries. He comes to the conclusion that it is these legal differences which have led to the respective ®nancial structures and are thus one reason for the differences in transmission mechanisms. He concludes that unless legal structures, are harmonised across Europe, the ®nancial structures, and hence transmission mechanisms, will remain diverse. In his commentary, Manfred Neumann ®rst draws attention to some inconsistencies in the construction of the indices which measure the importance of the lending channel in the individual countries. He states that the aggregation of different features to form such an index gives rise to dif®cult and unresolved problems. Furthermore, he calls into question the robustness of Cecchetti's empirical ®ndings on the links between the importance of the credit channel in a given country and the impact of monetary policy on its output. Ignazio Angeloni mentions, ®rst, that in most countries in the monetary union the relevant legal systems are not so different. Secondly, he expresses the view that the traditional transmission channels via the exchange rate and the interest rates, which Cecchetti has neglected, are becoming more and more similar in the wake of the changeover to monetary union. He also puts forward the argument that the effects of a lending channel in the European countries will converge as a result of forthcoming changes in the
6
Heinz Herrmann
respective ®nancial systems. Finally, he criticises some of Cecchetti's empirical ®ndings as contradicting the literature and also as being at variance with his theoretical arguments. Reinhard Schmidt, in Chapter 6, `Differences Between Financial Systems in European Countries: Consequences for EMU' also comments on the effects which typical differences in the ®nancial systems have on the transmission mechanism and likewise focuses on the credit channel, in particular. He describes some important features of a ®nancial system ± i.e. the ®nancial sector and the ®nancial side of the real sector. He classi®es four sub-systems, each of which may assume different forms and which, together, make up a ®nancial system: the ®nancial sector (bank-dominated versus ®nancial market-dominated), the ®nancial patterns (bank lending versus ®nancing via the capital market, bank deposits versus asset acquisition through securities), corporate governance (control by insiders or outsiders) and the corporate strategy (major versus minor importance of implicit contracts in business decisions). He regards the German and the British systems as being different but internally consistent, whereas he describes the French system as being in a state of ¯ux in which it is dif®cult to ®nd a unifying `logique'. For the purpose of assessing the monetary transmission mechanism, he differentiates between the two levels ± between the central bank and the ®nancial sector, on the one hand, and between the ®nancial sector and the real sector, on the other. He argues that the central bank in the German system can exert a greater in¯uence on the ®nancial sector than in the Anglo±Saxon system but that in the former the ®nancial sector is much better able to cushion the economy from monetary shocks. Although the differences at the respective levels are considerable, they tend to offset each other in the aggregate. Finally Schmidt comments on the stability of the ®nancial systems in the initial phase of monetary union. The introduction of the Euro itself is unlikely to change the systems fundamentally, but the increased interpenetration of elements between one ®nancial system and another may have far-reaching implications for the monetary transmission mechanism. In his commentary, Charles Goodhart agrees with Schmidt's characterisation of the different ®nancial systems. But he is more critical regarding the description of the transmission channels. He argues that the strength of the credit channel depends very much on the given circumstances ± for example, whether the banking system is well endowed with own capital. Regarding the interest rate channel he believes that the monetary policy effects are particularly strong in the United Kingdom. In this connection he raises the question why UK interest rates have shown greater volatility than German rates in the past. Finally, Goodhart notes that he is less pessimistic than Schmidt regarding the coming restructuring of the ®nancial systems and the associated problems. This view is shared by Alain Vienney. Vienney also criticises the overemphasis on the credit channel compared with the interest rate channel. In his view, neglecting the interest rate channel is the reason
Introduction 7
why Schmidt overstates the differences in the transmission of monetary policy between the Continental Europen countries, whereas he understates the the Anglo±Saxon countries. In this context, Vienney also differences vis-a-vis draws attention to the differences in the maturities of ®nancial relationships and in the balance sheet structures which still exist in the United Kingdom compared with continental Europe. A major factor in¯uencing the transmission of monetary policy, in addition to the ®nancial systems and the situtation on goods markets, are labour markets. A key factor here is the ¯exibility of prices and wages. Two concerns have emerged with a view to EMU: on the one hand, the fear that in a single currency area there will be a convergence of wages which is not accompanied by a corresponding convergence of labour productivity. The second concern in this respect is that, with a single currency and a single monetary policy, there will be no mechanisms to deal with asymmetric shocks. Hence, the question arises of whether there are indications that the labour markets will show suf®cient ¯exibility to assume the necessary burden of adjustment. In Chapter 7, `European Labour Markets and the Euro: How Much Flexibility Do We Really Need?', Michael Burda attempts to provide answers to that question. Burda ®rst discusses the general question of what role rigidities on the labour and goods markets may play in the convergence of countries in a monetary union and summarises his ®ndings by saying that, at the start of monetary union, rigidities and the lack of mobility on the labour markets must be considered problematical. He then outlines scenarios which show the future in a much more favourable light. He takes the view that monetary union will act as the engine for the freeing-up of labour markets and real wages in Europe. A factor contributing to this, he suggests, is that prices will tend to adjust more slowly to a new equilibrium since, inter alia, he considers the adjustment pressures in a large, more closed currency area to be lower in this respect. On the other hand, he envisages that the negotiating power of the trade unions will decline in the future Europe and that the pressure on governments to carry out institutional reforms ± for example, of the social security systems ± will increase, which in turn will boost the ¯exibility of wages. In view of this new combination of circumstances, Burda predicts an improved transmission of monetary policy, but he also warns against abuse. In his commentary, Hendrik Brouwer stresses how important it is for labour markets to become more ¯exible in view of the high level of unemployment and potential asymmetric shocks in monetary union. Although he supports some of Burda's arguments ± for example, that goods prices might become less ¯exible ± he disagrees with him in some important points. He mentions that there are indications of persisting labour market rigidities and he is less optimistic about the governments' commitment to reforms. Luis Campos e Cunha, too, is sceptical regarding Burda's central statement that the euro will be the `Trojan horse' which liberalised labour markets. In addition, he is not convinced by Burda's opinion that EMU will lead to less ¯exible prices. Finally,
8
Heinz Herrmann
he warns of the risk of inappropriate harmonisations of labour markets ± for example, of minimum wages. The conference volume ends with statements during a debate which was intended to re¯ect, in particular, the view of practitioners on the topics discussed and the problems ahead. In his introductory remarks Otmar Issing addresses two questions, in particular: `What do we know about the transmission process in Europe at the start of monetary union?' and `How will this process change in the coming years?' He reminds us that the participating countries share a number of common features, but also have some differences. But in his view, it would be dif®cult to make any clear-cut pronouncements, based on institutional differences or econometric studies, on the divergent effects of monetary policy in those countries. Looking to the future, he is of the opinion that a number of factors, not least the single monetary policy and the single currency, should result in a convergence of the transmission process. However, he draws attention to our imperfect knowledge, a factor which the ECB's monetary strategy takes into account. Francesco Giavazzi focuses in particular on the implications of the restructuring of the European banking markets. He refers to studies according to which the credit channel works differently in the individual countries. It seems likely that cross-border banking mergers would reduce such differences; actually, however, most mergers in Europe take place within national boundaries. Giavazzi states that this is worrying, for competition reasons, on the grounds of general considerations concerning an ef®cient ®nancial system in Europe and also in the context of a single transmission process. Claes Berg ®rst presents his view on the previous debate and questions ± in particular, Burda's optimistic conclusions regarding the ¯exibility of real wages. In his more general remarks he comments on the monetary policy strategy, emphasising the importance of published in¯ation forecasts. Futhermore, he considers whether a central bank should pursue an aggressive interest rate policy or whether a step-by-step policy would be preferable. He mentions that uncertainty does not always justify a smooth interest rate policy. Based on the economic situation in spring 1999, Ignazio Visco discusses some future challenges for economic policy and, in particular, for monetary policy. He stresses the progress achieved in convergence and in economic integration in Europe which, in principle, should facilitate a European monetary policy. At the same time he points to the uncertainty as to the appropriate monetary policy in the new currency area, an uncertainty engendered in part against the backgound of the historically low in¯ation rates. In this connection he also emphasises the necessity of a discussion on the appropriate policy mix in Europe. Manfred Neumann discusses two questions: `What is the optimal ®nancial system?' and `Are differences in the ®nancial systems of importance at all in the long run?' He comes to no de®nite conclusion regarding the advantages of the respective ®nancial systems. He stresses, however, that a system which is based more strongly on ®nancial markets (rather than on
Introduction 9
banks) is likely to have some advantages for the funding of innovative, young enterprises. At the same time, he states that a comparison of the growth performance of the United States and that of Germany does not indicate the superiority of the US ®nancial system. Also, he takes the view that differences between countries are a problem, especially whenever unexpected shocks emanate from monetary policy. From this he concludes that monetary policy should have a medium-term orientation as far as possible in order to avoid such shocks. References BIS (1995) `Financial Structure and the Monetary Policy Transmission Mechanism', Basle. Kashyap, A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending,' in N. G. Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press. King, R. G. and M. W. Watson (1997) `Testing Long Run Neutrality', Federal Reserve Bank of Richmond, Economic Quarterly, 83, pp. 69±101. Romer, C. and D. Romer (1989) `Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Schwartz', NBER Macroeconomics Annual 1989, Cambridge, MA, MIT Press. Sims, C. A. (1998) `The Role of Interest Rate Policy in the Generation and Propagation of Business Cycles: What has Changed Since the '30s?', in Federal Reserve Bank of Boston (ed.), Beyond Shocks: What Causes Business Cycles?, Boston: Federal Reserve Bank of Boston.
This page intentionally left blank
1
Analysis of the Monetary Transmission Mechanism: Methodological Issues Bennett T. McCallum1
1 Introduction The purpose of this chapter is to consider several methodological issues relevant for study of the monetary transmission process. These issues involve relative emphasis on monetary shocks as opposed to systematic policy adjustments; vector autoregression versus structural modelling research strategies; impulse response versus vector autocorrelation functions as diagnostic tools; and an evaluation of the so-called `narrative approach'. But while these methodological issues are stressed, the chapter's approach is signi®cantly substantive, in the sense that the issues will be considered in the context of a non-trivial quantitative analysis that is intended to be of interest on its own. As a preliminary matter, it may be useful to outline what meaning is here being given to the term `monetary transmission mechanism'. That this term evokes different responses from different scholars is well illustrated by a symposium on `The Monetary Transmission Mechanism' featured in the Fall 1995 issue of the Journal of Economic Perspectives. In the papers of that symposium, Bernanke and Gertler (1995) focus on the credit channel; Meltzer (1995) promotes monetarist emphasis on the importance of recognising multiple assets;2 Taylor (1995) outlines a particular econometric framework for studying the transmission mechanism; Obstfeld and Rogoff (1995) discuss foreign exchange rate policy and ®nancial crises; and Mishkin (1995) provides a brief overview. More generally, many writers on the subject restrict their attention to the effects of monetary policy shocks,3 while some are concerned only with effects on real variables. In the present chapter, however, the concept of the transmission process to be considered includes effects on both real and nominal variables of shocks and more especially the regular, systematic component of monetary policy. The implied de®nition, therefore, is similar to that expressed by Taylor (1995, p. 11): `the process through which monetary policy decisions are transmitted into changes in real GDP and in¯ation.' 11
12 Bennett T. McCallum
An outline of the chapter is as follows. In Section 2, it is argued that study of the systematic component of monetary policy actions is at least as important as the study of the unsystematic component, also known as policy shocks. Section 3 then presents some procedures for exhibiting effects on in¯ation, output and other variables of different systematic policy responses. These differences are, of course, model-speci®c: they depend upon the structural speci®cation of the model being utilised. In Section 4, consequently, some variants of the basic model utilised in Section 3 are considered. It is demonstrated that systematic policy effects are signi®cantly dependent upon speci®cations relating to price adjustment behaviour, habit formation in saving versus consumption decisions and the economy's openness to foreign trade. This dependence is expressed in terms of root-mean-square statistics for in¯ation targeting errors and output gap measures, and also in terms of the characteristics of impulse response functions for shocks other than the monetary policy shock. Section 5 concerns diagnostic tools to be used in the construction of structural models; here it is suggested that more attention should be given to vector autocorrelation functions (and correspondingly less to impulse response functions) than is typically the case in the vectorautoregression (VAR) literature. Finally, Section 6 offers a partial evaluation and criticism of the `narrative approach' introduced by Romer and Romer (1989) and a non-standard VAR procedure utilised by Sims (1998b), plus brief comments on relevant papers by Bernanke, Gertler and Watson (1997) and Dotsey (1999). Section 7 is a brief conclusion.
2
Shocks versus systematic policy
A large volume of literature exists, much of it highly sophisticated, in which the effects of monetary policy on output, prices, and other variables are discussed entirely in terms of policy shocks.4 In this context, policy shocks represent the random, unsystematic component of the monetary authorities' actions ± i.e. the portion that is not related to the state of the economy, current or past. A leading theme of the present chapter is that emphasis on the shock component has been overdone; that while both shocks and the systematic component of behaviour are important, it would be more fruitful to emphasise the latter. This point of view has been taken by a number of analysts, including Taylor (1995), Rotemberg and Woodford (1997, 1999) and Bernanke, Gertler, and Watson (1997), but needs to be stressed nevertheless because of the sheer volume of literature that differs in this crucial respect. Perhaps the simplest way of arguing for an emphasis on the systematic component of policy is to recognise that quantitatively the unsystematic portion of policy instrument variability is quite small in relation to the variability of the systematic component. An illustration is provided by the prominent study by Clarida, Gali and Gertler (1998) of policy behaviour since 1979 by central banks of the G-3 nations. In particular, their `baseline'
Methodological Issues 13
estimations of monthly Bundesbank, Bank of Japan and Federal Reserve reaction functions indicate that the fraction of monthly interest instrument variability that is unexplained by systematic determinants is only 1.9, 3.0 and 1.6 per cent, respectively.5 Also, Rotemberg and Woodford (1997) and McCallum and Nelson (1998) ®nd in the US quarterly data that only about 5 per cent of instrument variability is unexplained over roughly the same period. Indeed, it is conceivable that policy behaviour could be virtually devoid of any unsystematic component. In the limit, that is, the variance of the shock component could approach zero. But this would not imply that monetary policy is unimportant for price-level behaviour, a central bank's main responsibility. Nor would it imply that policy is unimportant for real cyclical activity unless the economy is of the rather special type that satis®es the `policy ineffectiveness' proposition.6 More generally, it should be kept in mind that when a central bank raises its interest rate instrument by (for example) 50 basis points `in order to head off in¯ation', the action is likely to represent a systematic response, not a shock. To illustrate the implications of some less extreme and less obvious phenomena, let us consider a simpli®ed analytical representation of monetary policy behaviour and its consequences for output and in¯ation. Here (and in the remainder of the chapter) let Rt , yt and pt denote a short-term nominal interest rate, the logarithm of real output, and the log of the price level. Also let yt be the natural rate value of yt , with pt the associated price level,7 and let vt and et represent shocks to spending and monetary policy behaviour. Thus et is the unsystematic component of policy. To keep the present example simple, we temporarily pretend that yt is a constant and normalise it as y 0, so that yt also measures output relative to its natural rate value. Our schematic model is given by three equations, as follows: yt b0 b1
Rt � Et pt1 Et yt1 vt pt � pt�1
1 �
p t�1 � pt�1 Et�1
p t � p t�1
Rt Et�1 pt1 0 1
Et�1 pt1 � et
b1 < 0
1:1
0 < < 1
1:2
1 > 0
1:3
The ®rst of these is an IS-type relation representing demand for current output ± i.e. saving versus spending behaviour. Optimising theory suggests that the variable Et yt1 should appear as indicated on the right-hand side; this has been argued by Kerr and King (1996), McCallum and Nelson (1999), and Rotemberg and Woodford (1997), among others. Neglect of this term would not simplify derivations substantially and would not affect the main points of the analysis.8 This future expected income term will accordingly be incorporated in the quantitative models of Sections 3±4. Equation (1.2) is one form of the P-bar price adjustment relation that was rationalised and utilised in papers by McCallum and Nelson (1997, 1998). It is not as widely used as the Calvo±Rotemberg (1983) model9 or variants of the
14 Bennett T. McCallum
Fuhrer±Moore (1995) speci®cation, but its theoretical properties are arguably superior.10 In any event, most of the points to be made here would carry over to other price adjustment speci®cations. And in our quantitative work below, the Fuhrer±Moore speci®cation will be considered in addition to (1.2). Note that the variable pt in (1.2) is the value of pt that would induce producers to make yt y t . Equation (1.3) represents monetary policy behaviour according to which an interest rate instrument Rt is set each period so as to raise the expected real rate of interest, Rt ± Et�1 pt1 , when the expected future in¯ation rate exceeds the target value . With 0 chosen to equal the average real rate of interest, (1.3) amounts to a special case of a forward-looking Taylor rule. Expectations are dated t � 1 in (1.3) so as to realistically limit the information available to the central bank when setting Rt . The notation is that Et ztj E
ztj j t , where t includes variables dated t and earlier. Solution of the foregoing model is facilitated by the fact, demonstrated in McCallum and Nelson (1998), that (1.2) implies that Et�1 y~ t y~ t�1 , where y~ t yt ± yt . In the present setting with yt 0, yt y~ t so we have Et�1 yt yt�1 . Then the minimal-state-variable (MSV) rational expectations solution11 for yt and pt is of the form yt 10 11 yt�1 12 vt 13 et
1:4
pt 20 21 yt�1 22 vt 23 et
1:5
and it is clear that 10 0, 11 . Also, since prices are fully predetermined, 22 23 0. Given these facts, the undetermined coef®cient procedure can be used to ®nd the remaining values of the ij . The solution, it turns out, is yt yt�1 1 =
1 1
1 � vt b1 1 =
1 1
1 � et
pt �
0 =1 �
b0 =b1 1
1 � =b1 1 yt�1
1:6
1:7
Now let us suppose that the above system were studied by some VAR approach that correctly identi®ed the policy shock term et . The coef®cient b1 1 /(1 1 )(1 ± ) is negative, so a VAR estimate of (1.6) would correctly ®nd a negative effect from a positive shock to Rt , provided that there had been enough sample-period variation in et . If et had stayed close to its mean value of zero, however, statistical procedures might ®nd no signi®cant effect in ®nite samples of a realistic size. In the case of pt , (1.7) indicates that the policy shock et plays no role when the effect of yt�1 is taken into account, as it would be in any VAR study. A study that looked for monetary effects on in¯ation by the Granger-causality method would therefore conclude that there are no such effects. The method relying upon the `fraction of explained variance' of the various shocks would attribute some portion to et since et�1 , et�2 , . . . help to explain yt�1 . But the fraction of pt variability attributed to monetary shocks would then be precisely the same as the fraction pertaining to yt variability. Note that since (1 ± )/b1 1 < 0,
Methodological Issues 15
these will be such that a surprise increase in Rt will have the effect of increasing ± rather than decreasing ± subsequent values of pt . This is a rather perverse property of this simpli®ed model.12 But it is nevertheless true that an increased value of the policy response parameter 1 will decrease the variability of in¯ation, pt; in the system (1.1)±(1.3). Furthermore, note that, since the unconditional mean Eyt equals zero by construction, relation (1.1) implies that the average value of the real rate of interest rt Rt ± Et pt1 equals ±b0 /b1 . Thus if the central bank chooses 0 to equal ± b0 /b1 , as a sensible policymaker would do, (1.7) reduces to pt
1 � =b1 1 yt�1
1:7
Thus on average, over a large number of periods, realised in¯ation will tend to equal , a target rate that is entirely determined by monetary policy.13 Within a given policy regime, the `long-run' ± i.e. unconditional mean ± value of in¯ation is entirely monetarily determined, but this fact would not be revealed by shock-oriented VAR procedures.14 Having argued that it is more important to focus on the systematic portion of monetary policy actions, rather than on shocks, we are then necessarily driven toward the study of structural models, rather than VARs.15 The reason is that even `identi®ed' or `semi-structural' VAR systems do not give rise to behavioural equations that can be presumed to be structural ± i.e. policyinvariant. The purpose of identi®ed VARs is to identify the unsystematic component of monetary policy, not to generate policy-invariant equation systems. But it is the latter that governs the effects of systematic or anticipated policy actions. This should be emphasised, for it is the crucial point of my argument. It is, I believe, consistent with the analysis and views of most creators and practitioners of the identi®ed VAR approach, including Bernanke, Blanchard and Watson, and Christiano, Eichenbaum, and Evans.16 My position would not be accepted by Sims (1998b), however, so some discussion of his approach is included below (in Section 6). It should be noted that my objections to several aspects of VAR analysis are not the same as those put forth by Rudebusch (1998). In fact, they are not actually objections to VAR analysis per se but rather are arguments for concentrating on the systematic component of policy rather than the shocks.
3 Effects of systematic policy In this section the purpose is to describe one approach to analysis of the effects of the systematic component of monetary policy. Since this undertaking requires use of a structural model, according to the foregoing argument, the results obtained will depend upon the adopted speci®cation of economic behaviour. My starting point will be the small-scale, open economy, quarterly model based on an optimising analysis that is developed and presented in
16 Bennett T. McCallum
McCallum and Nelson (1998). The following paragraphs will brie¯y outline that model and report some simulation results that serve to characterise the effects of monetary policy. Variants of the basic model will be considered in Section 4. Basing one's analysis on the assumption of explicit optimising behaviour by the modelled individuals in a general equilibrium setting is obviously not suf®cient ± and perhaps not necessary ± for the creation of a structural model that is speci®ed with reasonable accuracy relative to economic reality. The optimising general equilibrium approach can be helpful in this respect, however, since it eliminates potential internal logical inconsistencies that are possible when this source of intellectual discipline is absent. The model in McCallum and Nelson (1998), henceforth termed the M±N model, has a very simple basic structure since it depicts an economy in which all individuals are in®nite-lived and alike. As with many recent models designed for policy analysis, it assumes that goods prices are `sticky,' i.e. adjust only slowly in response to changes in conditions. It differs from many previous efforts in this genre, however, in three ways. First, the gradual price-adjustment speci®cation satis®es the strict version of the natural-rate hypothesis.17 Second, the modelled economy is open to international trade of goods and securities. And, third, individuals' utility functions do not feature time-separability, but instead depart in a manner that re¯ects `habit formation'. This last feature is speci®ed as follows. A typical agent desires at t to maximize Et
Ut Ut1 . . ., where the within-period measure Ut is speci®ed as Ut exp
vt
=
� 1Ct =Ct�1 h
�1=
1 � �1 Mt =pt 1�
1:8
Here Ct is a CES consumption index, Mt /Pt is real domestic money balances, vt is a stochastic preference shock and h is a parameter satisfying 0 h < 1. With h 0, preferences feature intertemporal separability, but with h > 0 there exists `habit formation' that makes consumption demand less volatile. The open-economy aspect of the model is one in which produced goods may be consumed in the home economy or sold abroad. Imports are exclusively raw materials, used as inputs in a production process that combines these materials and labour according to a CES production function. Capital accumulation is not modelled endogenously, but securities are traded internationally. The relative price of imports in terms of domestic goods, i.e, the real exchange rate, affects the demand for exports and imports, the latter in an explicit maximising fashion. Nominal exchange rates and the home country one-period nominal interest rate are related by a version of uncovered interest parity, one that realistically includes a stochastic `risk premium' term (as in Taylor, 1993b, and many multi-country econometric models). Price adjustments conform to the P-bar model, mentioned above, but with capacity output y t now treated as a variable that depends upon raw material
Methodological Issues 17
inputs and the state of technology, the latter driven by an exogenous stochastic shock that enters production in a labour-augmenting fashion.18 As mentioned above, price adjustment behaviour implies Et�1 y~ t y~ t�1 , so application of the unconditional expectation operator yields Ey~ t Ey~ t and this implies Ey~ t 0 regardless of the monetary policy rule employed. This natural rate property is not a feature of the Calvo±Rotemberg or Fuhrer± Moore models of price adjustment. Indeed, there are very few sticky-price models that have the natural rate property, the only other one that I know of being Gray±Fischer-style nominal contracts that imply limited persistence of y~ t magnitudes. The foregoing is intended to give the reader a broad overview of the M±N model; for a full description the reader may consult McCallum and Nelson (1998).19 Here the objective is to combine that structure with a policy rule formulation that permits a moderately straightforward `measure' of the effect of systematic policy activism. We begin with the following rule: Rt
1 � 3 Et�1 pt1 0 1
Et�1 pt � 2 Et�1 y~ t 3 Rt�1 et
1:9
In (1.9), one difference relative to (1.3) is the inclusion of a lagged Rt term, to re¯ect a form of interest rate smoothing that seems to characterise the behaviour of actual central banks. In light of estimates and previous experience, our basic experiments will assign a value of 0.8 for 3 . A second difference is the appearance of 2 Et�1 y~ t , as in the Taylor rule (Taylor, 1993a), but initially we shall set 2 0. Then with 2 0 and 3 0.8, the extent to which activist but systematic policy actions are taken is directly related to the magnitude of 1 . In that regard, a third difference relative to (1.3) is that feedback is taken from Et�1 pt , rather than Et�1 pt1 . The reason is that the former is more effective in this model, as will be seen below. Since the implicit primary objective of rule (1.9) is to keep in¯ation pt close to the target value , one measure of the effect of policy is the reduction (if any) in the rootmean-square-error (RMSE) value of pt � . In the simulations reported in this section, all constant terms are set to zero ± a standard practice in stochastic simulation work of this type20 ± so the standard deviation of pt can be interpreted as the RMSE value of pt � . Somewhat less tenuously, the standard deviation of y~ t can be interpreted as the RMSE value of yt ± yt . In all cases, the reported magnitudes are mean values averaged over 100 replications, with each simulation pertaining to a sample period of 200 quarters (after 53 start-up periods are discarded). Calculation of the rational expectation solutions is effected by means of Klein's (1997) algorithm. The most basic results are given in Table 1.1. In the ®rst row of the ®rst panel we see how the RMSE value of p � decreases as additional policy response to expected target errors is applied. With 1 0.1, there is almost no response to such errors; with 1 0 there would merely be a gradual adjustment of
18 Bennett T. McCallum Table 1.1 Simulation results for variants of policy rule (1.9) Basic model. 3 0.8 Case
std dev. of
1 0.1
0.5
1.0
5.0
50.0
Et�1 pt as target 2 0
p y~ R
10.54 1.45 2.17
7.37 1.83 2.55
5.48 2.12 2.86
1.91 2.71 3.77
0.24 3.08 4.50
Et�1 pt1 as target 2 0
p y~ R
11.21 1.35 2.06
9.76 1.52 2.14
8.57 1.65 2.29
4.51 2.29 3.25
** ** **
1 0.5 on pt 2 on Et�1 y~ t
p y~ R
2 0.1 7.45 1.79 2.54
0.5 8.29 1.70 2.96
1.0 9.10 1.59 3.32
5.0 13.37 0.97 4.97
50.0 18.77 0.17 6.98
Notes: The 2 values actually used are 1 /4 of the values listed; the latter correspond to units of measurement in annualised percentage points, as are typically reported in the Taylor rule literature. That statement applies to all results in this chapter. means MSV solutions are not available with existing software.
Rt � Et�1 pt1 toward its long-run average value. As 1 is increased, with policy response strength increased, the standard deviation of pt falls distinctly. In the second panel, feedback response is taken from Et�1 pt1 � . Clearly, the variability of pt is much greater than when Et�1 pt is the target variable, especially for large values of 1 . In the model at hand, then, the stabilising effect of monetary policy on the in¯ation rate is greater when Et�1 pt , rather than Et�1 pt1 , is the variable responded to. That property does not obtain for all model speci®cations, of course. In both of the ®rst two panels, we see that application of stronger feedback to in¯ation rate discrepancies has the effect of increasing the variability of y~ t , the output gap. In the third panel we consider policy responses to the output gap, as well as to in¯ation. In particular, we assume that the interest rate instrument is adjusted upward when Et�1 y~ t is positive ± i.e. when output is expected to exceed its natural rate value. As 2 is increased ± i.e. moving to the right in the table ± we see that the variability of y~ t falls, as one would expect. Thus it is the case that systematic monetary policy can be used to stabilise output (i.e. keep yt close to y t ) in this model, despite its highly classical long-run properties. When 2 is increased with constant 1 and 3 , the variability of in¯ation increases. Another way to see the effect of the systematic portion of monetary policy is to compare impulse response functions for different values of policy rule parameters. Let us return to the case represented in the ®rst panel of Table 1.1 i.e. with 2 0, 3 0.8, and 1 varied over a wide range. In this context, the impulse response function for the policy shock itself is not as interesting as for some of the other shocks. Let us consider ®rst a shock to the expectational IS function ± i.e. a shock to preferences that increases the demand for current
Methodological Issues 19
consumption in relation to future consumption. Impulse response functions for the variables yt , pt , pt , qt , st , and Rt are shown in Figure 1.1a (for a unit shock) when 1 0.1 ± i.e. when policy response is very weak. By comparison, the same responses are shown in Figure 1.1b for a very strong policy response, with 1 5. We see that the response of output to the shock is slightly greater in the second case, but that the responses of in¯ation and the price level (pt and pt ) are much smaller; please note the vertical axis scalings. Furthermore, the response pro®les for qt and st , the real and nominal exchange rates, are not even of the same shape in the two panels.21 Clearly, the systematic component of monetary policy has major effects on the way in which the economy responds to demand shocks (IS shocks) in this model. Also of interest is the difference of the response patterns to a shock to the uncovered interest parity (UIP) relation ± i.e. a foreign exchange risk premium shock. Figure 1.2 includes panels for the same two policy rules as re¯ected in Figure 1.1. Thus the top panel, Figure 2a, obtains when policy responds very weakly to in¯ation target misses while the bottom panel, Figure 2b, is for very strong responses. Again, the responses of pt and pt are distinctly muted by stronger policy behaviour (larger 1 values). By the `overshooting' mechanism, consequently, the nominal and real exchange rate responses are larger when 1 is large. Finally, let us consider a technology shock, one that increases the value of yt .22 Output and real exchange rate responses are not dissimilar with 1 0.1 and 1 5.0, but the response of nominal variables is drastically different with the different 1 values ± see Figures 1.3a and 1.3b. With 1 0.1, in¯ation falls and then very slowly returns to zero, whereas with 1 5.0 in¯ation brie¯y rises. As a result, the time pro®les for the price level and the nominal exchange rate are extremely different. All in all, the differences depicted in Figures 1.1± 1.3 re¯ect the effects of the systematic component of monetary policy behaviour in response to shocks of the type that are crucial in the implementation of monetary policy.
4 Model speci®cation The previous section has suggested some procedures for characterising the effects of the systematic component of monetary policy for a given structural model. But of course different models generate very different effects, so it is essential to have a strategy for developing a good structural model. Most researchers would agree that it is desirable for a model to be consistent with both economic theory and empirical evidence, but that dual requirement is only a starting point for consideration of numerous issues. Like many economists, I have been persuaded that it is a desirable practice to begin with the construction of a general equilibrium model in which individual agents are depicted as solving dynamic optimisation problems. As mentioned above, such a step is neither necessary nor suf®cient for obtaining
0.2
0.2
0.1
0
0
q
y
20 Bennett T. McCallum
–0.2 0
10
20
0
0
10
20
0
10
20
s
p
0
–0.5
–0.5 0
10
20 0.05
–0.1
0
Δp
R
0
–0.2 0
–0.05 10
20
0
10
20
10
20
10
20
10
20
0.2
0.4
0.1
0.2
0
q
y
a Responses to unit shock to IS; μ1 = = 0.1 0.1
0
10
0 0
20
0
p
S
0.5
–0.05
0
10
–0.05 0 0
20
–0.02
–0.02
R
Δp
0 –0.01
0
0
–0.04
0 10 20 b Responses to unit shock to IS; μ1 = 5.0
Figure 1.1 Impulse response functions, basic model (IS shock)
0.1
2
0.05
1
q
y
Methodological Issues 21
0 0
10
20
0
10
20
0
10
20
–0.1 0
10
20
0
10
20
0
10
20
0
10
20
0
2
–0.5
0
s
p
0
–2
–1 0
10
20 0
–0.5
–0.05
R
Δp
0
–1 0
10
20
0.4
4
0.1
2
q
y
a Response to unit shock to UIP; μ1 = 0.1
0
0 10
20
0
5
–0.1
0
s
p
0
–5
–0.2 0
10
20 0
0
–0.1
Δp
R
0.2
–0.2
0
10
20
–0.2
b Response to unit shock to UIP; μ1 = 5.0
Figure 1.2 Impulse response functions; basic model (UIP shock)
0.4
1
0.2
0.5
q
y
22 Bennett T. McCallum
0
0 10
20
0
2
–1
0
s
p
0
–2
–2 10
20
0
0.1
–0.1
0
R
Δp
0
0
10
20
0
10
20
0
10
20
–0.1
–0.2 0
10
20
0.4
1
0.2
0.5
q
y
– μ = 0.1 a Responses to unit shock to y; 1
0
0 10
20
0.05
1
0
0.5
s
p
0
–0.05
10
20
0
10
20
0
10
20
0 0
10
20
0.05
0.1
0
0
R
Δp
0
–0.1
–0.05 0
10
20
– μ = 5.0 b Responses to unit shock to y; 1
Figure 1.3 Impulse response functions, basic model (y shock)
Methodological Issues 23 Table 1.2 Simulation results for model variants 2 0.0; 3 0.8 Case
std. dev. of
1 0.1
0.5
1.0
5.0
50.0
h 0.8 Et�1 pt as target
p y~ R
10.54 1.45 2.17
7.37 1.83 2.55
5.48 2.12 2.86
1.91 2.71 3.77
0.24 3.08 4.50
h 0.0 Et�1 pt as target 2 0
p y~ R
7.01 1.38 3.35
6.09 2.02 4.02
5.33 2.83 5.03
2.54 5.47 8.09
0.37 7.11 9.88
Closed econ. Et�1 pt as target
p y~ R
18.33 1.50 1.47
5.98 1.42 1.03
2.87 1.41 0.99
0.59 1.42 0.86
0.06 1.44 0.84
F±M price adj Et�1 pt as target
p y~ R
3.87 1.77 2.84
3.48 2.14 3.15
3.15 2.45 3.51
2.54 3.96 6.87
1.87 8.27 27.56
F±M price adj Et�1 pt1 as target
p y~ R
3.97 1.81 2.89
3.07 2.03 3.02
3.36 2.39 3.57
2.63 3.54 5.99
2.00 6.81 20.47
a good model, but is useful in tending to reduce inconsistencies and forcing the modeller to think about the economy in a disciplined way. But adherence to dynamic optimising general equilibrium analysis still leaves room for enormous differences among models ± even ones that are of the same scale and include the same variables. In this section I will attempt to discuss a few of the crucial speci®cational issues, illustrating their importance by various comparisons with the model introduced in section 3. One non-standard feature of that model is the presence of `habit formation' in consumption behaviour. Much more common is a speci®cation with a time-separable intertemporal utility function ± as utilised, for example, by Rotemberg and Woodford (1997), Kerr and King (1996), or McCallum and Nelson (1997). This more standard speci®cation can be obtained as a special case of the model of Section 3 simply by setting the parameter h at the value of zero ± see (1.8). If that is done and simulations like those of Table 1.1 are conducted, the results with Et�1 pt as the target variable are shown in the second panel of Table 1.2, where the ®rst panel repeats a comparable reference case from Table 1.1. It will be seen that the differences in RMSE values for in¯ation (for various 1 values) are somewhat smaller than in the reference case, and that the RMSE differences for y~ t are substantially larger. Intuitively,
24 Bennett T. McCallum
setting h 0.0 eliminates a source of consumption persistence that obtains with h 0.8. This change leaves output free to respond more strongly to shocks and simultaneously makes in¯ation more controllable by policy actions. The difference in behaviour when h 0 rather that h 0.8 shows up even more dramatically in impulse response functions. For the purpose of this comparison, we set 1 at the intermediate value of 1.0 while keeping 2 0 and 3 0.8. The responses to a monetary policy shock are shown in Figures 1.4a and 1.4b, where the case with h 0 is in part b. Comparing the upper left panels from the two sets, we see that the response of yt to a unit Rt shock is almost three times as large when habit formation is absent. Even more drastically, we see that with h 0 the response of in¯ation (and the price level) is in the opposite direction from that with h 0.8: an unanticipated increase in Rt causes pt to rise temporarily, rather than fall. This is the same property that was observed to obtain in the simple analytical model of Section 2. Most readers will probably ®nd it implausible, although it is reminiscent of the `price puzzle' empirical results that have some tendency to arise in VAR studies with models that include no commodity-price variable (Sims, 1992). But the main point to be stressed here is that behaviour is quite different in models with h 0 and h 0.8.23 This difference also obtains in response to a vt shock, as is illustrated in Figure 1.5. Next we consider the effect of removing the open-economy features of our basic model ± i.e. converting it to a closed-economy speci®cation.24 Comparing the third panel of Table 1.2 with the ®rst (reference) panel shows that the in¯ation RMSE values are much more sensitive to 1 in the closedeconomy speci®cation ± i.e. in¯ation is more readily controllable by monetary policy. The different 1 settings have much less effect on y~ t variability, however. One problem with the open-economy speci®cation of our model is that exports and imports are assumed to respond promptly ± within the period ± to changes in their determinants, i.e. the real exchange rate, foreign income and domestic income. It might be more realistic to assume instead that imports and exports respond in a distributed-lag fashion. The third speci®cational change to be considered involves the price adjustment relation. With the P-bar speci®cation, our model generates a great deal of persistence in both yt and y~ t , but not much for the in¯ation variable pt: This absence can be seen in terms of the response of pt to a policy shock in the lower-left panel in the top half of Figure 1.4; there is some persistence but not much. The autocorrelation functions for y~ t and pt , which of course re¯ect responses to all shocks, are as shown in Table 1.3. From the study of Nelson (1998), it is known that as of 1997 most existing quantitative models designed to incorporate both sticky prices and optimising behaviour feature little if any in¯ation persistence.25 One notable exception is the model of Fuhrer and Moore (1995), which was designed so as to provide a good match to the US autocorrelation functions for y~ t , pt , and Rt .26
Methodological Issues 25
0
y
q
0
–0.5
0
10
–5
20
0
10
20
–1
0
10
20
0
10
20
s
p
0
0
–5
–2 0
10
20 1
0
0.5
Δp
R
1
–1 0
10
0 20
a Responses to unit shock to policy rule, with h = 0.8 0
0
y
q
–1
–2
–10 0
10
0
20
2
10
1
0
s
p
–5
10
20
–10
0 0
10
0
20 2
0.2
1
10
20
10
20
Δp
R
0.4
0 0
10
0 20
0
a b Response to unit shock to policy rule, with h = 0.0
Figure 1.4 Impulse responses to monetary shock
0.4
0.4
0.1
0.2
q
y
26 Bennett T. McCallum
0
0 10
10
20
0
10
20
0
10
20
0
0.2
–0.1
0
–0.2
–0.2 0
10
20
0
0
–0.05
–0.02
R
Δp
0
20
s
p
0
–0.04
–0.1 0
10
20
0.2
1
0.1
0.5
q
y
a Responses to unit shock to IS; h = 0.8
0
0 10
10
20
0
10
20
0
10
20
0
1
–0.5
0
–1
–1 0
10
20
0
0
–0.5
–0.1
R
Δp
0
20
s
p
0
–0.2
–1 0
10
20
b Responses to unit shock to IS; h = 0.0
Figure 1.5 Impulse responses to shock to IS
Methodological Issues 27 Table 1.3 Autocorrelation functions Policy parameters: 1 1.0, 2 0, 3 0.8 US dataa Lag 0 1 2 3 4 5
Basic model
Model with (1.10) replacing (1.2)
y~t
pt
y~t
pt
y~t
pt
1.000 0.970 0.910 0.841 0.769 0.703
1.000 .875 .827 .798 .776 .719
1.000 .870 .758 .655 .567 .487
1.000 .283 .051 ±.013 ±.017 ±.019
1.000 .904 .814 .726 .637 .549
1.000 .821 .666 .531 .415 .315
Note: a Quarterly, 1955:1±1996:4. It should be noted that the the output gap in the ®rst column is measured as in McCallum and Nelson (1997), not by any detrending procedure based only on the output series itself.
Consequently, it is of interest to determine how replacement of the P-bar price-adjustment speci®cation with that of Fuhrer and Moore (F±M) would affect our model. In fact, we will adopt a slightly modi®ed version of the F±M speci®cation with two-period contracts, a version that has been used by Fuhrer (1997) and Isard, Laxton, and Eliasson (1999), among others. Speci®cally, we consider the following price adjustment model: pt
1 � !Et pt1 !pt�1 y~ t ut
1:10
With ! 0.5, this relation is almost identical to the F±M speci®cation, as has been shown by Walsh (1998, pp. 224±5).27 Here ut re¯ects the random, unsystematic component of pricing behaviour; it is assumed to be white noise. For inclusion in our simulation analysis, numerical values must be attached to and 2u E(u2t ). On the basis of results in Isard, Laxton, and Eliasson (1999), I have adopted 0.0032 and 2u (0.0025)2 .28 The fourth panel of Table 1.2 reports RMSE values for simulations with the same policy rule settings as before. As can be seen, the extent to which in¯ation variability depends upon 1 (our measure of policy activism) is much less than with the P-bar model. That is because the F±M speci®cation features much more built-in in¯ation inertia, and also because of the presence in (1.10) of the ut component that is not present in our P-bar cases. The sensitivity of y~ t variability to the policy rule is, correspondingly, considerably greater than with the P-bar model. Autocorrelation coef®cients for pt and y~ t are shown in the two rightmost columns of Table 1.3. In keeping with our understanding of the nature of the F±M speci®cation, in¯ation persistence is much greater than with the P-bar model, and much closer to that found in the US data. The persistence of y~ t is about the same as in our basic model ± i.e. quite substantial although less than in the US data. In the ®fth panel of Table 1.2, the policy
28 Bennett T. McCallum
feedback rule responds to Et�1 pt1 rather than Et�1 pt , so as to determine whether this type of `preemptive' response is more effective in the presence of additional in¯ation inertia. As can be seen, however, the results are not greatly affected by this change. Impulse response functions for the F±M pricing speci®cation are shown in Figures 1.6 and 1.7; the policy rule has 1 1.0 and responds to Et�1 pt1 . The additional in¯ation inertia provided by this model shows up quite clearly in the lower left-hand panels. It is worth noting that although our P-bar model does not give rise to much in¯ation persistence, it does account for a considerable amount of persistence in output.29 This ®nding con¯icts with claims made recently by various writers, including Chari, Kehoe and McGrattan (1995), Christiano, Eichenbaum and Evans (1997) and Andersen (1998). The reason that such a disagreement is possible is that these authors all presume that slow adjustment or staggering of goods prices is combined with continuous clearing of the labour market. But what is assumed in the models given above ± as well as in the work of McCallum and Nelson (1997,1998), Taylor (1979, 1993b), Fuhrer and Moore (1995) and many others ± is that ®rms produce whatever quantity is demanded at the prevailing price with labour supplying as much labour as is needed (given capital, technology and the production function). Current wages in this arrangement are irrelevant for labour quantity determination, except via effects on prices, as in the `instalment payment' discussion of Hall (1980). Labour supply conditions are important only in the determination of y t , not yt ± y t .30 As a consequence, these models do not imply that a contractionary monetary shock leads to a rise in pro®ts, as suggested by Christiano, Eichenbaum and Evans (1997). One important speci®cational issue that will not be explored quantitatively concerns the absence of any monetary variables in the basic model of Section 3. In this respect that model is consistent with most recent analysis of monetary policymaking, as represented in the NBER conference volume edited by Taylor (1999). But is it actually reasonable to conduct monetary policy analysis using an analytical framework that includes no money demand function and indeed no reference to any monetary aggregate, either narrow or broad? At a super®cial level, this question is answered by the well known point that if a money demand function were appended to a basic model such as (1.1)±(1.3), its only role would be to determine how much money would have to be supplied to implement the interest instrument policy rule; implied paths of yt , pt and Rt would be entirely unaffected. At a less super®cial level, however, the question becomes one that asks whether an optimising speci®cation, with the medium-of-exchange role of money properly recognised, would yield an expectational IS function that includes no real money balance terms. The answer to that question is that such terms are absent only when the implied `indirect utility function' for the optimising household is additively separable in consumption and real money balances. Thus formulations of the expectational IS function of the type that
1
5
0
0
q
y
Methodological Issues 29
–5
–1 0
10
20
10
20
0
10
20
0
10
20
s
p
–0.5 –1
–5 0
10
20
0
1
–0.05
0
R
Δp
0 0
0
–1
–0.1 0
10
20
a Responses to unit shock to policy rule
0
0
q
y
0.2
–0.05
–0.2 0
10
20
10
20
0
10
20
0
10
20
0
s
p 0
0 0.05
0.05
–0.05 0 x10–3
10
20 0.01
R
Δp
5
0.005 0
0 0
10
20
b Responses to unit shock to IS
Figure 1.6 Impulse responses with F±M (1.10)
0.1
2
0
0
q
y
30 Bennett T. McCallum
–0.1
–2 10
0.02
2
0.01
1
0
20
0 x10–3
10
20
10
20
0
10
20
0
10
20
0
10
20
20 4
0
2
R
Δp
10
5
–5 0
0 10 20 0 a Responses to unit shock to UIP
1
2
0.5
1
q
y
10
0 0 x10–3
0
0 0
10
20
0
2
–0.5
0
s
p
0
20
s
p
0
–1
–2 0
10
20 0.1
–0.05
0
R
Δp
0
–0.1
–0.1 0
10
20
–
Responses unit shocktotoytY bb Responses toto unit shock
Figure 1.7 Impulse responses with F±M (1.10)
Methodological Issues 31
are prevalent in the literature ± and used above ± depend upon this separability assumption. To evaluate whether such an assumption is appropriate, one would have to consider alternative ways of modelling the role of the medium of exchange ± e.g. money in utility function, shopping time, transaction cost or cash-in-advance setups ± and alternative functional forms (complete with quantitative properties). Such a study is far beyond the scope of the present paper, so I will end this discussion simply by noting that separability is not compatible with the shopping-time formulation utilised by McCallum and Goodfriend (1987).
5 Model diagnostics In the previous section it has been demonstrated that changes in speci®c details of a structural model can make major differences in its policy-relevant dynamic properties. Consequently, it is important to have a strategy for conducting model diagnostics, so as to ascertain readily and reliably which models or model variants are more nearly consistent with actual macroeconomic data. In this regard there are clearly many ways to proceed, since there are alternative ways of presenting the various second moments relevant to a model's performance. Again I would like to suggest, however, a procedure that differs from the ones most common in the VAR-oriented literature. More speci®cally, I would suggest that vector autocorrelation functions of the type utilised by Fuhrer and Moore (1995) ± but augmented by univariate variance statistics ± may be a more fruitful source of information than the impulse response functions that are more frequently emphasised. The basic point is as follows. There are reasonably robust procedures, developed by Christiano, Eichenbaum and Evans (1997) and Bernanke and Mihov (1998), for identifying monetary policy shocks, but these procedures do not identify the other structural disturbances of a dynamic macroeconomic model.31 Therefore, these procedures do not automatically provide data analysis counterparts to be compared with impulse response functions for the candidate model's structural shocks other than the policy instrument shock.32 By contrast, vector autocorrelation functions for actual data constitute pure descriptive statistics that can be readily compared with analogous statistics implied by a candidate model. It is of course true that autocorrelations in two sets of (actual or hypothetical) data could agree while the autocovariances nevertheless differed in magnitude. Accordingly, one needs to modify the Fuhrer±Moore statistics to re¯ect autocovariances rather than autocorrelations. Or, equivalently, one could augment the autocorrelations with magnitudes of the variances of each variable in the (actual or hypothetical) data set. To me the latter possibility seems somewhat more attractive, since it divides the comparison into two parts. The univariate variances indicate whether the variability of the model's variables matches that in the data, while
32 Bennett T. McCallum Table 1.4 Model and data variability p
y~
R
a US data, 1955:1±1996:4
variancea std dev.b
0.36 2.41
4.97 2.23
0.49 2.80
b Basic model (0.5, 0.4, 0.8)
variance std. dev.
4.11 8.11
3.10 1.76
0.53 2.92
c Basic model But with h 0
variance std dev.
2.56 6.40
3.13 1.77
1.00 4.00
d Model with (1.10)
variance std dev.
0.87 3.74
3.52 1.88
0.65 3.21
e Model with (1.10) and h 0
variance std dev.
1.04 4.07
2.56 1.60
0.94 3.89
Notes: a Variance statistics are quarterly fractional units multiplied by 104 in all panels. b Standard deviations are in percentages, annualised for p and R, in all panels.
1
1
1
0
0
0
–1
–1 0
20
–1 0
40
20
40
1
1
1
0
0
0
–1
–1 0
20
40
20
40
1
1
0
0
0
20
40
40
0
20
40
0
20
40
–1
–1 0
20
–1 0
1
–1
0
0
20
40
– Autocorrelation in US US data date Autocorrelationfunctions functionsfor forΔΔp,yy˜,, RR in
Figure 1.8 Autocorrelation functions for US data
Methodological Issues 33
1
1
1
0
0
0
–1
–1 0
20
–1 0
40
20
40
1
1
1
0
0
0
–1
–1 0
20
40
20
40
1
1
0
0
0
20
40
40
0
20
40
0
20
40
–1
–1 0
20
–1 0
1
–1
0
0
20
40
Δp, R ininbasic basicmodel model Autocorrelation functions for Δ p, yy˜~,, R
Figure 1.9 Autocorrelation functions for basic model
the autocorrelation magnitudes and patterns re¯ect the nature of the dynamic interrelationships. Any major discrepancy on any of these dimensions ± any discrepancy between a model's properties and actual data ± re¯ects a weakness in the model's speci®cation. This argument presumes, clearly, that the policy rule in the model simulations and the various shock variances are realistically matched to the ones that prevailed over the sample period during which the data was generated. The foregoing objection to impulse response methods does not pertain, of course, to VAR systems in which all shocks, not just the one associated with monetary policy actions, are identi®ed. Examples are provided by Sims (1998a), Blanchard and Watson (1986) and many others. But the identifying restrictions in these systems are much more demanding and less credible than in the semi-structural systems promoted by Bernanke and Mihov (1998a), Christiano, Eichenbaum and Evans (1998) and others who seek robustness. But whether this point of view is persuasive or not, the vector autocorrelation strategy seems at least somewhat attractive because of its purely descriptive nature (as mentioned by Fuhrer and Moore, 1995). Accordingly, an illustrative
34 Bennett T. McCallum
1
1
1
0
0
0
–1
–1 0
20
–1 0
40
20
40
1
1
1
0
0
0
–1
–1 0
20
40
20
40
1
1
0
0
0
20
40
40
0
20
40
0
20
40
–1
–1 0
20
–1 0
1
–1
0
0
20
40
Autocorrelation functions for Δp, y~, R in basic model, h = 0.0
Figure 1.10 Autocorrelation functions for basic model (h 0.0)
application will be presented as the remainder of this section. For the following experiments, we use policy rule (1.9) with parameter values 1 0.5, 2 0.4 and 3 0.8. These are chosen to be representative of actual policy behaviour, as estimated by McCallum and Nelson (1998) following Clarida, Gali and Gertler (1998). We begin by combining this rule with our basic model and comparing its autocorrelation properties (plus variances) with those of actual data. For the latter, I use seasonally adjusted observations over 1955:1±1996:4 on pt , y~ t and Rt as described in McCallum and Nelson (1998). The three variances ± alternatively reported as annualised percentage standard deviations ± are shown in panel a of Table 1.4, with autocorrelations presented in Figure 1.8. A comparison of panels a and b of Table 1.4 indicates that the basic model implies variances of a realistic magnitude for y~ and R, but much too large for p. In addition, the autocorrelations depicted in Figure 1.9 fail badly to match those of Figure 1.8 in all panels except those for the own autocorrelations of y~ and R. Two of the three contemporaneous correlation coef®cients are of the
Methodological Issues 35
1
1
1
0
0
0
–1
–1 0
20
–1 0
40
20
40
1
1
1
0
0
0
–1
–1 0
20
40
20
40
1
1
0
0
0
20
40
40
0
20
40
0
20
40
–1
–1 0
20
–1 0
1
–1
0
0
20
40
Autocorrelation functions for Δp, ~ y, R in model with (1.10)
Figure 1.11 Autocorrelation functions for model with (1.10)
same sign in the model and in the data, but only one (for p and R) represents a reasonably close quantitative match. The third panel in Table 1.4 and Figure 1.10 pertain to the same model except with h 0 ± i.e. with habit formation eliminated from the households' saving decision. Surprisingly, this elimination slightly increases the persistence of in¯ation. But it does not overcome the other major problems with the basic model. Next we turn to the model in which the price adjustment (1.10) replaces the P-bar speci®cation. Now the variance magnitudes, reported in panel d of Table 1.4, are closer to those in the data. And the own autocorrelation functions shown in Figure 1.11 are distinctly more similar to those of Figure 1.8. Indeed, they provide a match that might be judged as semi-respectable. But the cross autocorrelations match quite poorly, especially those involving y~ . Thus this chapter's ®ndings are basically consistent with those reported by Fuhrer (1997, 1998).33 Setting h 0, in Figure 1.12 and panel e of Table 1.4, worsens the match between model and data, especially in terms of the variance magnitudes.
36 Bennett T. McCallum
1
1
1
0
0
0
–1
–1 0
20
–1 0
40
20
40
1
1
1
0
0
0
–1
–1 0
20
40
20
40
1
1
0
0
0
20
40
40
0
20
40
0
20
40
–1
–1 0
20
–1 0
1
–1
0
0
20
40
Autocorrelation functions for Δ p, y˜y~,, RR in Δp, in model model with with (1.10) (1.10) and and hh == 0.0 0.0
Figure 1.12 Autocorrelation functions for model with (1.10) (h 0.0)
What can be concluded from these exercises concerning the usefulness of the variance plus vector-autocorrelation approach to model diagnostics? The basic similarity across Figures 1.9±1.12 of several of the panels suggests that the autocorrelation properties are not as sensitive to a model's speci®cation as are the impulse response properties. This is admittedly a mark against the former approach, but arguably one that is not as serious as those against the impulse response approach that are mentioned on p. 31. The crucial fact, I would suggest, is that the variances and autocorrelations together are able to indicate (1) clear discrepancies relative to the actual data and (2) clear differences among model variants ± both without requiring any (inherently dubious) identi®cation assumptions other than those used in developing the model.
6
Other approaches
Considerable attention has been devoted, during recent years, to the `narrative approach' to measuring the effects of monetary policy that was pioneered by Romer and Romer (RR) (1989). As is well known, the RR study generated a
Methodological Issues 37
number of dates at which the Fed is judged to have `exogenously' adopted a more stringent policy stance in order to reduce in¯ationary pressures. Shapiro (1994), Leeper (1997) and others have noted, however, that responses to in¯ation pressures are clearly not exogenous in the sense relevant to the systematic versus shock decomposition. Shapiro's study represents an improvement in that regard, but would still seem open to the criticism that it builds upon a measure of policy actions that differs from the usual ones in the following three respects: (1) Traditional measures use variables that can range over a near-continuum of values, rather than only two, so can distinguish between major and minor actions. (2) The RR dummy-variable measure recognises as non-zero actions only those in a contractionary direction, leaving decisions to be unusually expansionary to be included together with normal behaviour. (3) Values of the RR dummy are based on what it is that the Federal Open Market Committee's records say, not on what the Open-Market Desk actually does. All in all, then, it is dif®cult to understand the enthusiastic reception that this approach has received. In any event, the present chapter is concerned more with the systematic portion of policy rather than the portion toward which the RR approach is ostensibly directed. A striking and unusual analysis was put forward by Sims (1998b). In this paper, Sims utilises VAR procedures but in a different and bolder fashion than that ± typi®ed by Bernanke and Mihov (1998) or Christiano, Eichenbaum and Evans (1998) ± mentioned in Section 2. In particular, Sims (1998b) estimates an identi®ed VAR system for US monthly data from the interwar period 1919:08±1939:12 and then conducts counterfactual historical simulation policy analysis by replacing the estimated monetary policy rule ± represented by the VAR equation explaining movements in the Federal Reserve discount rate ± with one estimated on postwar data 1948:08±1997:10. The striking ®nding emphasised by Sims is that this replacement has very little effect on the estimated time path of real output that would have obtained over 1919± 39, given the estimated shocks of that era. In other words, monetary policy as practised during postwar years would not have prevented the Great Depression. This dramatic conclusion is reinforced by Christiano's (1999) ®nding that Sims' conclusion is not overturned by replacement of his discount rate rule with one that generates a much more expansionary time path for the M1 money stock. Both of these studies are, clearly, open to Lucas Critique objections. Both authors recognise that problem, Sims contending that the usual objection is philosophically ¯awed (1998b, pp. 154±6) and Christiano leaving it for consideration by his readers (1999, p. 4). My own belief is that the relationship between macroeconomic variables ± nominal income or prices and output ± and monetary policy variables during the interwar years cannot be satisfactorily represented by a model that does not include some variable representing the effects of ®nancial crises. In my (1990) study, for example, I found that the relationship between M1 growth and growth in the monetary
38 Bennett T. McCallum
base was strongly in¯uenced by a measure of current bank failures (prior to the creation of the FDIC).34 Another VAR-oriented approach to the study of systematic monetary policy responses was developed by Bernanke, Gertler and Watson (1997), who concluded that a large fraction of the US economy's real effects from oil price shocks since 1970 has resulted from the monetary policy response to these shocks, rather than from the shocks themselves. The study is concerned with attributing historical ¯uctuations to various sources, rather than with the type of characterisation attempted in the present chapter. Bernanke, Gertler and Watson (1997, p. 93) state that their method `certainly is not invulnerable to the Lucas critique'. Closer in spirit and approach to the present chapter is a study by Dotsey (1999). It, too, utilises simulations with a setup that features maximising behaviour and aspires to the development of a policy-invariant, structural model, and it reaches similar conclusions. One major difference relative to the present chapter is that Dotsey's comparisons are made across entirely different policy rule speci®cations, rather than across different parameter settings for variants of a single rule, as is typically the case in Sections 3±5 above.
7
Conclusions
Let us conclude with a very brief summary. The chapter has argued that, in studying the monetary policy transmission process, more emphasis should be given to the systematic portion of policy behaviour and correspondingly less to random shocks ± basically because shocks account for a very small fraction of policy instrument variability. Analysis of the effects of the systematic part of policy requires structural modelling, rather than VAR procedures, because the latter do not give rise to behavioural relationships that can plausibly be regarded as policy-invariant. By use of an illustrative structural speci®cation with variants, the chapter characterises the effects of policy parameter settings by means of impulse response functions and rootmean-square statistics for target errors. Different models give different answers to questions about the effects of systematic policy, so procedures for scrutinising model speci®cation are essential. In this regard, it is argued that vector autocorrelation functions, augmented by variance statistics for each of a model's variables, seem more promising than impulse response functions because the latter require shock identi®cation, which is inherently a dif®cult process. Notes 1. I am indebted to Miguel Casares, Larry Christiano, Marvin Goodfriend, Jeffrey Lacker, Ellen McGrattan, Allan Meltzer, Edward Nelson and Harald Uhlig for helpful commments and suggestions.
Methodological Issues 39 2. Meltzer's contribution to another 1995 symposium entitled `Channels of Monetary Policy', sponsored by the Federal Reserve Bank of St Louis, focuses instead on the role of nominal price stickiness. 3. Some examples are Cochrane (1994), Sims (1992), Christiano, Eichenbaum and Evans (1998), Chari, Kehoe and McGrattan (1996), and Bernanke and Blinder (1992). 4. An extensive and sophisticated surrey of this portion of the literature is provided by Christiano, Eichenbaum and Evans (1997). Also see Bernanke and Mihov (1998) and Sims (1992). 5. I thank Richard Clarida for providing me with the relevant standard deviations. 6. It is well known that most models with non-instantaneous price adjustment behaviour do not satisfy this proposition. 7. The natural rate concept used in this chapter is the value of yt that would prevail if prices were fully ¯exible ± i.e. if there were no nominal stickiness in the economy. 8. That some fundamental points can be unaffected by this type of neglect is demonstrated by two examples in McCallum and Nelson (1999). Such is de®nitely not true in general, however. 9. References are Calvo (1983) and Rotemberg (1982). 10. In particular, the P-bar model satis®es the natural rate hypothesis ± i.e. that E(yt ± y t ) 0 for any monetary policy rule, which is not the case for the other two speci®cations. 11. For an extensive discussion of the MSV solution concept, see McCallum (1999). 12. The perverse response of pt to a policy shock is not a general implication of the Pbar model, as will be seen below. 13. If the central bank is mistaken in its belief about the average value of r, this will result in an average in¯ation rate that differs from the target . 14. Is there any evidence in approaches oriented toward structural relations such as (1.1)±(1.3)? Given the system (1.1)±(1.3), there is no hypothesis of this type testable with data from a single regime. But there is the possibility of testing speci®cations (1.1) and (1.2) against alternatives that imply the presence of money illusion. Incidentally, as 1 ! 1 the solution becomes arbitrarily close to one in which Rt is set so as to make Et�1 pt1 *. This indicates that there is not much difference between `instrument rules' and the `forecast targeting' procedure emphasised by Svensson (1999). 15. This statement presumes that expectations are rational. 16. Bernanke, Gertler and Watson (1997, p. 92) state that `It is not possible to infer the effects of changes in policy rules from a standard identi®ed VAR system'. Important references to the literature include Bernanke (1986), Bernanke and Blinder (1992), Blanchard and Watson (1986), Christiano and Eichenbaum (1992) and Christiano, Eichenbaum, and Evans (1998). 17. This version is due to Lucas (1972). For a brief discussion, see McCallum and Nelson (1997). 18. As mentioned above, we treat capital as exogenously determined. 19. The model is calibrated by reference to relationships estimated in various studies with quarterly data. A value of 0.8 for h was estimated by Fuhrer (1998). 20. I am not entirely happy with this practice, which implicitly attributes knowledge to policymakers that they could not actually possess. 21. The asymptotic effect on qt is nevertheless the same ± zero ± in the two cases. 22. Although y t depends on raw material inputs, it also has a technology-shock term, as in the real business cycle literature. This shock process is autoregressive of order 1, with parameter 0.95.
40 Bennett T. McCallum 23. This difference has been usefully emphasised by Fuhrer (1997, 1998). A paper by Estrella and Fuhrer (1998) stresses that `standard models' with h 0 and a Calvo± Rotemberg price adjustment speci®cation are seriously inconsistent with the data. I agree with that judgement but see no reason to conclude that all optimising models with rational expectations are unsatisfactory in this regard. 24. To close down the basic model, it is necessary not only to eliminate exports and imports, but also to adjust the variance of the shock term driving y t (because the latter no longer depends on imported raw materials). For more explanation, see McCallum and Nelson (1998). 25. To re¯ect price level stickiness, some departure from full optimizing behaviour is required ± e.g. some additional constraint must be imposed relative to a ¯exible price general equilibrium system. There is considerable scope for dispute concerning the relative `rationality' of different departures. 26. The Fuhrer±Moore (1995) paper does not use optimising analysis to generate its consumption behaviour, but instead posits a non-expectational IS function. Also it uses detrended yt as its measure of the output gap. 27. The difference is that (1.10) includes only y~ t in place of 0.5Et (y~ t y~ t 1 ). 28. To get 0.0032 from the Isard, Laxton and Eliasson (1999) value of 3.2, one divides by 400, because they express in¯ation in annualised percentage points, and then divides by 2.5 to re¯ect the slope of an Okun's Law relationship between y~ t and unemployment. 29. Simulation results indicate that yt features signi®cantly more persistence than does y~ t . 30. Recall that y t is the natural rate value of yt that would prevail in the absence of any nominal frictions. 31. There exists some controversy even over the robustness of these procedures. For recent contributions on this topic, see Faust (1998) and Uhlig (1999). 32. It would be possible to judge a model's ®t entirely on the basis of the impulse response functions for the policy shock, as in Rotemberg and Woodford (1997), but this seems undesirable given the small contribution to overall variability coming from this source. 33. They are also somewhat in the spirit of Estrella and Fuhrer (1998), but do not involve the Calvo±Rotemberg pricing equation that was a component of the MN model criticised by Estrella and Fuhrer. 34. My study found that a policy rule that adjusts the monetary base so as to attempt to keep nominal income on a steady growth path would have made the 1930s' fall in nominal income much less severe than actually occurred. This activist feedback rule would have resulted in a much greater expansion of M1 than in Christiano's counterfactual simulation (which was in turn more expansionary than Sims'). What this monetary stimulus would have done for real output depends, of course, on the model utilised.
References Andersen, T. M. (1998) `Persistency in Sticky-price Models', European Economic Review, 42, p. 593±603. Bernanke, B. S. (1986) `Alternative Explanations of the Money-income Correlation', Carnegie±Rochester Conference Series on Public Policy, 25, 49±99. Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of Monetary Transmission', American Economic Review, 82, 901±21. Bernanke, B. S. and M. Gertler (1995) `Inside the Black Box: The Credit Channel of Monetary Policy Transmission', Journal of Economic Perspectives, 9, 27±48.
Methodological Issues 41 Bernanke, B. S., M. Gertler and M. W. Watson (1997) `Systematic Monetary Policy and the Effects of Oil Price Shocks', Brookings Papers on Economic Activity, No. 1, 91±142. Bernanke, B. S. and I. Mihov (1998) `Measuring Monetary Policy', Quarterly Journal of Economics, 113, 869±902. Blanchard, O. J. and M. W. Watson (1986) `Are Business Cycles All Alike?', in R. J. Gordon (ed.), The American Business Cycle, Chicago, University of Chicago Press. Calvo, G. A. (1983) `Staggered prices in a Utility-maximizing Framework', Journal of Monetary Economics, 12, 383±98. Chari, V. V., P. J. Kehoe and E. R. McGrattan (1996) `Sticky Price Models of the Business Cycle: Can the Contract Multiplier Solve the Persistence Problem', NBER Working Paper 5809. Christiano, L. J. (1999) `Discussion', in Beyond Shocks: What Causes Business Cycles?, Boston, Federal Reserve Bank of Boston. Christiano, L. J. and M. Eichenbaum (1992) `Identi®cation and the Liquidity Effect of a Monetary Policy Shock', in A. Cukierman, Z. Hercowitz and L. Leiderman (eds) Political Economy, Growth, and Business Cycles, Cambridge, MA, MIT Press. Christiano, L. J., M. Eichenbaum and C. L. Evans (1997) `Sticky Price and Limited Participation Models: A Comparison', European Economic Review, 41, 1201±49. ÐÐÐÐ (1998) `Monetary Policy Shocks: What Have We Learned and to What End?', NBER Working Paper 6400. Clarida, R., J. Gali and M. Gertler (1998) `Monetary Policy Rules in Practice: Some International Evidence', European Economic Review, 42, 1033±68. Cochrane, J. H. (1994) `Shocks', Carnegie±Rochester Conference Series on Public Policy, 41, 295±364. Dotsey, M. (1999) `The Importance of Systematic Monetary Policy for Economic Activity', Federal Reserve Bank of Richmond Economic Quarterly, 85, 41±59. Dotsey, M. and P. N. Ireland (1995) `Liquidity Effects and Transaction Technologies', Journal of Money, Credit, and Banking, 27, 1441±56. Estrella, A. and J. C. Fuhrer (1998) `Dynamic Inconsistencies: Counterfactual Implications of a Class of Rational Expectations Model', Federal Reserve Bank of Boston, Working Paper 98±5. Faust, J. (1998) `The Robustness of Identi®ed VAR Conclusions about Money', Carnegie± Rochester Conference Series on Public Policy, 49, 207±44. Fuhrer, J. C. (1997) `The (UN)importance of Forward-looking Behavior in Price Speci®cations', Journal of Money, Credit, and Banking, 29, 338±50. ÐÐÐÐ (1998) `An Optimization-based Model for Monetary Policy Analysis: Can Habit Formation Help?', Boston, Federal Reserve Bank of Boston, Working Paper. Fuhrer, J. C. and G. R. Moore (1995) `In¯ation Persistence', Quarterly Journal of Economics, 110, 127±59. Hall, R. E. (1980) `Employment ¯uctuations and Wage Rigidity', Brookings Papers on Economic Activity, 1, 91±123. Isard, P., D. Laxton and A.-C. Eliasson (1999) `Simple Monetary Policy Rules under Model Uncertainty', IMF Working Paper 99/75; also in P. Isard, A. Razin and A. K. Rose (eds) International Finance and Financial Crises, Boston, Kluwer Academic Publishers. Kerr, W. and R. G. King (1996) `Limits on Interest Rates Rules in the IS Model', Federal Reserve Bank of Richmond Economic Quarterly, 82, 47±75. Klein, P. (1997) `Using the Generalized Schur Form to Solve a System of Linear Expectational Difference Equations', University of Stockholm, Working Paper. Leeper, E. M. (1997) `Narrative and VAR Approaches to Monetary Policy: Common Identi®cation Problems', Journal of Monetary Economics, 40, 641±58.
42 Bennett T. McCallum Leeper, E. M., C. A. Sims and T. Zha (1996) `What Does Monetary Policy Do?', Brookings Papers on Economic Activity, 2, 1±63. Lucas, R. E., Jr. (1972) `Econometric Testing of the Natural Rate Hypothesis', in O. Eckstein (ed.), Econometrics of Price Determination, Washington, DC, Board of Governors of the Federal Reserve System. McCallum, B. T. (1990) `Could a Monetary Base Rule have Prevented the Great Depression?', Journal of Monetary Economics, 26, 3±26. ÐÐÐÐ (1999) `Role of the Minimal State Variable Criterion in Rational Expectations Models', Working Paper, NBER Working Paper 7087; also in P. Isard, A. Razin and A. K. Rose (eds) International Finance and Financial Crises, Boston, Kluwer Academic Publishers. McCallum, B. T. and M. S. Goodfriend (1987) `Demand for Money: Theoretical Studies', in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of Economics, Macmillan, also in J. Eatwell, M. Milgate, P. Newman, The New Palgrave Dictionary of Money and Finance, London. McCallum, B. T. and E. Nelson (1997) `Performance of Operational Policy Rules in an Estimated Semi-classical Structural Model', NBER Working Paper 6599; also in Taylor (1999). ÐÐÐÐ (1998) `Nominal Income Targeting in an Open-Economy Optimizing Model', NBER Working Paper, 6675, also in Journal of Monetary Economics, 43, 553±78. ÐÐÐÐ (1999) `An Optimizing IS-LM Speci®cation for Monetary Policy and Business Cycle Analysis', Journal of Money, Credit, and Banking, 31, 296±316. Meltzer, A. H. (1995) `Monetary, Credit (and Other) Transmission Processes: A Monetarist Perspective', Journal of Economic Perspectives, 9, 49±72. Mishkin, F. S. (1995) `Symposium on the Monetary Transmission Mechanism', Journal of Economic Perspectives, 9, 3±10. Nelson, E. (1998) `Sluggish In¯ation and Optimizing Models of the Business Cycle', Journal of Monetary Economics, 42, 303±22. Obstfeld, M. and K. Rogoff (1995) `The Mirage of Fixed Exchange Rates', Journal of Economic Perspectives, 9, 73±96. Romer, C. and D. Romer (1989) `Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Schwartz', NBER Macroeconomics Annual 1989, Cambridge, MA, MIT Press. Rotemberg, J. J. (1982) `Monopolistic Price Adjustment and Aggregate Output', Review of Economic Studies, 44, 517±31. Rotemberg, J. J. and M. Woodford (1997) `An Optimization Based Econometric Framework for the Evaluation of Monetary Policy', NBER Macroeconomics Annual 1997, Cambridge, MA, MIT Press. ÐÐÐÐ (1999) `Interest Rate Rules in an Estimated Sticky Price Model', NBER Working Paper 6618. Rudebusch, G. D. (1998) `Do Measures of Monetary Policy in a VAR Make Sense?' International Economic Review, 39. Shapiro, M. D. (1994) `Federal Reserve Policy: Cause and Effect', in N. G. Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press. Sims, C. A. (1992) `Interpreting the Macroeconomic Time Series Facts: The Effects of Monetary Policy', European Economic Review, 36, 975±1000. ÐÐÐÐ (1998a) `Comment on Glenn Rudebusch's ``Do measures of monetary policy in a VAR Make Sense?' '', International Economic Review, 39. ÐÐÐÐ (1998b) `The Role of Interest Rate Policy in the Generation and Propagation of Business Cycles: What has Changed Since the 30's?', in Federal Reserve Bank of Boston, Beyond Shocks: What Causes Business Cycles?, Boston, Federal Reserve Bank of Boston.
Methodological Issues 43 Svensson, L. E. O. (1999) `In¯ation Targeting: Some Extensions', Scandinavian Journal of Economics, 101, 337±367. Taylor, J. B. (1979) `Staggered Wage Setting in a Macro Model', American Economic Review, Papers & Proceedings, 69, 108±13. ÐÐÐÐ (1993a) `Discretion Versus Policy Rules in Practice', Carnegie±Rochester Conference Series on Public Policy, 39, 195±214. ÐÐÐÐ (1993b) Macroeconomic Policy in a World Economy, New York, W. W. Norton. ÐÐÐÐ (1995) `The Monetary Transmission Mechanism: An Empirical Framework', Journal of Economic Perspectives, 9, 11±26. ÐÐÐÐ (1999) (ed.), Monetary Policy Rules, Chicago, University of Chicago Press. Uhlig, H. (1997) `What are the Effects of Monetary Policy? Results from an Agnostic Identi®cation Procedure', Tilburg University, Working Paper. Walsh, C. E. (1998) Monetary Theory and Policy, Cambridge, MA, MIT Press.
Discussion
Lawrence J. Christiano
1
Introduction
Ben McCallum seeks to make two contributions. One is methodological and the other is more substantive. On the methodological side, the chapter reviews various empirical strategies that can used for evaluating a monetary model. On the substantive side, it presents a particular monetary model and displays a wealth of information about its dynamic properties. Through numerous well constructed experiments, we learn about the empirical implications of alternative ways of capturing price rigidity, about the impact of habit persistence in preferences and about the implications of open economy considerations. These experiments will be particularly useful for other researchers working to construct empirically compelling monetary models. I will organise my remarks about the methodological theme, though in the end I will touch on some substantive issues. My comments focus on the following three topics:
. Monetary policy shocks.
. Second moments and Maximum Likelihood.
. Testing a monetary model using a historical episode.
The perspective that I have on monetary policy shocks differs from that of McCallum, and I begin my discussion by explaining why. In my discussion of the use of second moments in empirical analysis, I relate McCallum's analysis to the existing literature. Finally, in proposing the use of a historical episode to help evaluate a model, I seek to add another empirical strategy to the set considered by McCallum. I argue that evaluating a model's ability to account for a particular episode can be a useful complement to the overall assessment of a model. The episode I consider is the take-off of in¯ation in the early 1970s, an experience that I believe represents an important challenge to McCallum's sticky-price model, and other sticky-price models as well. 44
Discussion 45
2 Monetary policy shocks To clarify the issues surrounding monetary policy shocks, it is useful to begin by de®ning a monetary policy rule: St f
t "t
2:1
Here, St denotes the variable (or, vector of variables) that the central bank controls directly. McCallum follows much of the recent literature by treating St as a short-term interest rate. The central bank's decision variable is assumed to be determined in part as a systematic function, f ; of the state of the economy,
t : But, in practice not all of the variation in St can be accounted for by f
t : The rest is determined by the monetary policy shock, "t : The literature on monetary policy shocks seeks to measure "t and to determine its impact on the major macroeconomic variables. This literature has found that "t accounts for only a small part of the variation in most variables, including St : McCallum infers from this that the analysis of monetary policy shocks is therefore not very useful. He argues that it is much more interesting to study the economic effects of the systematic part of the monetary policy rule, f
t , and the effects of alternative speci®cations of St : But this misses the point of the monetary policy shock literature, at least the part of that literature with which I am familiar.1 In that literature, the purpose of measuring monetary policy shocks, and studying their impact on the economy is precisely to be helpful in the ultimate objective of studying alternative speci®cations of f and S: To see why this is so, recall that to study alternative speci®cations of f and S; one needs a monetary model that can be used as a laboratory for this purpose. There is no practical alternative. We clearly cannot experiment with alternative f and Ss on the real world. And, history does not offer observations on the range of variation in f and S that interest us, for modern economies. So, to study alternative f and Ss, we need a model. And this is precisely where the analysis of monetary policy shocks has been helpful. In principle, there are many models that can serve as laboratories. The analysis of monetary policy shocks has narrowed the range of models useful for this purpose, and continues to provide a guide to further development of models. For example, the class of models called `Lucas misperception' models is rarely used in quantitative monetary studies today. This is due, in part, to the ®nding of the empirical monetary policy shock literature that the price level seems to be the last thing ± after employment, output, inventories, etc. ± to respond to a disturbance in "t : In contrast, the Lucas misperception model appears to imply that the mechanism by which a disturbance, "t ; impacts the economy involves as its ®rst step, a change in the aggregate price level.2 Another class of models, naive extensions on the monetary model of Cooley and Hansen (1989), emphasises anticipated in¯ation effects in the transmis-
46 Lawrence J. Christiano
sion of policy shocks, to the exclusion of liquidity effects. The empirical monetary shock literature appears to support the view that liquidity effects play an important role in translating a monetary policy disturbance into movements in interest rates and money. The empirical monetary policy shock literature also provides other facts that are useful as a guide to constructing monetary models. For example, the ®nding that an expansionary monetary policy action does not generate a signi®cant drop in the real wage severely limits the scope of naive sticky-wage models. The key mechanism in these models, by which an expansionary monetary policy shock is translated into an expansion in employment, operates by reducing the real wage. Similarly, the ®nding that expansionary monetary policy is associated with a rise in pro®ts limits the scope for naive sticky-price models, which tend to imply that pro®ts fall with a monetary expansion. The analysis of policy shocks has been usefully applied in other areas as well. For example, Rotemberg and Woodford (1992) analysed the macroeconomic impact of shocks to government spending to discriminate between alternative models of government spending. This work has been extended in interesting ways by Ramey and Shapiro (1998) and Eichenbaum, Fisher and Edelberg (1999). Ultimately, we are interested in models which incorporate all the signi®cant shocks driving the data. Analyses which focus on one shock are particularly useful as a ®rst step in this enterprise, when the researcher seeks broad guidance in discriminating between different types of models. As the set of models worthy of consideration is re®ned, then other methods for model estimation and diagnosis become relatively more useful. These methods, which include those preferred by McCallum, involve incorporating all shocks into a model and comparing a model's implications with the raw data, rather than just the part driven by one shock or another ± i.e. the impulse response functions and variance decompositions. This is the phase we are in in the analysis of monetary models. The set of models in this literature has been reduced, broadly, to two: models with sticky prices, such as the one studied by Ben McCallum, and `limited participation' models.3 In sum, there is a limited sense in which I agree with McCallum's opinion about the role of monetary policy shocks in the construction of models. I am sympathetic to the idea that it is time to `move on' to analyses which involve more shocks and which compare a model's implications with the raw data. However, it is a mistake to minimise the importance of the shock-analysis literature in this transition. The models we use as we `move on' are models that were selected from a much larger potential class largely as a result of the analysis of shocks. Moreover, in the further analysis of models, it is wise for researchers to bear in mind some of the results from the monetary policyshock literature ± for example, the implications of monetary shocks for pro®ts and real wages mentioned above.
Discussion 47
3 Second moments and Maximum Likelihood The empirical methods that McCallum prefers involve comparing a model's implications for the second moment properties of the data with the corresponding empirical objects. Here, I wish to point out that there is a time-honoured tradition of using this type of analysis in macroeconomics generally and in monetary models speci®cally. Moreover, there are further developments in this type of analysis which could be usefully applied in analyses like McCallum's. In a series of papers, Finn Kydland and Edward Prescott have emphasised the value of evaluating models based on second moment properties. Similarly, a series of papers by Lars Hansen and Thomas Sargent draws attention to Gaussian maximum likelihood and other econometric methods based on second moments in model estimation and testing. In the analysis of monetary models, Chari, Christiano and Eichenbaum (1995) and King and Watson (1996) also pursue the strategy of diagnosing models based on an analysis of second-moment properties of the data. More recently, there have been interesting new developments in this literature, which allow a researcher to evaluate a model based on different frequency components of the data. These include papers by Watson (1993), Diebold, Ohanian and Berkowitz (1998) and Christiano and Vigfusson (1999). Often researchers are interested in understanding how different frequency components of the data impact on model estimation and testing, and this literature provides the formal tools for doing this. This can be useful for two reasons. First, understanding which frequencies the model does poorly on can give guidance to further model development. Second, a researcher may want to react very differently to a statistical rejection of a model if it arises because of poor performance in the low frequencies or in the high frequencies. For example, with a monetary model, one might want to insist particularly that the model does well in the high frequencies.
4 Testing a model using a historical episode One way to test a model, not considered by McCallum, is to see how well it accounts for the events of a particular historical episode. Given the motivation for the type of work in this chapter, I believe there is one episode that deserves special attention. As I understand it, a primary motivation for research on monetary policy rules is to identify strategies for conducting monetary policy that will ensure that we not repeat the monetary policy mistakes of the past. One such mistake is the burst of in¯ation experienced by many countries in the 1970s. In view of this, a natural test of a model like McCallum's is to see whether it can account for this episode. After all, if the cures for past monetary disorders emerging from a particular model framework are to be compelling,
48 Lawrence J. Christiano
that framework must at least be able to provide a plausible account of why those disorders occurred in the ®rst place. There is a general class of explanations for the high in¯ation of the 1970s, according to which it was the result of a weakness either in monetary policy institutions or in the people in charge of them, that made economies like the United States vulnerable to a rise in in¯ationary expectations. These explanations suggest that high in¯ation expectations in the early 1970s ± triggered perhaps by the commodity and oil shortages of the time ± led central banks to react with an accommodative monetary policy.4 Consistent with the observations of Clarida, Gali and Gertler (1998), McCallum's model can articulate a particular version of this view. Monetary policy in McCallum's model is represented in the form of (1), in which St is a short-term rate of interest, Rt ; and f
t constant 3 Rt�1
1 � 3
1 1 Et t1 2 y t where t1 is the in¯ation rate in the quarter following quarter t; Et is the conditional expectation operator and yt is the deviation of output from potential. A feature of this model is that when 1 < 0; then it is possible for higher anticipated in¯ation to be self-ful®lling. Since nothing in this result depends on the value of 3 ; to simplify the intuition I set 3 0: Under these circumstances, higher expected in¯ation can be self-ful®lling via the following mechanism.5 Suppose agents expect higher in¯ation. With 1 < 0; this creates the expectation that the real interest rate will be low, which in turn stimulates investment demand. The rise in investment demand then triggers a rise in output and employment which ultimately translates into higher in¯ation, thereby validating the original jump in in¯ation expectations. Note that, according to McCallum's model, a burst of high in¯ation triggered by expectations in this way is associated with a strong economy. A problem with this argument, however, is that variables like investment and employment were actually low in the early 1970s, particularly during the 1975 recession. An alternative model that has been used for monetary analysis is the limited participation model studied by Christiano (1991), Fuerst (1992) and Christiano, Eichenbaum and Evans (1998b). Christiano and Gust (1999, 2000) show that this model also has equilibria in which higher expected in¯ation can be self-ful®lling when 1 < 0: However, Christiano and Gust (1999b) show that in these equilibria, output and investment are low, a feature that appears more consistent with the data. Thus, it appears that the sticky-price model of McCallum fails the test of the 1970s, a test that the limited participation model passes. There is, however, one important caveat that must be mentioned in this context. There are many things that happened in the 1970s ± the productivity slowdown began, there were oil shocks ± and possibly some of the evidence of weakness that
Discussion 49
Christiano and Gust (2000) attribute to self-ful®lling in¯ation expectations is actually due to these other factors.6 This possibility deserves further exploration. Still, I suspect that the evidence in fact represents a challenge to the kind of sticky price model studied by McCallum.
5 Conclusion In conclusion, McCallum's chapter contains valuable lessons for researchers constructing quantitative, monetary models for use in analysing the operating characteristics of monetary policy rules. The focus of McCallum's chapter is on IS±LM models with sticky prices. McCallum discusses various diagnostic tools for empirically evaluating these models. Drawing on work currently underway with Christopher Gust, I added another tool to the list. The idea is to test McCallum's model by determining whether it provides a plausible account for the burst of in¯ation observed in the early 1970s. There is some reason to think that it cannot. The account articulated by the model for this episode appears inconsistent with the weakness in employment observed at the time, particularly during the 1975 recession. This test suggests that an alternative framework for the analysis of monetary policy rules, the one based on the limited participation model of money, may be superior. Notes 1. This section summarises arguments in Christiano, Eichenbaum and Evans (1998a). 2. For a discussion of this model, see Cooley and Hansen (1997). 3. See Fuerst (1992); Christiano (1991); Christiano, Eichenbaum and Evans (1998) and Alvarez, Atkeson and Kehoe (1999). 4. For an informal discussion of how this might have happened, see Blinder (1982), especially p. 264. Formal analyses appear in Chari, Christiano and Eichenbaum (1998) and Clarida, Gali and Gertler (1998). 5. The observation that 1 < 0 can result in equilibria in which in¯ation expectations are self-ful®lling in a simpli®ed version of McCallum's model (an IS±LM model with rational expectations) was ®rst made in Kerr and King (1996). Ben McCallum was kind enough, in a private communication, to con®rm that this result also holds in the models analysed in his chapter. 6. In Christiano and Gust (2000)'s analysis, they simulate Clarida, Gali and Gertler's (1998) model and a limited participation model under the assumption that the jump in in¯ation expectations in the early 1970s was a reaction to a bad technology shock. This is consistent with the view that the higher in¯ation expectations at the time were in part due to the rise in oil and other commodity prices that occured in the wake of the shortages that occurred at the time. The policy rule used in the simulations uses the parameter values estimated by Clarida, Gali and Gertler (1998) using data from the 1960s and 1970s.
References Alvarez, F., A. Atkeson and P. Kehoe (1999) `Money and Interest Rates with Endogenously Segmented Markets', NBER Working Paper, 7060.
50 Lawrence J. Christiano Blinder, A. (1982) `Anatomy of Double-digit In¯ation in the 1970s', in R. Hall (ed.), In¯ation: Causes and Effects, National Bureau of Economic Research, Chicago, University of Chicago Press. Chari, V.V., L. J. Christiano and Martin Eichenbaum (1995) `Inside Money, Outside Money, and Short-term Interest Rates', Journal of Money, Credit, and Banking, 27, November, 1354±1401. ÐÐÐÐ (1998) `Expectation Traps and Discretion', Journal of Economic Theory, 81, 2, 462±92. Christiano, L. J. (1991) `Modeling the Liquidity Effect of a Money Shock', Federal Reserve Bank of Minneapolis' Quarterly Review, 15, 3±34. Christiano, L. J., M. Eichenbaum and C. Evans, (1998a) `Monetary Policy Shocks: What Have We Learned and to What End?', NBER Working Paper 6400; also in J. B. Taylor and M. Woodford (eds), Handbook of Macroeconomics, Amsterdam, Elsevier. ÐÐÐÐ (1998b) `Modeling Money', NBER Working Paper, number 6371. Christiano, L. J. and C. Gust (1999), `Taylor Rules in a Limited Participation Model', NBER Working Paper 7017. ÐÐÐÐ (2000) `The Expectations Trap Hypothesis', Federal Reserve Bank of Chicago Economic Perspectives, 25, 2, 21±39. Christiano, L. J. and R. J. Vigfusson, (1999), `Maximum Likelihood in the Frequency Domain: A Time to Build Example', NBER Technical Working Paper 7027. Clarida, R., J. Gali and M. Gertler (1998) `Monetary Policy Rules and Macroeconomic Stability: Evidence and Some Theory', New York University, manuscript. Cooley, T. F. and G. D. Hansen (1989) `The In¯ation Tax in a Real Business Cycle Model', American Economic Review, 79, 733±48. ÐÐÐÐ (1997) `Unanticipated Money Growth and the Business Cycle Reconsidered', Journal of Money, Credit and Banking, 29, 624±48. Diebold, F. X., L. E. Ohanian and J. Berkowitz (1998) `Dynamic Equilibrium Economies: A Framework for Comparing Models and Data', Review of Economic Studies, 65, 433±51. Eichenbaum, M. J. Fisher and W. Edelberg (1999) `Understanding the Effects of a Shock to Government Purchases', Review of Economic Dynamics, 2, 166±206. Fuerst, T. (1992) `Liquidity, Loanable Funds, and Real Activity', Journal of Monetary Economics, 29, 3±24. Kerr, W. and R. King (1996) `Limits on Interest Rate Rules in the IS±LM Model', Federal Reserve Bank of Richmond Economic Quarterly, 82, 47±75. King, R. G. and M. W. Watson (1996) `Money, Prices, Interest Rates and the Business Cycle', Review of Economics and Statistics, 78, 35±53. Ramey, V. and M. Shapiro, (1998) `Costly Capital Reallocation and the Effects of Government Spending', Carnegie-Rochester Conference Series on Public Policy, 48(1), 45±94. Rotemberg, J., and M. Woodford (1992) `Oligopolistic Pricing and the Effects of Aggregate Demand on Economic Activity', Journal of Political Economy, 100, 1153±1297. Watson, M. (1993) `Measures of Fit for Calibrated Models', Journal of Political Economy, 101, 1011±41.
Discussion
Harald Uhlig
1 Introduction How should monetary policy be conducted? Academics and central bankers often seem to differ in their opinions. Academics typically advocate openness (e.g. the publishing of minutes), whereas central bankers often emphasise the need for secrecy. Academics want central banks to be predictable, whereas central bankers like to keep the markets on their toes. Central bankers prefer what Alan Blinder (1989) called `enlightened discretion', whereas academics often advocate the use of rules. The word `rule' here is to be understood broadly, and simply means that ®nancial markets can ®gure out what and how much the central bank is going to do when and under which circumstances. Put differently, a rule is a systematic framework for predictability. The concept thus encompasses, for example, the in¯ation-targeting approach in the form advocated by Lars Svensson (Chapter 2 in this volume). A rule also allows central bankers to precommit and therefore not to yield to the temptation of doing something with short-run gains and long-run pains (e.g. lower unemployment now at the price of higher in¯ation later): a point due to the famous paper by Kydland and Prescott (1977) and apparently now deeply ingrained in the hearts, minds and actions of practical central bankers, but nonetheless worth stating again from time to time. The continued exchange of academics and central bankers is necessary. Academics need to understand better why central bankers dislike some of their suggestions. For example, are there good political or economic reasons for secrecy? Is the analysis of secrecy provided by academics so far (see Cukierman and Meltzer, 1986, for example) already complete? Conversely, central bankers need to understand academics better, too, if for no other reason than to understand more deeply what it is they are doing and for conducting even wiser, more focused and further forward-looking policy as a result. Yes, there are good reasons for following rules above and beyond their use as a commitment device. Often, `simple' rules such as money growth rules or the Taylor rule have been advocated, be it for enhancing the communication with the broader 51
52 Harald Uhlig
public, for framing the debate or for providing a ®rst, `rough' benchmark of comparison. Opponents like to use them to caricature the academic point of view. But these rules are really meant as a starting point for the analysis and its debate. Chapter 1 by Ben McCallum is a insightful and pathbreaking contribution to that end. The chapter provides an outline of a methodology for the analysis of monetary transmission mechanisms. Together with a few other leading researchers, McCallum aims at applying modern advances in macroeconomic theory to practical issues of monetary policy. One can brie¯y summarise that methodology as follows. First, write down one or better several models of the economy with solid microfoundations. Second, show with a careful comparison with the data, that the model indeed captures the observed behaviour of the economy. Third, experiment with suggestions for policy changes, using the models at hand. Fourth, provide a detailed, insightful and understandable account of the effects of policy and their changes to the policy maker. Ultimately, the aim must be to provide the policy maker with a deeper intuition about the cause and the effects of monetary policy and their changes. Rather than discuss the methodology in the abstract, McCallum provides a speci®c application, studying the Taylor rule in the context of a model framework developed by him and Nelson in a number of previous contributions. I have little doubt that McCallum's pioneering and ambitious contribution points us in the right direction. I liked the chapter a lot. If there are any disagreements, then they are with the details of the arguments and with the quality of the theoretical advances achieved so far. As a discussant, my role is to emphasise these points of departure rather than to restate the comparatively larger parts where author and discussant agree. In summary: the approach is a complement not a substitute to other approaches, in particular the VAR-based methodology, and the theory needs further development.
2
VARs versus theory
A simple view of monetary policy is to think of setting interest rates Rt as a (systematic) reaction to circumstances plus an unpredictable component, Rt f
past and available present data t Vector-autoregressions (VARs) identify the monetary policy shock t in the data and study its effects. This is useful, as they inform us about the effects of policy surprises on the economy and thus about the transmission mechanism in general. That literature has made great advances and provided deep insights in recent years. For example, there is now large agreement that these shocks are far from being a major source of business cycle ¯uctuations. Obviously, they shouldn't be if monetary policy is indeed successful in being predictable.
Discussion 53
Theories with solid microfoundations, on the other hand, are more suitable to contemplate changes in the systematic component f
. One may wonder, whether the systematic component has any effect at all other than in¯uencing the rate of in¯ation. One may also wonder whether there truly are changes in the systematic component: perhaps, the continuous trial-and-error evolution of the way monetary policy responds to current circumstances should more appropriately be understood to be part of the shock component t . But be this as it may: one certainly has to conclude that VARs and economic theory are complements, not competitors. Different problems require different tools. Good, advanced theories of money are still remarkably hard to come by. There is a simple reason: this is a very tough objective! Further advances here are not only desirable, they are crucial.
3 The Model The results in the chapter are based on an analysis of the model in McCallum and Nelson (1998). Despite the contrary appearance in chapter 1, the model is constructed according to the current standards in modern macroeconomic theory ± i.e. is of the stochastic, dynamic, quantitative, rational expectations applied dynamic general equilibrium variety. I, for one, ®nd it unfortunate, that the deep beauty and insight of these models apparently still has to be `sold' using the now outdated and cruder IS±LM language, but perhaps this is still necessary for another decade or two to reach the broadest audience possible. McCallum has rightly shown and emphasised that the IS±LM logic is not in con¯ict with these models, and, indeed, it isn't. The inherent logic provides an improvement on the IS±LM view, however. To take a perhaps strained analogy: there also is no con¯ict between the view that most animals are near-perfect at the tasks they do as if created by divine intervention for that purpose, and the view that this near-perfection is the result of genetic, Darwinian evolution: the latter view simply is more helpful for understanding what we see and for predicting what will happen if the environment changes. Likewise, a well built model in the modern tradition is going to be reliable and more helpful for analysing policy changes than the outdated and inferior IS± LM framework. The model by McCallum has a number of special features. The economy is modelled as open. Investment is exogenous. Labour supply and thus output gaps are demand-driven. Firms may ®nd themselves producing below capacity. There is habit formation in consumption.
4 Exogenous investment? To assume investment to be exogenous for the purpose of studying the effects of monetary policy is, of course, somewhat odd. Investment (including durable consumption) is the most volatile component of output and arguably
54 Harald Uhlig
the most interesting one for monetary policy. Whether directly through interest rates or indirectly through the credit channel, monetary policy mainly affects investment. The effect on consumption via a `price channel' is much too sluggish to be of major importance. Of course, having a more explicit role for investment can be done in the context of this model. It makes the model less elegant, but not much more complicated, but it would then have an additional and appealing channel for monetary policy. A key dif®culty here would be to reconcile the observed ¯uctuations in interest rates with the observed ¯uctuations in investment, a point not suf®ciently emphasised by the current real business cycle literature.
5
Theory versus Data
Jeff Fuhrer (1997) writes: It is too easy to write an elegant theoretical model, and too dif®cult to write a model that also replicates key dynamic elements found in the data. Policymakers will rightly be leery of the former, and at least somewhat more comfortable with the latter. It would be hard to disagree with this statement. How well does McCallum's model do? McCallum's Chapter 1 invites the reader to investigate this issue. It provides a number of comparisons to the data, that are similar in spirit to the practice in the real business cycle literature of comparing (®ltered) correlations at leads and lags, although more could be provided: Figure 1.8 (p. 32) offers only a partial assessment. There is also no principal difference between comparing raw correlation patterns versus comparing impulse response functions in (reduced-form) VARs: both VARs and correlation patterns provide summary information about the stochastic properties of the models and the data-and, in fact, one can essentially compute the former from the latter and vice versa. Which representation one ®nds more useful thus seems to be purely a question of taste, and I would have found it helpful to also provide a systematic comparison of the latter. The comparison to the data in Tables 1.3, 1.4 and Figures 1.8±1.12 (pp. 27± 36) shows that the differences to the data are still uncomfortably large. Policy makers should still be leery of the model at hand. That said, one should understand how dif®cult an exercise this is. While we are not at the end of the road yet, the chapter nonetheless provides a big step forward. Much to his credit, McCallum discusses all these issues openly and frankly rather than hiding them for someone else to ®nd. This is an honest, scienti®c approach which should serve as the model for how things should be done. Table 1.4 shows particularly big differences with the data regarding the in¯ation variability, except for model d. Should the reader conclude that one
Discussion 55
should prefer the Fuhrer±Moore framework over P-bar? Comparing Figures 1.9±1.12 shows that the second-order properties of the models do not seem to differ very much. How broad and robust really is the range of theoretical insights gained here? Some of the other ®gures show some puzzling light on McCallums model. Figure 1.1b (p. 20) shows prices to go down rather than up after a taste shock makes consumers want to consume more today. Comparing Figure 1.3a with Figure 1.3b (p. 22) shows that the path of interest rates does not matter for the path of output: output follows the same path, whether interest rates go up or go down. More intuition for why all this is happening, and whether or not this is reasonable, would be desirable. Some potential, additional problems may be present, but can be judged less clearly from the tables and ®gures shown. For example, demand-driven labour movements are known to easily produce counterfactual countercyclical productivity movements, an old problem of Keynesian-style macroeconomic theories. In McCallum's model, there are productivity shocks, too. But still: what is the net effect?
6 Is there a `price puzzle'? The role of habit formation in McCallum's model is an interesting one. The behaviour of the model without habit formation can be glanced from Figure 1.4b: it produces a `price puzzle' in that prices rise after a contractionary monetary policy shock. Introduce habit formation, and it disappears (see Figure 1.4a). While habit formation may be a very reasonable feature of preferences, it is somewhat ironic to see both the empirical VAR literature as well as the theory here struggle with making the `price puzzle' disappear. Should one perhaps contemplate the alternative that indeed prices truly rise rather than fall after a contractionary shock? There is a simple argument for possibly taking that alternative seriously at least for a small open economy. The argument has been put forth by Fabio Ghironi (1999) and runs as follows. Let R; R denote domestic and foreign interest rates. Let e denote the change in exchange rates, with an appreciation of the domestic currency indicated by e > 0. Let ; denote domestic and foreign in¯ation. If purchasing power parity (PPP) holds, then e � If uncovered interest parity (UIP) holds, then R R � e Taken together, one ®nds R � R
56 Harald Uhlig
indicating that there then has to be a `perverse' reaction of in¯ation to monetary policy: holding and R constant, in¯ation will rise one-for-one with a rise in the domestic interest rate R, i.e. in a monetary contraction! Of course, it is well known, that neither PPP nor UIP holds well in the data but still: these are key benchmark concepts in the theory of international trade. And the view that in¯ation slows down with a rise in interest rates is a key pillar of our understanding of monetary policy. The simple argument above shows that at least one of the three has to yield. More broadly, our understanding of the liquidity effect of monetary policy is still more patchy than our understanding of Fisherian relations between real and nominal interest rates.
7
Comparing alternative policy rules
For monetary policy and its effect on the economy, is there a sizeable difference between reacting to current in¯ation Et�1 pt versus reacting to in¯ation forecasts, Et�1 pt1 ? It is here where the chapter makes a direct contribution to an intense and current policy debate of direct practical consequence: the answer sheds important light on the question, whether monetary policy should be targeting in¯ation forecasts or whether it should react to current developments in prices. Table 1.1 seems to provide the answer. Indeed, McCallum writes, that the stabilising effect of monetary policy on the in¯ation rate is greater when Et�1 pt , rather than Et�1 pt1 , is the variable responded to. Obviously, this is true in the sense of tracing out the results of changing the numerical value in the policy rule. But is it true in the relevant economic sense? Figures D1.1 and 3.5 3 Output variability
2.5
E(t–1 ast) as t –(Δ1p) t)Infl( target
2 1.5
) t + 1) E(t–1 t –(Δ1p) t+1 Infl( as target
1 0.5 0 0
5
10
15
Inflation variability
Figure D1.1 Comparison of two rules: the `Taylor menu': in¯ation and output Figure D1.1 shows the `Taylor menu', juxtaposing the variance of in¯ation with the variance in output, using the numbers from the ®rst two rows in Table 1.1 in the chapter. As one can see, the results form a common line.
Discussion 57
Interest rate variability
5 4 3
ast) as E(t–1 t –(Δ1p) t)Infl( target
2
) t + 1) E(t–1 t –(Δ1p) t+1 Infl( as target
1 0
0
5 10 Inflation variability
15
Figure D1.2 Comparison of two rules: in¯ation versus interest rates Figure D1.2 juxtaposes the variance of in¯ation and the variance in interest rates, using the numbers from the ®rst two rows in Table 1.1 in the chapter. As in Figure D1.1, they form a common line.
D1.2 provide the answer, using the numbers provided by McCallum in Table 1.1. Figure D1.1 shows the implied tradeoff between the variability between in¯ation and output in McCallum's models. I would like to call this ®gure the Taylor menu, since focusing on this tradeoff has been popularised by John Taylor. In any case, Figure D1.1 shows that the numbers of the ®rst two rows (I didn't try all three) in Table 1.1 line up beautifully on a line: the only difference between reacting to Et�1 pt rather than Et�1 pt1 lies in the speci®c numerical value for 1 one has to choose in order to pick a particular point on this Taylor menu, but this is a question of technical execution of the policy rather than any particularly interesting difference (and can easily be understood as being due to the endogeneity of expectations when focussing on in¯ation forecasts). Figure D1.2 shows the same result when focusing on the tradeoff between in¯ation variability and interest rate variability. I ®nd this similarity to be a more striking and interesting result than the differences. Perhaps, the entire debate about whether monetary policy should target in¯ation forecasts or whether it should react to current price developments ± or, more broadly, perhaps the entire debate about the particular form of the policy rule to choose ± is largely a vain one. The discipline of examining the economic consequences is the relevant perspective. McCallum's chapter does this: indeed a theoretical framework like his is necessary to sort out `religious' differences from practical consequences. I read his results as indicating, that the practical differences are small. This might be an extremely important and interesting result, if it holds up robustly across
58 Harald Uhlig
many models. Central bank researchers should therefore devote substantial resources to investigating whether it does. Finally, McCallum's framework offers an opportunity which only models with complete micro foundations can offer: since preferences of the consumers are made explicit, one can evaluate the welfare consequences of different policy choices and one can search for the optimal policy rule and the optimal policy parameters within any given class. It is unfortunate that the chapter is silent on this issue, but this will surely become a standard aspect of this literature as it develops further.
8
The future
Central banks need to analyse the effects of policy changes. Insights into the effects of changes to the systematic `feedback' part of policy are particularly important: how should monetary policy react to developments in the economy? Providing substantive answers requires carefully spelled out theories, in which potential policy changes can be examined. These theories should make use of the tools of modern macroeconomics. They should be stochastic, dynamic, quantitative, rational expectations applied general equilibrium models. McCallum is one of the leading pioneers of that new direction, paving the way. These theories are advanced enough that they may soon be a useful guide for policy making. McCallum's chapter is a big step in the right direction. Analyses of this type should become the standard, by which monetary policy is conducted and assessed in the future. And the discussion should become more informed, moving away from the more primitive IS±LM language to the more elegant language of applied general equilibrium analysis. Indeed, this is already happening in the United States. To an outside observer, it is remarkable how far and how fast the various branches of the Federal Reserve System in the United States have moved in building up their expertise for analysing monetary and economic phenomena. By providing suf®cient freedom and resources to actively participate in frontier research on economics and monetary policy, and by encouraging publication in leading journals and presentations of ®ndings at learned meetings, the research departments at the board of the Federal Reserve System as well as at many regional Federal Reserve Banks have provided an environment so attractive that it has enabled them to recruit the best talent available. They now rival the leading economics departments at US universities in quality. There is a friendly, healthy competition between the various branches within the Federal Reserve System, keeping the analysts at each place honest and working hard to provide the best analysis possible, with the largest research department at the centre of the system. The relevance for monetary policy decision making is obvious: the weight and the clout of the members of the Board of Governors of the Federal Reserve System increases directly with the weight and the clout
Discussion 59
of the research department to which they have access, with Alan Greenspan's weight particularly large. For the public, the result has been the best monetary policy that the United states has ever enjoyed. Obviously, one cannot ascribe the entire success and the in¯uence of individual board members to the strength of the research departments. But one shouldn't run the risk of underestimating it either. It is hard to imagine that a similar logic will not hold true for the European System of Central Banks as well. Indeed, it rather seems that the race to the top has already begun. The economists at these central banks will use the best available tools to study the important monetary policy questions at hand. They will therefore be students of McCallum's work for a long time to come. References Blinder, A. S. (1989) Central Banking in Theory and Practice, Cambridge, MA, MIT Press. Cukierman, A. and A. H. Meltzer (1986) `A Theory of Ambiguity, Credibility, and In¯ation under Discretion and Asymmetric Information', Econometrica, 54, 1099± 1128. Fuhrer, J. (1997) `An Optimization-based Econometric Framework for the Evaluation of Monetary Policy ± Comment', NBER Macroeconomics Annual 1997, 346±55. Ghironi, F. (1999) `Alternative Monetary Rules for a Small Open Economy: The Case of Canada', University of Berkeley, Department of Economics, November, draft. Kydland, F. E. and E. C. Prescott (1977) `Rules Rather than Discretion: The Inconsistency of Optimal Plans', Journal of Political Economy, 85, 473±92. McCallum, B. T. and E. Nelson (1998) `Nominal Income Targeting in an Open-economy Optimizing Model', NBER Working Paper, 6675; also in Journal of Monetary Economics, 43.
2
Price Stability as a Target for Monetary Policy: De®ning and Maintaining Price Stability1 Lars E. O. Svensson
1
Introduction
The purpose of this chapter is to provide an up-to-date discussion of monetary policy with `price stability' as the primary objective. The chapter discusses how `price stability' can be de®ned, and how price stability can be maintained in practice. It also discusses some lessons for the Eurosystem. De®ning price stability involves deciding between price-level stability and low (including zero) in¯ation, choosing the appropriate price index and selecting the appropriate level for a quantitative target. It also involves deciding on the role of real variables, like output, in the objectives for monetary policy. Thus, de®ning price stability boils down to de®ning the monetary policy loss function. Maintaining price stability involves meeting the objectives of price stability ± that is, minimising the monetary policy loss function. I consider three main alternatives, namely commitment to a simple instrument rule (for instance, a commitment to following a Taylor rule), forecast targeting (for instance, in¯ation forecast targeting) and intermediate targeting (for instance, monetary targeting). A sizeable part of the literature on monetary policy seems to focus on the properties of optimal and simple reaction functions for monetary policy, like the performance of Taylor-type reaction functions (that is, linear reaction functions responding to deviations of in¯ation from an in¯ation target and to the output gap). This literature provides considerable insights into the characteristics of optimal monetary policy and the properties of different reaction functions, and thereby provides considerable guidance and benchmarks for actual monetary policy, but I argue that a commitment to any of these reaction functions is, for several reasons, neither a good nor a practical way of conducting monetary policy. Such commitment is not a substitute for a systematic operational framework for policy decisions by central banks. Instead, I believe forecast targeting provides such a systematic and operational framework. Indeed, I believe that the current best practice of conducting real world monetary policy can be interpreted as the application of 60
De®ning and Maintaining Price Stability 61
forecast targeting. Thus, most of this chapter is a discussion of forecast targeting. I examine its theoretical background and how, in practice, it can incorporate judgemental adjustments and extra-model information, the role of different indicators (including indicators of `risks to price stability') and, in particular, how it can incorporate complications like non-linearity and model uncertainty. The discussion of forecast targeting builds on Svensson (1999b). The new elements include a more explicit discussion of policy multipliers, judgemental adjustments, the choice between mean, median and mode forecasts, and the role of indicators (the latter builds on Svensson and Woodford, 2000). In particular, I discuss forecast targeting under non-linearities, nonadditive uncertainty and model uncertainty, and the related generalisation of what can be called mean forecast targeting to distribution forecast targeting. Intermediate targeting, in particular monetary targeting, is treated fairly brie¯y, for several reasons. The recent interest in monetary targeting has mainly been motivated by the view that monetary targeting is the reason behind the Bundesbank's outstanding record on in¯ation control and the possibility that the Eurosystem would choose monetary targeting as its monetary policy strategy. However, with regard to whether monetary targeting lies behind Bundesbank's success, ± as discussed, for instance, in Svensson (1999d) a number of studies of the Bundesbank's monetary policy, by both German and non-German scholars, has come to the unanimous conclusion that, in the frequent con¯icts between stabilising in¯ation around the in¯ation target and stabilising money growth around the money growth target, the Bundesbank has consistently given priority to the in¯ation target and disregarded the monetary target.2 Thus, the Bundesbank has actually been an in¯ation targeter in deeds and a monetary targeter in words only. Furthermore, although the Eurosystem has adopted a money-growth indicator, it has strongly rejected monetary targeting as a suitable strategy, on the grounds that the relation between prices and money may not be suf®ciently stable and that the monetary aggregates with the best stability properties may not be suf®ciently controllable (see Issing, 1998). (Furthermore, an extensive and convincing discussion some twenty-®ve years ago concluded that intermediate targeting was generally inferior ± see, for instance, Kareken, Muench and Wallace 1973, Friedman, 1975 and Bryant, 1980.) The discussion of the lessons for the Eurosystem builds on Svensson (1999d) The new elements includes further scrutiny of Eurosystem arguments in favour of its money growth indicator and against in¯ation-forecast targeting. In discussing monetary policy strategy, as in Svensson (1999d), I ®nd it helpful to distinguish two of its elements, namely the framework for policy decisions and communication. By the `framework for policy decisions', I mean the monetary policy procedures inside the central bank, which, from observations of various indicators, eventually result in decisions about the central bank's instruments ± in short, the principles for setting the
62 Lars E. O. Svensson
instruments (which, in the Eurosystem's case, will be a two-week repurchase rate). By `communication', I mean the central bank's way of communicating with outsiders (the general public, the ®nancial market, governments, policymakers and policy-making institutions which, in the Eurosystem's case, includes EU institutions and national governments and parliaments). Communication is part of the implementation of monetary policy, in that it affects the ef®ciency of monetary policy by, for instance, in¯uencing expectations, predictability and credibility. Communication also in¯uences how transparent policy is, which is crucial for central bank incentives and for accountability and arguably also for the political legitimacy of the policy. In terms of the distinction between decision framework and communication, this chapter almost exclusively deals with the decision framework. I have extensively discussed communication and transparency in in¯ation targeting in Svensson (1997a) and (1999b) and the same issue with regard to the Eurosystem in Svensson (1999d). The concluding Section 5 includes some comments on transparency and forecast targeting. A large part of the monetary policy literature uses the concept of `rules' in the narrow sense of a prescribed reaction function for monetary policy. As in previous papers, for instance Svensson (1997a) and (1999b), I ®nd it helpful to use monetary policy rules in a wider sense, namely as `a prescribed guide for monetary policy'. This allows `instrument rules', (prescribed reaction functions), as well as `targeting rules' (prescribed loss functions or prescribed conditions that the target variables, or forecasts of the target variables, shall ful®l). Furthermore (as in Cecchetti, (1997), for instance) `targeting' here refers to loss functions and `target variables' to variables in the loss function. Thus `targeting variable Yt ' means minimising a loss function that is increasing in the deviation between the variable and a target level. In contrast, in some of the literature `targeting variable Yt ' refers to a reaction function where the instrument responds to the same deviation. As discussed in Svensson (1999b section 2.4) and (1999a), these two meanings of `targeting variable Yt ' are not equivalent. `Responding to variable Yt ' seems to be a more appropriate description of the latter situation. Section 2 discusses the de®nition of price stability, Section 3 discusses maintaining price stability, Section 4 discusses lessons for the Eurosystem and Section 5 presents some conclusions.
2
De®ning price stability
Price-level stability versus low in¯ation How to de®ne `price stability'? The most obvious meaning of `price stability' would seem to be a stable price level, `price-level stability'. Nevertheless, in most current discussions and formulations of monetary policy, price stability
De®ning and Maintaining Price Stability 63
instead means a situation with low and stable in¯ation, `low in¯ation' (including zero in¯ation). The former de®nition implies that the price level is stationary (or at least trend-stationary). The latter de®nition implies base drift in the price level, so that the price level will include a unit root and be non(trend-)stationary. Indeed, the price-level variance increases without bound with the forecast horizon. Thus, to refer to low in¯ation as price stability is indeed something of a misnomer. Let me refer to a monetary policy regime as price-level targeting or in¯ation targeting, depending upon whether the goal is a stable price level or a low and stable in¯ation rate. We can represent (strict)3 price-level targeting with an intertemporal loss function Et
1 X
Lt;
2:1
0
to be minimised, where (0 < < 1) is a discount factor and the period loss function is the quadratic loss function Lt
1
pt � pt 2 2
2:2
Here, pt denotes the (log) price level in period t and pt denotes the (log) pricelevel target. The price-level target could be a constant or a (slowly) increasing path, pt pt�1
2:3
where 0 is a constant (low or zero) in¯ation rate.4 Similarly, we can represent (strict) in¯ation targeting with a period loss function given by Lt
1
t � 2 2
2:4
where t pt � pt�1 denotes (the) in¯ation (rate) and denotes a low (or zero) in¯ation target. Following Cecchetti (1997), we can use more compact notation by representing in¯ation targeting in (2.2) by the state-dependent price-level target pt pt�1
2:5
instead of (2.3.), or by representing price-level targeting in (2.4.) by the statedependent in¯ation target t pt � pt�1
2:6
instead of a constant . Hence, (2.5.) illustrates the base drift in in¯ation targeting; the in¯ation target applies from the realised price level pt�1 rather
64 Lars E. O. Svensson
than from the target price level pt�1 . Similarly, (2.6) illustrates that the in¯ation target becomes endogenous and time-varying under price-level targeting. In the real world, there are currently an increasing number of monetary policy regimes with explicit or implicit in¯ation targeting, but there are no regimes with explicit or implicit price-level targeting. Whereas the Gold Standard may be interpreted as implying implicit price-level targeting, so far the only regime in history with explicit price-level targeting occurred in Sweden during the 1930s (see Fisher, 1934, and Berg and Jonung, 1999; this regime was quite successful in avoiding de¯ation). Even if there are no current examples of price-level targeting regimes, pricelevel targeting has been subject to an increasing interest in the monetary policy literature. At the 1984 Jackson Hole Symposium, Hall (1984) argued for price-level targeting. Several recent papers compare in¯ation targeting and price-level targeting, some of which are collected in Bank of Canada (1994). Some papers compare in¯ation and price-level targeting by simulating the effect of postulated reaction functions. Other papers compare the properties of postulated simple stochastic processes for in¯ation and the price level (see Fischer, 1994). A frequent result, which has emerged as the conventional wisdom, is that the choice between price-level targeting and in¯ation targeting involves a tradeoff between low-frequency price-level variability on the one hand and high-frequency in¯ation and output variability on the other. Thus, price-level targeting has the advantage of reduced long-term variability of the price level. This should be bene®cial for long-term nominal contracts and intertemporal decisions, but it would come at the cost of increased short-term variability of in¯ation and output. The intuition is straightforward: in order to stabilise the price level under price-level targeting, higher-than-average in¯ation must be succeeded by lower-than-average in¯ation. This would seem to result in higher in¯ation variability than under in¯ation targeting, since base drift is accepted in the latter case and higherthan-average in¯ation need only be succeeded by average in¯ation. Via nominal rigidities, the higher in¯ation variability would then seem to result in higher output variability.5 However, this intuition may be misleadingly simple. In more realistic models of in¯ation targeting and price-level targeting with more complicated dynamics, the relative variability of in¯ation in the two regimes becomes an open issue. As shown in Svensson (1997b, appendix), this is the case if there is serial correlation in the deviation between the target variable and the target level ± for instance, if the price level displays mean reversion towards the pricelevel target under price-level targeting and in¯ation displays mean reversion towards the in¯ation target under in¯ation targeting. Svensson (1999e) gives an example where the absence of a commitment mechanism and at least moderate persistence in the Phillips curve imply that in¯ation variability becomes lower under price-level targeting than under in¯ation targeting,
De®ning and Maintaining Price Stability 65
without output variability becoming higher.6 For some empirical macromodels (both small and large), reaction functions with responses of the instrument to price level deviations from a price-level target lead to as good or better overall performance (in terms of in¯ation and output variances) than with responses to in¯ation deviation from in¯ation targets.7 I believe these results show that the relative properties of price-level targeting and in¯ation targeting are far from settled. In particular, the potential bene®ts from reduced long-term price-level variability and uncertainty are not yet well understood. Still, I believe that low and stable in¯ation may be a suf®ciently ambitious undertaking for central banks at present. However, once central banks have mastered in¯ation targeting, in perhaps another ®ve or ten years, it may be time to increase the ambitions and consider price-level targeting. By then, research and experience may provide better guidance about which regime is preferable. The rest of the chapter will refer to `low in¯ation', corresponding to ( 2.4), with possible additional terms in the loss function, rather than `price-level stability', corresponding to (2.2). Reluctantly, I will occasionally refer to `low in¯ation' as `price stability', without using quotation marks. Some of the discussion below is applicable to both price-level stability and low in¯ation, however. The loss function Is (2.4) an appropriate loss function for a monetary policy aimed at low in¯ation? As reported below, there seems to be considerable agreement among academics and central bankers that the appropriate loss function is instead of the form Lt
1
t � 2
yt � yt 2 2
2:7
where yt is (log) output, yt is potential output, so that yt � yt is the output gap, and > 0 is the relative weight on output-gap stabilisation. As in Svensson (1999c) and (2000b), the case when 0 and only in¯ation enters the loss function can be called strict in¯ation targeting, whereas the case when > 0 and the output gap (or concern about stability of the real economy in general) enters the loss function can be called ¯exible in¯ation targeting.8 Whereas there may previously have been some controversy about whether in¯ation targeting involves concern about real variability, represented by output-gap variability and corresponding to the second term in (2.7), there is now considerable agreement in the literature that this is indeed the case. In¯ation-targeting central banks are not what King (1997) called an `in¯ation nutter'. For instance, Fischer (1996), King (1996), Taylor (1996) and Svensson (1996) in Federal Reserve Bank of Kansas City (1996), all discuss in¯ation targeting with reference to a loss function of the form (2.7) with > 0. As shown in Svensson (1997a) and Ball (1997), concern about output-gap
66 Lars E. O. Svensson
stability translates into a more gradualist policy. Thus, if in¯ation moves away from the in¯ation target, it is brought back to target more gradually. Equivalently, in¯ation-targeting central banks lengthen their horizon and aim at meeting the in¯ation target further in the future. As further discussed in Svensson (1999c), concerns about output-gap stability, simple forms of model uncertainty and interest rate smoothing all have similar effects under in¯ation targeting ± namely, a more gradualist policy. The Sveriges Riksbank has explicitly expressed very similar views.9 The Chancellor's remit to Bank of the England (HM Treasury, 1997) mentions `undesired volatility of output'.10 The Minutes from Bank of England's Monetary Policy Committee (Bank of England, 1999) are also explicit about stabilising the output gap.11 Several contributions and discussions by central bankers and academics in Lowe (1997) express similar views. Ball (1999) and Svensson (1998b) give examples of a gradualist approach of the Reserve Bank of New Zealand. Indeed, a quote from the ECB (European Central Bank, 1999 p. 47) also gives some support for an interpretation with > 0, as well as some weight on minimizing interest rate variability: a medium-term orientation of monetary policy is important in order to permit a gradualist and measured response [to some threats to price stability]. Such a central bank response will not introduce unnecessary and possibly self-sustaining uncertainty into short-term interest rates or the real economy Thus, it is seems non-controversial that real world in¯ation targeting is actually ¯exible in¯ation targeting, corresponding to > 0 in (2.7). The loss function (2.7) highlights an asymmetry between in¯ation and output under in¯ation targeting. There is both a level goal and a stability goal for in¯ation, and the level goal ± that is, the in¯ation target ± is subject to choice. For output, there is only a stability goal and no level goal. Or, to put it differently, the level goal is not subject to choice; it is given by potential output. Therefore, I believe it is appropriate to label minimising (2.7) as `(¯exible) in¯ation targeting' rather than `in¯ation-and-output-gap targeting', especially since the label is already used for the monetary policy regimes in New Zealand, Canada, United Kingdom, Sweden and Australia. What index and which level? Which price index would be most appropriate? Stabilising the CPI should simplify consumers' economic calculations and decisions. The CPI has the advantage of being easily understood, frequently published, published by authorities separate from central banks and very rarely revised. Interest-related costs cause well known problems with the CPI, though: an interest rate increase to lower in¯ation has a perverse short-term effect in increasing in¯ation. It makes sense to disregard this short-term effect in monetary policy
De®ning and Maintaining Price Stability 67
decisions, but it still presents a pedagogical problem in the central bank's communication with the general public. To avoid this problem, the Bank of England and the Reserve Bank of New Zealand have in¯ation targets de®ned in terms of CPIX (RPIX in Britain), the CPI less interest-related costs.12 The Eurosystem has also de®ned price stability in terms of the HICP (Harmonised Index of Consumer Prices), which excludes interest costs. Furthermore, changes in indirect taxes and subsidies can have considerable short-run effects on the CPI. Different measures of underlying in¯ation ± core in¯ation ± try to eliminate such effects. Eliminating components over which monetary policy has little or no in¯uence serves to avoid misleading impressions of the degree of control. The disadvantage with subtracting too many components from the index used for the in¯ation target is that the index becomes more remote from what matters to consumers and less transparent to the general public. It may also be dif®cult to compute in a well de®ned and transparent way. Opinions generally differ on what components to deduct from the CPI. My own view is that deducting interest-related costs and using CPIX, together with transparent explanation of index movements caused by changes in indirect taxes and subsidies, is an appropriate compromise. What level of the in¯ation target is appropriate? Although zero in¯ation would seem to be a natural focal point, all countries with in¯ation targets have selected positive in¯ation targets. The in¯ation targets (point targets or midpoints of the target range) ranging between 1.5 per cent (per year) in New Zealand, 2 per cent in Canada, Sweden and Finland (before joining EMU), and 2.5 per cent in the United Kingdom and Australia (the Reserve Bank of Australia has an in¯ation target in the 2±3 per cent range for average in¯ation over an unspeci®ed business cycle). The Bundesbank had a 2 per cent in¯ation target for many years (called `unavoidable in¯ation', `price norm', or `medium-term price assumption'). During 1997 and 1998, it was lowered to 1.5±2 per cent (which could perhaps be translated into a point in¯ation target of 1.75 per cent). The EMI ± (European Monetary Institute, 1997) de®ned price stability as 0±2 per cent. The Eurosystem has announced `annual increases in the HICP below 2 per cent' as its de®nition of price stability, which has been interpreted as intervals 0±2 per cent or 1±2 per cent; the Eurosystem used a point in¯ation target of 1.5 per cent in constructing its reference value for money growth. The Eurosystem's de®nition of price stability is further scrutinised on page 88. That the in¯ation target exceeds zero can be motivated by measurement bias, non-negative nominal interest rates and possible downward nominal price and wage rigidities.13 2 per cent is the borderline in Akerlof, Dickens and Perry (1996), who study the effects of downward rigidity of nominal wages. 1 per cent is the borderline in Orphanides and Wieland (1998), who examine the consequences of non-negative nominal interest rates. These studies indicate that in¯ation targets below those borderlines risk reducing average output or increasing average unemployment.14 Altogether, announcing an
68 Lars E. O. Svensson
explicit in¯ation target (a point target or a range) may be more important than whether the target (the mid-point of the range) is 1.5, 2 or 2.5 per cent. A symmetric in¯ation target implies that in¯ation below the target is considered equally bad as in¯ation the same distance above the target (which is the case if in¯ation targeting is represented by a symmetric loss function like (2.7)). This would seem to be a precondition for in¯ation expectations being focused on the in¯ation target. A point target with or without a tolerance interval would, from this point of view, be better than just a range. A range would, in turn, be better than an asymmetric formulation like `below 2 per cent'. These aspects may be particularly important when persistent de¯ation is a possibility, of which recent developments in Japan remind us. A symmetric in¯ation target would seem to be the best defence against persistent de¯ation and against the appearance of de¯ationary expectations. Interestingly, a price-level target may have special advantages relative to an in¯ation target in avoiding persistent de¯ation, since an unanticipated de¯ation which makes the price level fall below the price-level target will, if the price-level target is credible, result in increased in¯ation expectations that will, in themselves, reduce the real interest rate and stabilise the economy.15
3
Maintaining price stability
The basic problem of maintaining price stability is thus to set the monetary policy instrument (or instruments) so as to minimise the intertemporal loss function (2.1) with the period loss function (2.7), subject to current information about the current and future state of the economy and the transmission mechanism. The transmission mechanism is taken to be represented by a linear model in state±space form Xt1 X ut1 A t Bit
2:8 xt1jt xt 0 where Xt
t ; yt ; yt ; ::; 10 (where 0 denotes transpose) is a column vector of nX predetermined variables (also called state variables), xt is a column vector of nx forward-looking variables (also called non-predetermined variables), it is a column vector of ni central bank instruments (also called control variables), ut1 is a column vector of nX exogenous iid shocks with zero means and a constant covariance matrix uu and A and B are matrices of appropriate dimensions. The predetermined variables include in¯ation, output, potential output and other variables. I use the convention that the last element of the vector of predetermined variables is unity. This is a convenient way of allowing non-zero means of the variables; the last column of A is then a function of these means. Although the framework is general enough for handling multiple monetary policy instruments I will, realistically, assume that there is only one
De®ning and Maintaining Price Stability 69
instrument (ni 1) and take that instrument to be a short nominal interest rate (for instance, an overnight interest rate or a one- or two-week repurchase rate). The expression xt1jt denotes Et xt1 , the expectation of xt1 conditional upon all information available in period t, including any information about the state of the economy and the model of the economy. The forward-looking variables include asset prices, like exchange rates and interest rates of longer maturity than the instrument, and other variables partially or fully determined by the expectations of future variables. Thus, at this stage I assume that there are no non-linearities in the transmission mechanism (or that shocks and deviations from a steady state are moderate so a linear approximation is acceptable). Furthermore, I make the assumption that the model is known, that the central bank and the private sector have the same information and that the predetermined and forwardlooking variables in period t are observable in period t. I will discuss generalisations of those assumptions below. A more general representation of the monetary policy loss function is to let Yt denote a column vector of nY target variables, given by Xt Yt C Ci it xt where C and Ci are matrices of appropriate dimension. Let Y denote the column vector of nY target levels, and let the period loss function be Lt
Yt � Y 0 W
Yt � Y where W is a positive-semide®nite weight matrix. The period loss function (2.7.) is a special case of this more general loss function, where the target variables are given by Yt
t ; yt � yt 0 , the target levels by Y
; 0 and the weight matrix W is a diagonal matrix with the diagonal
1=2; =2. Given this representation of the loss function and the transmission mechanism, the problem is now to ®nd the principles for setting the instrument it in each period t. I will consider two such main principles, ®rst what can be called `commitment to an instrument rule' or `interest-rate targeting' and then `forecast targeting' (p. 73). A third principle, `intermediatevariable targeting', especially monetary targeting, is brie¯y considered on p. 86. Commitment to a simple instrument rule: interest rate targeting Make the unrealistic assumption that the central bank can commit, once and for all, in period t 0 to a particular reaction function for all future periods. Furthermore, assume that the model (2.8) is known, that the predetermined and forward-looking variables are observable in each period, and that X0 is given. Under these assumptions, it is possible to ®nd the optimal reaction
70 Lars E. O. Svensson
function under commitment that minimises (2.1) in period 0 (see Backus and Drif®ll, 1986; Currie and Levine, 1993; and Soderlind, 1998). This reaction function will be a linear function of the predetermined variables and the predetermined Lagrange multipliers (shadow prices) of the forward-looking variables, it f Xt 't
2:9
for t 0 and X0 given, where f and ' are row vectors with nX and nx elements (called response coef®cients, or reaction coef®cients), respectively. Furthermore, the multipliers t ful®l t1 M21 Xt M22 t
2:10
for t 0 and 0 0, where M21 and M22 are matrices of appropriate dimension. It follows from (2.9) and (2.10) that the optimal reaction function under commitment can be written as a distributed lag of past predetermined variables, it f Xt '
t X 1
�1 M22 M21 Xt�
2:11
If there are no forward-looking variables, there is no distinction between commitment and discretion. Furthermore, the optimal reaction function is a linear function of the current predetermined variables only, it f Xt
2:12
Even when there are forward-looking variables, many papers consider the optimal reaction function under commitment over the class of reaction functions (2.12) of the current predetermined variables only (mostly without notifying the reader that this is a restriction). The optimal reaction function under commitment is normally a function of all the predetermined variables (and the lagged predetermined variables) and is, in this sense, a rather complex construction. Consider also the class of simple reaction functions, the class of linear reaction functions restricted to being `simple' in the sense of having few arguments (for instance, some of the elements of vector f and all the elements of vector ' are restricted to zero). A typical simple reaction function is the much-discussed Taylor rule, where the instrument responds only to current or lagged in¯ation and the output gap. Let the optimal simple reaction function under commitment be the reaction function in a particular class of simple reaction functions that minimises (2.1) in period 0, given X0 : An optimal reaction function under commitment is likely to be too complex, in the sense of involving speci®c responses to a large number of predetermined variables, to be veri®able. Therefore it is dif®cult to conceive of
De®ning and Maintaining Price Stability 71
a commitment of the central bank to this reaction function. A simple reaction function is easier to verify. Therefore, in principle we can conceive of a commitment to a simple reaction function, a commitment to a simple instrument rule. Such a commitment could also be expressed as a targeting rule, more precisely a commitment to the particular loss function corresponding to `interest rate targeting'. Then, for a particular simple reaction function f , a time-varying interest rate target, it , is de®ned as it f Xt
2:13
Then, instead of the period loss function (2.7), the central bank is committed to the new period loss function Lt
1
it � it 2 2
2:14
Clearly, a trivial ®rst-order condition for minimising (2.14) is given by16 it it
2:15
Thus, (2.15) can either be interpreted as a targeting rule, a ®rst-order condition resulting from the commitment to the particular loss function (2.14) with the interest target (2.13), or it can be interpreted directly as an instrument rule, a prescribed rule for setting the instrument as a function of observed variables. As an example, we can consider a Taylor-type reaction function with smoothing, which corresponds to an interest rate target given by it
1 � p r g
t � gy
yt � yt pit�1 where r is the average real interest rate, g and gy are the long-run response coef®cients and
0 < 1 is a smoothing parameter. Furthermore, we realise that under this commitment to a simple policy rule, the central bank need no longer be forward-looking. It need only be forwardlooking once and for all in period 0, when it decides to which simple reaction function it will commit. After that, it need never be forward-looking; to set the instrument according to the prescribed reaction function, or it simply needs to minimise the period loss function each period with the prescribed interest rate target. Although most of the current and previous discussion of monetary policy rules is in terms of commitment to alternative instrument rules (see, for instance, McCallum, 1997, and the contributions in Bryant, Hooper and Mann, 1993 and Taylor, 1999), I do not believe that a commitment to an instrument rule is either a practical or desirable way of maintaining price stability, for several reasons. First, there are overwhelming practical dif®culties in deciding once and for all which instrument rule to follow. The optimal reaction function will involve
72 Lars E. O. Svensson
speci®c responses to a large number of (current and lagged) information variables and is therefore unlikely to be veri®able. Furthermore, results by Levin, Wieland and Williams (1998) indicate that the optimal reaction function (in their case with the restriction that ' is zero) is quite sensitive to the model. This is problematic, since the model is, in practice, not precisely known. A simple reaction function may be more robust, in the sense of performing reasonably well in different models. This is an idea promoted and examined in several papers by McCallum and recently restated in McCallum (1997). The results of Levin, Wieland and Williams also indicate that a simple reaction function may be quite robust in this sense. On the other hand, as shown by Currie and Levine (1992), the optimal simple reaction function does not only depend on the model and the loss function but also on the stochastic properties of the shocks and the initial state of the economy, X0 , so that the performance of simple rules generally depends on these stochastic properties (certainty-equivalence does not hold for simple reaction functions in linear models with quadratic loss functions, in contrast to the case for the optimal unrestricted reaction function).17 Second, a commitment to an instrument rule does not leave any room for judgemental adjustments and extra-model information. As argued further below, the use of judgemental adjustments and extra-model information is both unavoidable in practice and desirable in principle. Also, there is no room for revision of the instrument rule, when new information results in a revision of the model. By disregarding such information, a commitment to an instrument rule would be inef®cient. Third, although a commitment to a complex instrument rule also seems inconceivable in principle, since it will hardly be veri®able, a commitment to a simple instrument rule is, in principle, feasible ± for instance, by an interest rate targeting regime as above. Still, such a commitment is unheard of in the history of monetary policy, for obvious reasons. It would involve committing the decision making body of the central bank to reacting in a prescribed way to prescribed information. Monetary policy could be delegated to the staff, or even to a computer. It would be completely static and not forward-looking. Such a degradation of the decision making process would naturally be strongly resisted by any central bank and, I believe, arguments about its inef®ciency would easily convince legislators to reject it as well. In practice, there is therefore no commitment mechanism that commits the decision making body to reacting in a prescribed way to prescribed information. In practice, decision making under considerable discretion is unavoidable, and nothing prevents the decision making body from reconsidering their decisions more or less from scratch, without being bound by previous decisions and commitments. As Blinder (1998, p. 49) puts it, `Enlightened discretion is the rule'. Fourth, in the absence of a commitment mechanism, a prescribed simple instrument rule would not be incentive-compatible. There would be frequent incentives to deviate, often for very good reasons, due to new, unforeseen,
De®ning and Maintaining Price Stability 73
information (stock market crash, Asian crisis, Brazil ¯oating) and corresponding sound judgemental adjustment. Although alternative instrument rules can serve as informative guidelines (see, for instance, the contributions in Taylor, 1999 ± or, with regard to the performance of a Taylor rule for the Eurosystem ± Gerlach and Schnabel, 1998, Peersman and Smets, 1998 and Taylor, 1998) and decisions ex post may sometimes be similar to those prescribed by the simple instrument rules, they are not a substitute for a decision making procedure for the central bank. Interest rate targeting for the Eurosystem was indeed rejected by the European Monetary Institute (EMI), the predecessor of the European Central Bank (ECB), in European Monetary Institute (1997, p. 1) (with, arguably, not the most exhaustive argument): [T]he use of an interest rate as an intermediate target is not considered appropriate given dif®culties in identifying the equilibrium real interest rate which would be consistent with price stability. Indeed, instead of having a decision making procedure and being forwardlooking only once and for all, at the time of a commitment to a simple rule, the central bank needs to have a continuous decision making procedure and be continuously forward-looking. To quote Greenspan (1994, p. 244): Implicit in any monetary policy action or inaction, is an expectation of how the future will unfold, that is, a forecast. The belief that some formal set of rules for policy implementation can effectively eliminate that problem is, in my judgement, an illusion. There is no way to avoid making a forecast, explicitly or implicitly.18 Therefore, I now turn to a practical and realistic, and indeed already practised, way of maintaining price stability, namely by way of `forecast targeting'.19 Forecast targeting As a background, recall that, with a quadratic loss function and a linear model, with the assumption of a known model and only additive uncertainty, certainty-equivalence applies. The problem of minimising the loss function can be separated into a deterministic problem involving conditional forecasts, the conditional means of current and future variables and a stochastic problem involving deviations between realised outcomes and conditional forecasts. The solution to the deterministic problem is the same as to the stochastic problem (see Chow, 1975, for models with only predetermined variables and Currie and Levine, 1993, for models with forward-looking variables as well). Thus, the discussion can focus on the deterministic problem involving conditional forecasts.
74 Lars E. O. Svensson
For any variable t , let tjt for 0 denote the expectation Et t given information in a ®xed period t. The information in period t includes the information available about the state of the economy as well as about the model, (2.8).20 Let jt denote the future path tjt ; t1jt ; t2jt ; ::: Consider the set I t of given paths ijt
itjt ; it1jt ; ::: of instrument settings, for which there exist bounded paths jt and yjt � yjt of future in¯ation and output gaps, respectively. For each ijt 2 I t , let jt
ijt denote the corresponding path for variable and y (yjt is taken to be exogenous), and call it the corresponding conditional forecast (conditional on information in period t, ijt and the model (2.8)). Let n o YT jt
ijt ; yjt
ijt � yjt jijt 2 I t denote the set of feasible conditional forecasts of the target variables. Constructing conditional forecasts in a backward-looking model (that is, a model without forward-looking variables) is straightforward. Constructing such forecasts in a forward-looking model raises some speci®c dif®culties, which are explained and resolved in Svensson (1998a, Appendix A).21 Due to the certainty-equivalence, the stochastic optimisation problem of minimising the expected intertemporal loss function (2.1) over future random target variables in (2.7), subject to (2.8), is equivalent to the deterministic problem of minimizing the deterministic intertemporal loss function 1 X
L~ t jt
2:16
0
with the deterministic period loss function 1 2 L~ t jt
tjt � 2
ytjt � yt jt 2
2:17
(where the stochastic t and yt � yt have been replaced by the deterministic tjt and ytjt � yt ) subject to
jt; yjt � yjt 2 Y t
2:18
Thus, the problem of the central bank is to choose the path ijt ; such that the resulting jt and yjt minimise (2.1) with (2.17). The ®rst element of ijt , itjt , is then the appropriate instrument setting for period t; it . If there is no new relevant information in period t 1, the instrument setting in period t will be the second element in ijt . If there is new relevant information, that information is used for solving the problem again in period t 1.22 This procedure thus involves making conditional forecasts of in¯ation, output and the output gap for alternative interest rate paths, using all relevant information about the current and future state of the economy and the transmission mechanism. It involves making consistent assumptions about
De®ning and Maintaining Price Stability 75
exogenous and endogenous variables (for instance, that exchange rates and interest rates are consistent with appropriate parity conditions). As discussed further below, it also allows judgemental adjustments and extra-model information (p. 77), as well as partially observable states of the economy (p. 79). As discussed on p. 82, forecast targeting can even be adapted to take non-linearities and model uncertainty into account, which both result in nonadditive uncertainty. The procedure requires estimates of policy multipliers, the effects on the conditional forecasts of changes in the instrument. The policy multipliers are easily calculated in a simpli®ed model with only predetermined variables Xt1 AXt Bit ut1
2:19
(see Svensson 1998a, Appendix A, for a case with forward-looking variables). The conditional forecast for Xtjt then ful®lls Xtjt AXt�1jt Bit�1jt A Xtjt
�1 X
A�1�s Bitsjt
2:20
s0
for 1, so that the policy multipliers dXtjt =ditsjt ; 0 s � 1, are given by dXtjt A�1�s B ditsjt
2:21
Optimality criterion What is the criterion for having found an optimal interest rate path and corresponding conditional forecasts of in¯ation and the output gap? One criterion can be formulated as follows. Suppose the central bank staff have constructed a potential optimal combination of an interest-rate path and such conditional forecasts. Consider a change dijt
ditjt ; dit1jt ; ::: in the interestrate path ijt . This will result in changes dXjt
0; dXt1jt ; dXt2jt ; ::: in the predetermined variables, given by dXtjt
�1 X dXtjt s0
dits
ditsjt
Let djt and dyjt denote the corresponding changes in the in¯ation and output forecasts (the output-gap forecast yjt is taken to be exogenous). A necessary condition for optimality is then that the corresponding change in the intertemporal loss function is non-negative ± that is d
1 X
Ltjt
0
1 X
tjt � dtjt 0
ytjt � ytjt dytjt 0
2:22
76 Lars E. O. Svensson
Given jt , yjt , yjt d;jt and dyjt ; as well as , and ; this expression is easily checked, and a relatively easy way of making pairwise comparisons of alternative interest rate paths and conditional forecasts. For instance, a delay in an interest rate increase can be compared with an immediate increase, a small increase now can be compared with a larger increase two quarters later, etc. Within the simple model in Svensson (1997a, 1999c), I have shown that the ®rst-order conditions for an optimum can be expressed as particularly simple targeting rules (in the form of equations for the conditional forecasts of the target variables). For instance, the in¯ation gap t1jt � and the output gap ytjt � ytjt should be of the opposite signs and related according to t1jt � �
y c
; 1
ytjt � ytjt 1 � c
where the coef®cient y is the sensitivity of the change in in¯ation to the output gap and the coef®cient c
is an increasing function of that ful®ls 0 c
< 1, c
0 0 and c
1 1. Alternatively, the targeting rule can be expressed as the in¯ation forecast approaching the in¯ation target at a constant rate t1jt � c
tjt � ; 1 In practice, the decision making body may get far by just visually examining alternative in¯ation and output-gap forecasts and then choosing the one that is the best compromise between hitting the in¯ation target at an appropriate horizon and avoiding output-gap stability. If done in a consistent way, this will be equivalent to minimising the intertemporal loss function. Compared to many other intertemporal decision problems that households, ®rms and investors solve one way or another (usually without the assistance of a substantial staff of economics PhDs), this particular decision problem does not seem to be overly complicated or dif®cult. Instrument assumptions In the above discussion, the problem is to ®nd the appropriate interest rate path ijt , which requires constructing conditional forecasts for exogenous interest rate paths. Forecasts for unchanged interest rates, where the interest rate path ful®ls itjt it�1; 0 can be used as indicating `risks to price stability'. They are used by Bank of England and Sveriges Riksbank to motivate changes in the interest rate and their direction (see p. 81f.) Conditional forecasts can also be constructed for given reaction functions, in which case the interest rate path is endogenous and ful®ls
De®ning and Maintaining Price Stability 77
itjt f Xtjt for a given f .23 Some central banks, notably the Bank of Canada and the Reserve Bank of New Zealand, construct forecasts conditional on reaction functions involving what Rudebusch and Svensson (1999a) call `responding to model-consistent forecasts', and what Batini and Haldane (1999) call `forecastbased' reaction functions. This implies an interest rate path ful®lling itjt f Xtjt gXtT jt; where T > 0 is the forecast horizon (typically some six±eight quarters), and sometimes f 0: Among reaction functions involving responses to forecasts, it would seem more natural that the forecast responded to is one for unchanged interest rates. Indeed, under strict in¯ation targeting, the optimal instrument adjustment is proportional to the deviation of the conditional forecast for unchanged interest rate from the in¯ation target, as demonstrated in Svensson (1999d).24 Judgemental adjustments A major advantage of forecast targeting relative to commitment to an instrument rule is that it allows a systematic and disciplined way of incorporating judgemental adjustment and extra-model information, by `®ltering information through the forecasts'. Given that every model of the transmission mechanism is an abstraction and a simpli®cation, and given that there is considerable uncertainty about the details of the transmission mechanism, monetary policy will never, it seems, be able to rely on models and the information entering models alone. There will always be an important role for additional extra-model information and judgemental adjustments of the instrument. Never using such information and judgement would neither be ef®cient nor incentive-compatible for the decision making body of the central bank. At the same time, such information and informal adjustment opens up monetary policy to arbitrariness and potential abuse. Forecast targeting allows some system and discipline in the use of extra-model information and judgemental adjustments. Judgemental adjustments can be made in several ways in the framework outlined here. One is in the form of adjustments of the coef®cients of the model. This means that the coef®cients of matrices A and B become timevarying, Atjt and Btjt , which is easily incorporated when constructing the forecasts (as long as the time-variation is deterministic; see p. 84 on nonadditive uncertainty). Another kind of judgemental adjustments consists of simple additive shifts in the forecasts. For the model with only predetermined variables, this means that the model in period t can be represented as Xt1jt AXtjt Bitjt jtjt
78 Lars E. O. Svensson
where the column vector jtjt corresponds to additive judgemental adjustments to Xt in period t. Suppose, as above, that the last element in Xt is unity, in order handle non-zero means. Add the additive judgemental adjustment jtjt to the last column of the matrix A to form the matrix Atjt . Then, the system can be written Xt1jt Atjt Xtjt Bitjt formally as a model with time-dependent coef®cients.25 Only if a particular piece of information can be convincingly shown to affect the conditional forecasts at relevant horizons does it warrant a change in the instrument. It is not correct to adjust the instrument without `®ltering the information through the forecasts'. Thus, targeting rules and forecast targeting bring some system and discipline to the use of judgments and extra-model information, and provide some protection against the arbitrariness in use and temptations of abuse that might easily arise. This is, I believe, one aspect of what Bernanke and Mishkin (1997) call `constrained discretion' (although they are not explicit about the role of forecast targeting).26 Mean, median or mode forecasts? Is it the mean, the median or the mode forecast that is relevant in forecast targeting? Under a linear model with additive uncertainty and a quadratic loss function, the previous discussion have demonstrated that certainty-equivalence holds and that it is the mean forecast that is relevant. When would median or mode forecasts be relevant? Vickers (1998) and Wallis (1999) have provided systematic discussions of this issue. Under the maintained assumption of a linear model with additive uncertainty and objectives corresponding to minimum expected loss, the nature of the loss function determines which forecast is relevant. Thus, with a quadratic loss function, it is the mean forecast. Under a linear loss function (that is, linear in the deviation from target, V-shaped), it is the median forecast. Under an `all or nothing' loss function with extreme favourable weight on the target level, it is the mode forecast. Finally, under a loss function assigning a constant loss outside a tolerance interval and a zero loss inside, the forecast should maximise the probability of being inside the interval, which implies that the upper and lower bounds should have the same probability density.27 The mean, median and mode alternatives differ only when the probability distribution is asymmetric. When they differ, unless certainty-equivalence is explicitly assumed not to hold, it seems that the mean forecast should still have prominence and never be excluded, also when the mode and/or the median is reported. Still, both the Bank of England and the Sveriges Riksbank report the mode forecast as their central forecast, in spite of the mode being associated with a relatively bizarre loss function, the all-or-nothing one.28 The Bank of England and the Sveriges Riksbank, however, aim to illustrate not only a point forecast but the probability distribution of the forecasted
De®ning and Maintaining Price Stability 79
variable, so as to indicate the uncertainty of the forecast, as well as the `balance of risks', ± that is, whether the distribution is symmetric or not. Thus, the Bank of England produces its famous fan chart, which can be interpreted as showing iso-level contours for the density function of the probability distribution, where each shade of colour (red for in¯ation, green for the output forecast) encloses a given fraction of the probability mass (see Britton, Fisher and Whitley, 1998). Thus, the fan chart displays the density function in the same way as a contour map illustrates a hill (with the difference that maps usually have contours with equal vertical distance whereas the fan chart has equal increments of the enclosed probability mass). Furthermore, the mode will be at the centre of the narrow central band, the deepest shade of colour which denotes the most likely outcomes, the highest portion of the probability density. As noted by Wallis (1999), this means that the Bank of England uses non-central prediction (or con®dence) intervals, which are different from standard central prediction intervals in that they do not have the same probability mass above and below the intervals (when the distribution is asymmetric). The Sveriges Riksbank instead reports the mode together with standard central prediction intervals, with the same probability above and below the prediction interval (see Blix and Sellin, 1998). As a consequence, for an asymmetric distribution, the mode will not be at the centre of the most narrow con®dence interval (instead, the centre of the most narrow con®dence interval is the median). As long as these alternative graphs are clearly understood as different ways of illustrating the whole distribution, and policy is based on the whole distribution, the ways of illustration need not have any effect on the policy. Still, with the ways of illustration that these central banks have chosen, I believe that most observers are led to focus on the mode, when the mean arguably in many cases is a more appropriate focus. Therefore I believe that it might be better to plot the mean rather than the mode, together with the different con®dence intervals. I believe this might be better also under distribution forecast targeting, to be discussed on p. 82, when the whole distribution matters.29 The role of indicators So far, the maintained assumption in the discussion has been that the central bank can directly observe the state of the economy, more precisely, observe the predetermined and forward-looking variables, Xt and xt in period t. In order to discuss the role of indicators, I now assume that the predetermined and forward-looking variables are not necessarily observable. Consequently, introduce a vector of nZ observable variables, indicators, Zt . The indicators depend on the predetermined and forward-looking variables and the instrument according to Xt Di it vt Zt D xt
80 Lars E. O. Svensson
where D and Di are matrices of appropriate dimension and vt is a vector of nZ iid shocks with zero means and constant covariance matrix vv : These shocks may be interpreted as measurement errors. Assume that the central bank's information in period t is represented by the information set It Zt�; 0; A; B; C; Ci ; D; Di ; uu; vv g That is, in period t, the central bank knows the current and past indicators, in addition to the model and the stochastic properties of the shocks. This formulation allows some variables to be directly observable, some to be observed with measurement error and some to be completely unobservable. The discussion in previous sections, the full information case, corresponds to the special case Zt
X0t ; x0t ; i0t 0 . For simplicity, assume that there are no forward-looking variables, so that the model is (2.19), and let the indicators depend on the predetermined variables according to Zt DXt vt
2:23 This setup is examined in more detail in Chow (1975), Tinsley (1975) and LeRoy and Waud (1977).30 The case of partial information with forwardlooking variables is examined in Pearlman, Currie and Levine (1986), Pearlman (1992), Aoki (1998) and Svensson and Woodford (2000). The present discussion follows Svensson and Woodford (2000) (although without forward-looking variables). Under these assumptions, certainty-equivalence continues to hold. In period t, the central bank needs to form the estimate Xtjt of the predetermined variables in order to construct its conditional forecasts and set its instrument. The optimal estimate is given by a Kalman ®lter, with the updating equation Xtjt Xtjt�1 K
Zt � Ztjt�1
2:24
where the matrix K is the Kalman gain matrix.31 The elements of the gain matrix give the optimal weights on the indicators in estimating the predetermined variables. Using (2.23), the updating equation can be written Xtjt
I � KD Xtjt�1 KZt
2:25 so the matrices K and I � KD give the weights on the indicators Zt and the prior information Xtjt�1 . If D I and vv 0, Zt coincides with the predetermined variables; then K I and all weight is on the indicators and none on the prior information. Generally, a row in K gives the optimal weights on the indicators in estimating the corresponding predetermined variable. A column in K gives the weights given to the corresponding indicator in estimating the different predetermined variables. Assume that a particular indicator, Zjt , is equal to one of the predetermined variables, Xkt , plus a measurement error, vjt ;
De®ning and Maintaining Price Stability 81
Zjt Xkt vjt If the variance of the measurement error approaches zero, all the elements in row k of K approaches zero except element j, which approaches unity. Thus, all the weight in estimating Xkt falls on Zjt . If the variance of the measurement error goes to in®nity, Zjt becomes a useless indicator. The elements in column j of K then all become zero. The indicator Zjt gets zero weight in estimating the predetermined variables. Assume, for simplicity, that forecast targeting has resulted in a reaction function f in the past. That is, it� f Xt�jt� for 1 t. Then, by (2.9), Xj�1
A Bf X�1j�1 for 1 t. Using this in (2.25), we can write the updating equation as a distributed lag of past indicators, Xtjt
1 � KD
A Bf Xt�1jt�1 KZt
1 � KD
A Bf t X0j0 t�1 X
1 � KD
A Bf KZt� 0
This gives the weight on indicators Zt� as
1 � KD
A Bf K for 0. These equations illustrate the gradual updating of the estimate of the predetermined variables. We can summarise the effects of the indicators in period t on the forecasts Xtjt in terms of `indictor multipliers,' dXtjt =dZt , given by dXtjt dXtjt dXtjt A K dZt dXtjt dZt Thus, the indicator multiplier is the product of the effect of the estimate of the current state of the forecast, dXtjt =dXtjt , which by (2.20) is given by A , and the effect of the indicators on the estimate of the current state, with by (2.24) is given by the gain matrix K. It follows that an indicator will affect the instrument setting via affecting the current state of the economy, then the forecasts, and ®nally the instrument. Schematically, Zt ! Xtjt ! Xtjt ! it In particular, the weights on any given indicator depend exclusively on its power in predicting future in¯ation and output gap. Monetary aggregates, for instance, have no special role beyond that, and any weight on monetary aggregates will exclusively depend on its predictive performance. An indicator of `risks to price stability' An indicator of what the Eurosystem calls `risks to price stability' ± that is, in¯ationary or de¯ationary pressure, or risks of overshooting or undershooting
82 Lars E. O. Svensson
the in¯ation target, ± should be useful when discussing monetary policy that aims to maintain price stability. What requirements should such an indicator ful®l? It would seem that, ®rst, it should signal in which direction and to what extent the in¯ation target will be missed if policy is not adjusted. Second, it should signal in which direction and to what extent the instrument should be adjusted. Finally, it should be intuitive and easy to understand, so that it can be used to communicate with the public and explain why an instrument change is warranted or not. The Eurosystem has put forward a money growth indicator, namely, the deviation between current M3 growth and a reference value as an indicator of risks to price stability - indeed, the ®rst of the `two pillars' of its monetary strategy. As discussed in Svensson (1999d), such a money growth indicator seems quite unsuitable for this purpose, since it is largely just a noisy indicator of the deviation of current in¯ation from the in¯ation target (which deviation can be more easily observed directly).32 Instead, the obvious candidate is a conditional in¯ation forecast, conditional upon unchanged monetary policy in the form of an unchanged interest rate. That is, it is constructed for the interest rate path that ful®ls itjt it�1 , 0. Constructing this in¯ation forecast is straightforward in a model without predetermined variables, as is apparent from (2.20). It is somewhat more complicated in a model with forward-looking variables, as shown in Svensson (1998a Appendix A).33 This indicator signals whether and in which direction the in¯ation target is likely to be missed, if policy is not adjusted, and it thereby also signals in which direction the instrument needs to be adjusted. The Bank of England and the Sveriges Riksbank, in their quarterly In¯ation Reports, use conditional in¯ation forecasts for unchanged interest rates as their main vehicle for motivating why the instruments need to be adjusted or not. Complications to mean forecast targeting and generalisation to distribution forecast targeting Non-linearities and non-additive uncertainty in the model are two complications to forecast targeting. These complications both imply that the certaintyequivalence underlying conditional mean forecasts no longer holds. These two complications are discussed on p. 82 and p. 84.34 A solution to the complications is to move from conditional mean forecast targeting to conditional distribution forecast targeting, which is discussed on p. 85. Non-linearities So far, the maintained assumption has been that the model is linear. Sources of non-linearity that have been discussed in the literature include non-linear Phillips curves (see Debelle and Laxton 1997; Gordon, 1997; Isard and Laxton, 1996), non-negativity of nominal interest rates and downward nominal rigidity of prices and/or wages (see references on pp. 66±8). Suppose now that the model is non-linear. Assume that the model remains known, that there are no
De®ning and Maintaining Price Stability 83
forward-looking variables, that the predetermined variables are observable, and that the model can be written as Xt1 M
Xt ; it ; ut1
2:26
where M
is a non-linear function. This has two consequences. First, the conditional mean forecasts, Xtjt , are now non-linear functions of the interest rate path, ijt , the current state of the economy, Xt , and the covariances of the shocks, uu . Second, the policy multipliers, dXt =ditsjt ; will be stochastic, and the `forecast' policy multipliers, dXtjt =ditsjt ; will be endogenous and not constant. Then, certainty-equivalence no longer applies, and using the period loss function (2.17) is no longer equivalent to using (2.7). The reason why certainty-equivalence no longer holds can be demonstrated with reference to the optimality criterion (2.22). Consider the change in the intertemporal loss function (2.1.) with (2.7.) of a change dijt for the optimal instrument path, which implies the optimality criterion 0 dEt
1 X
Lt
0
d
1 X
1 Et
t � 2
yt � yt 2 2 0
1
X Et
t � dt Et
yt � yt dyt 0
1 n o X dytjt
tjt � dtjt
ytjt � ytjt 0
1 X Covt t; dt Covt yt; � yt; dyt
2:27
0
where Covt ; denotes the covariance conditional on information available in period t. With a linear model, the changes dt and dyt caused by the change dijt are deterministic and independent of t and yt , so that the covariance terms in (2.27) vanish. The change dXt in the random variable Xt is identical to the change dXtjt in the conditional forecast Xtjt . Then the optimality criterion (2.22) applies, and it is suf®cient to think in terms of conditional forecasts. With a nonlinear model, the changes dt and dyt are stochastic and depend on t and yt and therefore the covariance terms in (2.27) are not necessarily zero. In order to see this, note that with the linear model (2.19), the change dXt from a change ditsjt
0 s < � 1) is given by dXt
dXt dXtjt ditsjt ditsjt A�1�s B ditsjt dits ditsjt
84 Lars E. O. Svensson
where we use that dXt =dits dXtjt =ditsjt A�1�s B, so dXt is deterministic and independent of Xt (since the policy multipliers are constant). With the non-linear model 2.26, the same change is ! r�s�1 Y @M
Xtsr; itsr; utsr1 @M
Xts; its; uts1 dXt ditsjt @X @i r1 so that dXt is stochastic and generally correlated with Xt (since the policy multipliers dXt =ditsjt are now endogenous and not constant).35 With a non-linear model, the optimal reaction function is nonlinear, it f
Xt and it will generally depend on the covariances uu . The covariance terms in (2.26) also imply that the optimal policy may imply a bias, in the sense that it is optimal to, on average, either overshoot or undershoot the in¯ation target (also, the optimal average output gap may not be zero). Doing forecast targeting under a non-linear model and selecting the instrument such that (2.17) is minimised would still imply a non-linear reaction function, since the conditional forecasts would be non-linear functions of the instrument path. It would imply disregarding the optimal bias, though. How costly forecast targeting would be relative to the optimal policy would, of course, depend on the degree of non-linearity in the transmission mechanism, which is an empirical question. My reading of the literature is that there is considerable controversy about the extent and the relevance of any nonlinearity (see for instance Gordon, 1997, and Isard and Laxton, 1996). Non-additive uncertainty So far, only additive uncertainty has been considered, appearing as the additive shock ut1 in (2.8). Uncertainty about parameters in the model ± that is, uncertainty in the coef®cients of the matrices A and B in (2.8) ± results in multiplicative uncertainty, an example of non-additive uncertainty. Multiplicative uncertainty has consequences similar to non-linearity, in that certainty-equivalence no longer holds, even if the model remains linear and the loss function is quadratic. Assume that the model is linear, that there are no forward-looking variables and that the predetermined variables are observable, but assume now that the model has time-varying stochastic parameters, with known stochastic properties. Then the model can be written Xt1 At1 Xt Bt1 it ut1 ;
2:28
where At and Bt are stochastic processes with known stochastic properties. It follows that Xt can be written
De®ning and Maintaining Price Stability 85
Xt
�1 Y s0
At1s Xt
�1 X
�s�1 Y
s0
r1
! At1sr
Bts1 its
�s Y X s1
! Atsr uts
r1
Then the policy multipliers, dXt dits
�s�1 Y
! At1sr
Bts1
r1
for 0 s � 1, are stochastic. It follows that a change in the interest rate path, dijt , results in stochastic changes dt and dyt in in¯ation and output. Once more, the covariance terms in (2.27) do not vanish, and certaintyequivalence no longer applies. As shown in the classic paper by Brainard (1967) and, for instance, in a more recent application to strict in¯ation targeting in Svensson (1999c), the optimal instrument response to disturbances usually becomes more cautious, and a bias enters such that the optimal policy implies that the in¯ation either overshoots or undershoots the in¯ation target. However, some covariance patterns for the parameters, as well as uncertainty about the in¯ation persistency, may make the optimal instrument response more sensitive to disturbances than in the absence of parameter È derstro uncertainty (see So È m, 1999a). Monetary policy under model uncertainty is currently a very active research area, with a number of recent papers.36 Practically all work is directed towards the understanding of how the optimal reaction function changes with model uncertainty. But how should practical monetary policy incorporate model uncertainty? This is the issue to be discussed next. Generalised forecast targeting: distribution forecast targeting instead of mean forecast targeting One possibility for handling non-linearity and model uncertainty might seem to be to commit, once and for all, to a simple instrument rule, either to an optimal reaction function that minimises the expected intertemporal loss, taking Bayesian priors on the nature of the model uncertainty into account, or a simple rule with reasonable robustness properties across potential models. However, I believe the objections to this solution raised on p. 71±73 apply with even greater force under non-linearity and model uncertainty, since the unavoidability and desirability of judgemental adjustments, extra-model information and updated Bayesian priors are even more apparent. Therefore, I believe that the solution must be found elsewhere, namely in a generalisation of forecast targeting. Non-linearity and model uncertainty both imply that certainty-equivalence does not apply. This, in turn, means that it is not optimal to rely on forecast targeting with conditional mean forecasts ± that is, using conditional mean forecasts as intermediate variables. Thus mean forecast targeting is not
86 Lars E. O. Svensson
optimal. Still, we can consider a kind of generalised forecast targeting, distribution forecast targeting, as a way of handling non-linearity and multiplicative uncertainty. Distribution forecast targeting simply consists of constructing conditional probability distributions of the target variables instead of means only. For any random variable t , let t denote the random path
t ; t1 ; ::: (recall that jt denotes the path of the conditional means
tjt ; t1jt ; :::). Then, for a given interest rate path ijt , the central bank staff should construct the joint conditional density function of the random path of in¯ation and the output gap, denoted 't
t ; yt � y t ; ijt , conditional upon all information available in period t and the interest rate path ijt . Then, this conditional probability distribution is used to evaluate the loss function (2.1) with (2.7). This can either be done numerically, or informally by the decision making body of the bank being presented with the probability distributions for a few alternative interest rate paths and then deciding which path and distribution provides the best compromise. This alternative is actually much more feasible than many readers might think. The Bank of England and the Sveriges Riksbank have already developed methods for constructing con®dence intervals for the forecasts published in their In¯ation Reports (see Blix and Sellin, 1998 and Britton, Fisher and Whitley, 1998). The Bank of England presents fan charts for both in¯ation and output, and the Sveriges Riksbank gives con®dence intervals for its in¯ation forecasts.37 Furthermore, scrutiny of the motivations for interest rate changes (including the minutes for the Bank of England's Monetary Policy Committee and the Riksbank's Executive Board) indicate that both banks occasionally depart from certainty-equivalence and take properties of the whole distribution into account in their decisions, for instance, when the risk is unbalanced and `downside risk' differs from `upside risk'. The result of distribution forecast targeting can be compared with what would result if only conditional mean forecasts were considered. In practice, there may not be a big difference in the resulting instrument setting, in which case there is not much point in letting distribution forecast targeting replace mean forecast targeting. Distribution forecast targeting seems meaningful and worth the effort only if it results in signi®cantly different, and more adequate, instrument settings than mean forecast targeting. This remains to be examined. Intermediate targeting When would intermediate targeting be optimal? Assume, for simplicity, strict in¯ation targeting, (2.4). Consider, for simplicity, the model without forwardlooking variables, (2.19). Decompose the vector of predetermined variables according to Xt
t ; X02t ; X03t 0 , where the ®rst element is in¯ation and the rest of the variables are decomposed into two vectors, X2t and X3t . Suppose that
De®ning and Maintaining Price Stability 87
the two vectors X2t and X3t can be chosen such that the model (2.19) of the transmission mechanism ful®ls 2
3 2 t1 0 4 X2;t1 5 4 A21 X3;t1 A31
A12 A21 A32
32 3 2 3 0 0 A23 54 X2t 5 4 B2 5it ut1 A33 X3t B3
that is, where the A and B matrices are such that A11 0, A13 0 and B1 0. Then in¯ation ful®lls t1 A12 X2t u1;t1;
2:29
and is exclusively determined by variables X2t ; and variables X2t are the only predictors of in¯ation (aside from the zero-mean exogenous shock u1;t1 ). Under these assumptions, the instrument it affects in¯ation exclusively by ®rst affecting X2;t1 and then by X2;t1 affecting t2 . Schematically, we have it ! X2;t1 ! t2 Because of this property of X2t ; its elements can be called intermediate variables. Let X2 ful®ll A12 X2 :
2:30
Substituting (2.29) and (2.30) into (2.4) for t 1 and taking the expectation in period t result in 2 2 Lt1jt 12
t1jt � 2 21u1
X2t � X2 0 W
X2t � X2 12u1 2 where u1 is the variance of u1t and the weight matrix W ful®ls W 12 A012 A12 . Clearly, using the period loss function
L~ t
X2t � X2 0 W
X2t � X2 is equivalent to using (2.4). Now we can call X2t intermediate target variables, X2 intermediate target levels, and minimising (2.1) with L~ t instead of (2.7) we can call intermediate targeting. We thus have a situation where intermediate targeting is as good as strict in¯ation targeting. In particular, assume that in¯ation is exclusively determined by money growth according to t1 mt u1;t1 where mt mt � mt�1 and mt is the log of a monetary aggregate. Then, the instrument exclusively affects in¯ation via ®rst affecting money-growth ± that is, it ! mt1 ! t2
88 Lars E. O. Svensson
Thus, let X2t mt , A12 1, X2 and W 12 ; and strict in¯ation targeting can be replaced by strict money growth targeting with the period loss function 1 L~ t
mt � 2 2 Both kinds of targeting will be equivalent. In the example above, the transmission mechanism is recursive in a special way, such that the target variables (in the above case only in¯ation) are determined only by a set of intermediate variables (the only exception being zero-mean exogenous shocks). Clearly, this is an extremely special case. In the real world, and in reasonable models, the transmission mechanism is too complex for intermediate variables in this sense to exist ± that is, the transmission mechanism is not recursive in the above sense.38 Therefore, intermediate targeting in general, and monetary targeting in particular, is not a good monetary policy strategy. However, there is one exception to the general non-existence of intermediate variables. As discussed in Svensson (1997a, 1999b), one set of intermediate variables always exists, namely conditional forecasts. For any vector Yt of target variables, we can write Yt Ytjt "t where Ytjt is a conditional forecast of Yt , conditional on information available in period t, and "t1 is an error term uncorrelated with information in period t. Formally, conditional forecasts can be seen as intermediate variables, and forecast targeting can be seen as intermediate targeting. As King (1994) stated early in the history of in¯ation targeting, in¯ation targeting means having in¯ation forecasts as intermediate targets.
4
Lessons for the Eurosystem
De®ning price stability In its ®rst announcement of its monetary policy strategy, on 13 October 1998, the Eurosystem (European Central Bank, 1998a) stated: Price stability shall be de®ned as a year-on-year increase in the Harmonised Index of Consumer Prices (HICP) for the euro area of below 2%. Thus, only an upper bound of 2 per cent was unambiguously announced. About a month later, the Eurosystem clari®ed that `increase' in this de®nition shall be interpreted as excluding de¯ation. This might seem to imply that the lower bound for in¯ation is zero. On the other hand, the Eurosystem also stated that it had not announced a lower bound, giving the reason that the size of any measurement bias in the HICP is not known. On 1 December, 1998, the
De®ning and Maintaining Price Stability 89
Eurosystem (European Central Bank, 1998b) announced a monetary reference value for M3 of 4.5 per cent. Then, subtraction of the sum of its estimates of potential output growth and trend decline in velocity from 4.5 per cent revealed that it had applied an in¯ation target of 1.5 per cent. If this is the middle of an interval with an upper bound of 2 per cent, that interval is obviously 1±2 per cent. Hence, the Eurosystem seems to have implicitly announced a lower bound of 1 per cent. Clearly, observers should not have to piece together the de®nition from different statements, including the announcement of the reference value. Any remaining ambiguity with regard to the de®nition seems to serve no purpose and should be eliminated. Instead of the ambiguous and asymmetric statement `below 2%', the Eurosystem had better state the unambiguous and symmetric 1.5 per cent as its in¯ation target, or the interval 1±2 per cent, possibly with the addendum that this de®nition may be somewhat modi®ed when more evidence about the quality of the HICP becomes available. Absent such a statement, it seems that observers should currently interpret the Eurosystem as having an in¯ation target of 1.5 per cent, and evaluate it accordingly. As argued on pp. 62±5, it is by no means clear that interpreting price stability as an in¯ation target should be the ®nal word, since that implies accepting a unit root and non-stationarity of the price level, making `price stability' a misnomer. After some ®ve or ten years of successful in¯ation targeting, I believe the Eurosystem should seriously consider the pros and cons of moving to price-level targeting. Maintaining price stability In addition to the de®nition of price stability, on 13 October, the Eurosystem announced what was later called the `two pillars' of its strategy:
. `a prominent role for money with a reference value for the growth of a monetary aggregate' and
. `a major role for a broadly-based assessment of the outlook for future price developments'.39
With regard to the role of money, the Eurosystem has emphasised that the reference value should not be interpreted as an intermediate target for money growth. Indeed, it has rejected monetary targeting, on the grounds that the relationship between money and prices may not be suf®ciently stable, and that it is not clear that the monetary aggregates with the most stable relationship is suf®ciently controllable in the short run. As Issing (1998) summarises: In these circumstances, relying on a pure monetary targeting strategy would constitute an unrealistic, and therefore misguided, commitment.
90 Lars E. O. Svensson
Instead, the Eurosystem plans to use money growth as an indicator of `risks to price stability,' such that deviations of current money growth from the reference value signals risks to price stability. The reference value will be derived in a manner that ensures, as far as possible, that deviations of monetary growth from the value will signal risks to price stability. In the ®rst instance, such a deviation will prompt further analysis to identify and interpret the economic disturbance that caused the deviation, and evaluate whether the disturbance requires a policy move to counter risks to price stability. (Issing, 1998) As I argue in some detail in Svensson (1999d), there is little ground for such a prominent role for money. It is easily shown in the simple and conventional model used there, that such a money growth indicator will be a relatively useless indicator of risks to price stability ± and, indeed, mostly a noisy indicator of the deviation of current in¯ation from the in¯ation target. As argued on p. 79±82, the weight on money as an indicator should be strictly determined by its predictive power in forecasting in¯ation. As argued on p. 81±82 and demonstrated in some detail in Svensson (1999d), the best indicator of `risks to price stability' is an in¯ation forecast conditional on an unchanged interest rate.40 I believe it worthwhile to look more closely at the Eurosystem's arguments in favour of a prominent role to money, stated very clearly in Issing (1998). Three arguments for giving a prominent role for money are provided: (1) `In¯ation is fundamentally monetary in origin over the longer term', (2) `It creates a ®rm `nominal' anchor for monetary policy and therefore helps to stabilise private in¯ation expectations at longer horizons', and (3) `[It] emphasises the responsibility of the ESCB for the monetary impulses to in¯ation, which a central bank can control more readily than in¯ation itself'. With regard to argument (1), it is based on the empirical high long-run correlation between money and prices. This correlation, however, holds in any model where demand for money is demand for real money, for instance in the simple model used in Svensson (1999d) to demonstrate the inferiority of the Eurosystem's money growth indicator. The correlation is actually a relation between two endogenous variables, and says nothing about causality. The direction of causality is determined by the monetary policy pursued. Under strict monetary targeting, when the central bank aims at maintaining a given money growth rate regardless of what happens to prices, money growth becomes exogenous in the relation and causes in¯ation, which is endogenous. Under in¯ation targeting, when the central bank aims at maintaining a given in¯ation rate regardless of what happens to money, in¯ation becomes exogenous in the relation and causes money growth, which is endogenous. Hence, argument (1) is neutral to the monetary strategy. With regard to argument (2), it seems that the de®nition of price stability provides the best nominal anchor and is the best stabiliser of in¯ation
De®ning and Maintaining Price Stability 91
expectations. Emphasising a second nominal anchor seems redundant and even misguided, since more than one nominal anchor could confuse, rather than stabilise, private expectations. With regard to argument (3), it is not clear how `monetary impulses to in¯ation' can be de®ned in any unambiguous and useful way. I am not sure the argument means anything but monetary aggregates being easier to control than in¯ation. Furthermore, it seems obvious that the Maastricht Treaty does not assign `stability of monetary impulses to in¯ation' as the primary objective of the Eurosystem; it assigns `price stability', period. Thus, the ®rst pillar is unlikely to provide much support for the maintenance of price stability. Instead, the maintenance of price stability must rely on the second pillar, the `broadly-based assessment of the outlook for future price developments'. This is, of course, nothing but a long euphemism for in¯ationforecast targeting. Indeed, I believe the success of the Eurosystem maintenance of price stability depends strongly on its learning to do forecast targeting, as practised by an increasing number of in¯ation-targeting central banks. Strangely enough, Eurosystem statements argue that in¯ation-forecast targeting would be unsuitable for the Eurosystem, on the grounds that forecasting in¯ation will be dif®cult and that the understanding of the transmission mechanism is imperfect. To quote Issing (1998) further:41 In the uncertain environment likely to exist at the outset of Monetary Union, forecasting in¯ation will be dif®cult, not least because of the many conceptual, empirical and practical uncertainties faced by the ESCB at the start of Stage Three. Forecasting models estimated using historic data may not offer a reliable guide to the behaviour of the euro area economy under Monetary Union. Forecast uncertainty is likely to be relatively large. Forecasting in¯ation requires thorough knowledge of the properties of the new euro area-wide data series and experience and understanding of the transmission mechanism of monetary policy in the new euro area economy. Both are likely to be quite different from what we have been used to in the existing environment of eleven distinct national economies prior to Monetary Union. Certainly, forecasting and forecast targeting will not be easy, and forecast uncertainty is likely to be relatively large. Nevertheless, forecasting is simply necessary, given `the need for monetary policy to have a forward-looking, medium-term orientation' that Issing and the Eurosystem emphasise. Furthermore, forecast targeting implies using existing information in the most ef®cient and ¯exible way. It incorporates both model and extra-model information, allows judgemental adjustments, takes additive uncertainty for granted and even allows imperfect understanding of the transmission mechanism and model uncertainty, as I have tried to explain in this chapter. Of course, the less the uncertainty and the better the understanding of the
92 Lars E. O. Svensson
transmission mechanism, the more successful forecast targeting is likely to be. But this does not mean that there is some better way of maintaining price stability if there is more uncertainty and less understanding of the transmission mechanism.42
5
Conclusions
This chapter argues that forecast targeting is the best way of maintaining price stability, on the grounds that with lags and uncertainty in the transmission mechanism, forecast targeting is the most ef®cient and ¯exible way of using available information. By generalising forecast targeting from mean forecast targeting to distribution forecast targeting, it should also be the best way of handling model uncertainty. Indeed, I believe the current best practice in central banks' maintaining price stability must be understood as forecast targeting. The chapter has, so far, discussed only the framework for policy decisions and not at all the central bank's communication, degree of transparency and degree of accountability. Under forecast targeting, the conditional forecasts for in¯ation and the output gap are the crucial inputs in the policy decision. Therefore, policy decisions are best explained and motivated, and policy is best understood and anticipated by the public, with reference to these conditional forecasts. This has the bene®cial effect that any criticism of the policy must be more speci®c: for instance, is it the target or the central bank's forecast that is wrong? Furthermore, making these forecasts public provides the best opportunity for outside observers to monitor and evaluate the central bank's policy, and making sure that its decisions are consistent with its objective. Then policy can be evaluated almost in real time, without waiting some two years to see the outcome of an in¯ation rate that is, by then, contaminated by a number of intervening shocks. Finally, making the forecasts public provides the strongest incentives for the central bank to improve its competence and do the best possible job. These are strong arguments in favour of making these forecasts public, a practice already followed by the Reserve Bank of New Zealand, the Bank of England, and the Sveriges Riksbank. Against this background, the Eurosystem's refusal to publish its forecasts, citing far from convincing arguments,43 is very dif®cult to understand, except perhaps as an expression of an initial lack of con®dence and experience and a desire to further improve its competence before going public (but if so, why not announce that the forecasts will eventually be made public?). I see no reason why the Eurosystem should not aim for the current best standard of transparency, as demonstrated by the central banks already mentioned.
De®ning and Maintaining Price Stability 93
Notes 1. The paper which this chapter is based upon was presented at the Bundesbank conference on `The Monetary Transmission Process: Recent Developments and Lessons for Europe', 26±27 March, 1999. I thank the discussants Mervyn King and Jose Vi~nals, and Claes Berg, Donald Brash, John Faust, Torsten Persson, Anders Vredin and participants in seminars at IIES and Sveriges Riksbank for comments. Special thanks are due to Jon Faust and Dale Henderson. Over the years, I have bene®ted a great deal from many discussions with them on monetary policy and from their very constructive (sometimes relentless) criticism of previous work of mine. Dale has also directed me to an early, very relevant, literature. I also thank È nnblad for editorial and secretarial assistance and Marcus Salomonsson Christina Lo for research assistance. Needless to say, the views expressed and any errors are my own responsibility. 2. This literature includes Bernanke et al (1998), Bernanke and Mihov (1997), Clarida, Gali and Gertler (1998a), Clarida and Gertler (1997), Laubach and Posen (1997), Neumann (1997), von Hagen (1995) (note a crucial typo: the coef®cient for money supply in Table 1 should be 0.07 instead of 0.7). 3. `Strict' and `¯exible' targeting is de®ned below. 4. If arguments in favour of a small positive in¯ation rate is accepted, an upwardsloping price-level target path may be preferable. 5. An interesting issue is to what extent the degree of nominal rigidity depends on whether there is in¯ation or price-level targeting. 6. This result requires at least moderate output persistence with a Lucas-type Phillips curve, and does not hold for a Lucas-type Phillips curve without persistence. Kiley (1998) shows that the result does not hold for a Calvo-type Phillips curve without persistence. 7. See, for instance, McCallum and Nelson (1999) and Williams (1997). 8. As in¯ation-targeting central banks, like other central banks, also seem to smooth instruments, the loss function (2.7.) may also includes the term
it � it�1 2 with > 0. 9. See box on p. 26 in Sveriges Riksbank (1997) as well as Heikensten and Vredin (1998). 10. `actual in¯ation will on occasions depart from its target as a result of shocks and disturbances. Attempts to keep in¯ation at the in¯ation target in these circumstances may cause undesirable volatility in output'. 11. See Bank of England (1998), para. 40: `[I]n any given circumstances, a variety of different interest rate paths could in principle achieve the in¯ation target. What factors were relevant to the preferred pro®le of rates?... There was a broad consensus that the Committee should in principle be concerned about deviations of the level of output from capacity'. 12. The Reserve Bank's target was previously de®ned in terms of a somewhat complex underlying in¯ation rate. In the Policy Target Agreement of December 1997, there was a change to the more transparent CPIX. 13. On the other hand, the argument that in¯ation increases capital market distortions, examined in Feldstein (1997, 1999), would, under the assumption of unchanged nominal taxation of capital, motivate a zero or even a negative in¯ation target. 14. For reasons explained in Gordon (1996), I believe that Akerlof, Dickens and Perry (1996) reach too pessimistic a conclusion. On the other hand, their data is from the
94 Lars E. O. Svensson
15. 16. 17.
18. 19. 20.
21.
22. 23.
24.
25. 26.
United States and Canada, and downward nominal wage rigidity may be more relevant in Europe. The conclusions of Orphanides and Wieland (1998) are sensitive to assumptions about the size of shocks and the average real interest rate; the latter is taken to be 1 per cent for the United States. If the average real rate is higher in Europe, and the shocks not much larger than in the United States, nonnegative interest rates may be of less consequence in Europe. Wolman (1998a, 1998b) provides a rigorous examination of the consequences of non-negative interest rates in a more explicit model, and ®nds relatively small effects. Also, Wolman (1998b) ®nds that a reaction function responding to price-level deviations from a price-level target (rather than in¯ation deviation from an in¯ation target) has good properties for low in¯ation rates. Note that this simple ®rst-order condition only arises if the variables in Xt are predetermined. There is an additional philosophical objection to once-and-for-all commitment: how come the once-and-for-all commitment can be done in period 0? Why was it not already done before, so nothing remains to be committed to in period 0? Why is there something special about period 0? I found this appropriate quote in Budd (1998). See Budd (1998) for an interesting and detailed discussion of the advantages of explicitly considering forecasts rather than formulating reaction functions from observed variables to the instrument. It is important that these expectations are conditional on the central bank's model, and hence are `structural', rather than being private sector expectations, in order to avoid the problems of non-existence or indeterminacy of equilibria, arising from responding mechanically to private sector expectations, as has been emphasised in Woodford (1994) and further discussed in Bernanke and Woodford (1997). The conditional forecasts for arbitrary interest-rate path derived in Svensson (1998a, Appendix A] assume that the interest rate paths are `credible' that is, anticipated and allowed to in¯uence the forward-looking variables. Leeper and Zha (1999) discuss an alternative way of constructing forecasts for arbitrary interest rate paths, by assuming that these interest rate paths result from unanticipated deviations from a normal reaction function. The consequences of imposing the restriction of time-consistency of it remain to be examined. That is, that the elements itjt in it shall be consistent with the decision in period t conditional on Xtjt (see n. 6, p. 93). Note that one way of taking the discretionary nature of decision making into account is to set it in period t under the restriction that the reaction function that will apply in period t for 1 will be itjt ftjt Xtjt , where ftjt is the reaction function that is likely to result from the decision in period t . Note that an equation like it gXtTjt , where XtTjt is a model-consistent forecast (including this equation), especially in a model with forward-looking models, is a rather complex equilibrium condition. For reasons detailed in Svensson (1999b, section 2.3.1), I am rather sceptical about these equilibrium conditions as reaction functions representing in¯ation targeting. See Tinsley (1975) and Reifschneider, Stockton and Wilcox (1997) for further discussion of judgemental adjustments. Mervyn King has emphasised that it is important that the decision making body of the central bank agrees with the forecast. This requires iterations between the staff and the decision making body, with the decision making body having the last say on the forecast.
De®ning and Maintaining Price Stability 95 27. Thus, for the `strict' case with in¯ation as the only argument in the loss function, the four loss functions are (1) Lt 12
t � 2 , (2) Lt jt � j, (3) Lt k �
t � , where
x is the so-called R 1Dirac delta function with the properties
x 0 for x 6 0;
0 1; and �1
xdx 1, and (4) Lt 0 for jt � j a > 0, Lt k > 0 for jt � j > a. 28. Another problem with reporting the mode forecast is evident in the hypothetical case when the forecast is bimodel with approximately equal probability density at the two modes. 29. Two separate arguments are sometimes presented in favour of emphasising the mode forecast. First, in presenting and discussing the forecast, it may often be natural and intuitive to consider a most likely scenario together with one or two alternative scenarios. The most likely scenario would then correspond to the mode forecast. Second, before that stage, in constructing the forecast, it may be practical to start with a most likely scenario and then add various uncertainties and complications later on. Whereas the ®rst argument may be a legitimate argument in favor of the mode, the second is not, since the presentation and the construction can be independent. Furthermore, the mode and the median have the property that they are not affected by outlines, which may or may not be an advantage, depending on one's view. 30. See Orphanides (1998) and Smets (1998) for recent related work. 31. In a steady state, the Kalman gain is given by K PD0
DPD0 vv �1 , where the covariance matrix P of the forecast errors Xt � Xtjt�1 is given by P MP � PD0
DPD0 vv �1 DPM 0 uu , where M is the transition matrix in the transition equation Xt1 MXt ut1 : 32. That such a money growth indicator is unsuitable on its own is fairly obvious, since money is not the only, not even the major, predictor of in¯ation at the horizons relevant for monetary policy (see Estrella and Mishkin, 1998). What is perhaps less obvious is that the money growth indicator is unsuitable even for a completely stable money-demand function without velocity shocks (see Svensson, 1999d, and Rudebusch and Svensson, 1999b). 33. More speci®cally, with forward-looking variables, the interest rate is kept unchanged for a few periods (four±six quarters, say), but then, there is a shift to `normal' policy, or to some policy stabilizing in¯ation and determining the future forward-looking variables. 34. A non-quadratic loss function would also imply that certainty-equivalence no longer holds (see p. 78±79). 35. Note that we use the convention that tr s r 1 for t < s: 36. Recent work on and discussion of monetary policy under model uncertainty includes Blinder (1998), Cecchetti (1997), Clarida, Gali and Gertler (1998b), Estrella and Mishkin (1998), Levin, Wieland and Williams (1999), McCallum (1997), Onatski and Stock (1998), Peersman and Smets (1998), Rudebusch (1998), Sack È derstro
(1988), Sargent (1998) Smets (1998), So È m (1999a, 1999b), Stock (1998), Svensson (1999c), and Wieland (1996, 1998).
In contrast to the standard Brainard result in favour of caution, work on so-called robust control (where the reaction function is chosen so as to minimise expected loss for the most unfavourable model) indicates that model uncertainty may well result in more aggressive optimal responses (see, for instance, Sargent, 1998, and Stock, 1998). 37. The Bank of England's fan charts for in¯ation and output should probably be interpreted as marginal distributions. However, since the distributions for in¯ation
96 Lars E. O. Svensson and the output gap are unlikely to be independent, distribution forecast targeting requires the joint distribution to be conveyed. This may require some further innovation in display, beyond the already beautiful fan chart. 38. As discussed in Wallis (1999), the Bank of England's fan charts present prediction intervals that differ from normal con®dence intervals, central prediction intervals. The Sveriges Riksbank, however, presents normal con®dence intervals, (see Blix and Sellin, 1998). Both banks present the mode as their point forecast, whereas it seems to me that it would be more natural and consistent with the theory to present the mean (or, in distribution forecast targeting, at least the median). 39. See Bryant (1980), Friedman (1975), Kalchbrenner and Tinsley (1975) and Kareken, Muench and Wallace (1973) for this and other arguments against intermediate targeting in general and monetary targeting in particular. See Rudebusch and Svensson (1999b) for simulations of monetary targeting in the U.S. with lessons for the Eurosystem. These simulations show that monetary targeting in the United States would be quite inef®cient compared to ¯exible in¯ation targeting, in the sense of bringing higher variability of both in¯ation and the output gap. 40. Since the ®rst version of this paper was written, an extensive discussion and motivation of Eurosystem strategy has been presented by Angeloni, Gaspar and Tristani (1999). 41. Rudebusch and Svensson (1999b) examine monetary targeting in an empirical model of in¯ation, output and money for US data and draw some lessons for the Eurosystem. They ®nd that monetary targeting would be very inef®cient compared to in¯ation targeting, in the sense of increasing the variability of both in¯ation and output. Counter to conventional wisdom, this is the case also if money demand shocks are set to zero so the money demand is completely stable. Gerlach and Svensson (1999) examine the indicator properties of monetary aggregates for the euro area. Somewhat surprisingly, they ®nd considerable empirical support for the so-called P model of Hallman, Porter and Small È dter and Reimers (1994). This implies that (1991), adapted to Germany by To monetary aggregates, in the form of the `real money gap', the gap between current real balances and long-run equilibrium real balances, has considerable predictive power for future in¯ation. They ®nd little or no empirical support for the Eurosystem's money growth indicator, though. Indeed, the theoretical analysis in Svensson (2000a) shows that the P model, although emphasising the role of the real money gap in forecasting and controlling in¯ation, does not provide any support for a Bundesbank-style money growth target or a Eurosystem-style money-growth indicator. 42. See also Angeloni, Gaspar and Tristani (1999). 43. As I argue in Svensson (1999d), it may be sobering to recall that the introduction of in¯ation targeting in the United Kingdom, Sweden and Finland occurred under rather dramatic circumstances. The countries went through dramatic boom±bust experiences, very serious banking and ®nancial sector crises, and a dramatic sudden shift from a ®xed exchange rate to a new monetary policy regime with a ¯oating exchange rate. Furthermore, this occurred in a situation with very low credibility for monetary policy, with high and unstable in¯ation expectations, much above the announced in¯ation targets. At least for Sweden (where I am naturally more informed) the central bank's commitment to the ®xed exchange rate was so strong, that there was no contingency planning. When the krona was ¯oated in November 1992, the new in¯ation-targeting regime, which was announced in January 1993,
De®ning and Maintaining Price Stability 97 had to be conceived from scratch (although, of course, with the bene®t of the experiences mainly from New Zealand and Canada). It is not easy to rank dif®culties and uncertainty about the transmission mechanism, but it seems to me that the dif®culties facing the Eurosystem are still not of the same magnitude as the dif®culties that the central banks of the United Kingdom, Sweden and Finland were facing. Since those central banks have, nevertheless, managed quite well, the odds for the Eurosystem may be quite good, provided it adopts a similar framework for policy decisions. 44. See, for instance, Duisenberg (1998): `publishing an in¯ation forecast would obscure rather than clarify what the Governing Council is actually doing. The public would be presented with a single number intended to summarise a thorough and comprehensive analysis of a wide range of indicator variables. However, such a summary would inevitably be simplistic. Moreover, because publishing a single in¯ation forecast would be likely to suggest that monetary policy reacts mechanistically to this forecast, publication might mislead the public and therefore run counter to the principle of clarity'.
References Akerlof, G. A., W. T. Dickens and G. L. Perry (1996), `The Macroeconomics of Low In¯ation', Brookings Papers on Economic Activity, 1, 1±76. Angeloni, I., V. Gaspar and O. Tristani (1999) `The Monetary Policy Strategy of the ECB', ECB Working Paper. Aoki, K. (1998). `On the Optimal Monetary Policy Response to Noisy Indicators', Princeton University, Working Paper. Backus, D. and J. Drif®ll (1986) `The Consistency of Optimal Policy in Stochastic Rational Expectations Models', CEPR Discussion Paper, 124. Ball, L. (1997) `Ef®cient Rules for Monetary Policy', NBER Working Paper, 5952. ÐÐÐÐ (1999) `Policy Rules for Open Economies', in Taylor (1999). Bank of Canada (1994) Economic Behavior and Policy Choice under Price Stability, Ottawa, Bank of Canada. Bank of England (1998) Minutes of the Monetary Policy Committee meeting on 4±5 March 1998, Bank of England. Batini, N. and A.G. Haldane (1999) `Forward-looking Rules for Monetary Policy', in Taylor (1999). Berg, C. and L. Jonung (1999) `Pioneering Price Level Targeting: The Swedish Experience 1931±1937', Journal of Monetary Economics, 43, 3, 525±51. Bernanke, B.S., T. Laubach, F.S. Mishkin and A.S. Posen (1998) In¯ation Targeting: Lessons from the International Experience, Princeton University Press. Bernanke, B.S., and I. Mihov (1997) `What Does the Bundesbank Target?', European Economic Review, 41, 1025±54. Bernanke, B.S., and F.S. Mishkin (1997) `In¯ation Targeting: A New Framework for Monetary Policy?', Journal of Economic Perspectives; 11, 97±116. Bernanke, B.S. and M. Woodford (1997) `In¯ation Forecasts and Monetary Policy', Journal of Money, Credit, and Banking, 29, 653±84. Blinder, A.S. (1998) Central Banking in Theory and Practice, Cambridge, MA, MIT Press. Blix, M. and P. Sellin (1998) `Uncertainty Bands for In¯ation Forecasts', Sveriges Riksbank Working Paper Series, 65. Brainard, W. (1967) `Uncertainty and the Effectiveness of Policy', American Economic Review, Papers and Proceeding, 57, 411±25.
98 Lars E. O. Svensson Britton, E., P. Fisher and J. Whitley (1998) `The In¯ation Report Projections: Understanding the Fan Chart', Bank of England Quarterly Bulletin, 38, 30±7. Bryant, R.C. (1980) Money and Monetary Policy in Interdependent Nations, Washington D.C, Brookings Institution. Bryant, R.C., P. Hooper and C.L. Mann (eds) (1993) Evaluating Policy Regimes: New Research in Empirical Macroeconomics, Washington, DC, The Brookings Institution. Budd, A. (1998) `Economic Policy, with and without Forecasts', Bank of England Quarterly Bulletin, 86, 379±84. Cecchetti, S.G. (1995) `In¯ation Indicators and In¯ation Policy', NBER Macroeconomics Annual, 10, 189±219. ÐÐÐÐ (1997) `Central Bank Policy Rules: Conceptual Issues and Practical Considerations', Federal Reserve Bank of New York, Working Paper. Chow, G.C. (1975) Analysis and Control of Dynamic Economic Systems, New York, John Wiley. Clarida, R., J. Gali and M. Gertler (1998a) `Monetary Policy Rules in Practice: Some International Evidence', European Economic Review, 42, 1033±67. ÐÐÐÐ (1998b) `The Science of Monetary Policy, A New Keynesian Approach', NBER Working Paper 7147. Clarida, R. and M. Gertler (1997) `How the Bundesbank Conducts Monetary Policy', in C.D. Romer and D.H. Romer (eds), Reducing In¯ation: Motivation and Strategy, Chicago, Chicago University Press. Clark, P., D. Laxton and D. Rose (1996) `Asymmetry in the U.S. Output±In¯ation Nexus', IMF Staff Papers, 43, 216±51. Currie, D. and P. Levine (1992) `Should Rules Be Simple?', Economics and Planning, 25, 113±38, reprinted in Currie and Levine (1993). ÐÐÐÐ (1993) Rules, Reputation and Macroeconomic Policy Coordination, Cambridge, Cambridge University Press. Debelle, G. and D. Laxton (1997) `Is the Phillips Curve Really a Curve? Some Evidence for Canada, the United Kingdom, and the United States', IMF Staff Papers, 44, 249±82. Duisenberg, W.F.(1998) `The ESCB's Stability-oriented Monetary Policy Strategy', speech in Dublin, 10 November < http://www.ecb.int >. Estrella, A. and F.S. Mishkin (1997) `Is there a Role for Monetary Aggregates in the Conduct of Monetary Policy?', Journal of Monetary Economics, 40, 279±304. ÐÐÐÐ (1998) `Rethinking the Role of NAIRU in Monetary Policy: Implications of Model Formulation and Uncertainty', NBER Working Paper, 6518. European Central Bank (1998a) `A Stability-oriented Monetary Policy Strategy for the ESCB', Press Release, 13 October < http://www.ecb.int >. ÐÐÐÐ (1998b) `The Quantitative Reference Value for Monetary Growth', Press Release, 1 December < http://www.ecb.int >. ÐÐÐÐ (1999) `The Stability-oriented Monetary Policy Strategy of the Eurosystem', ECB Monthly Bulletin, January, < http://www.ecb.int >, 39±50. European Monetary Institute (1997) The Single Monetary Policy in Stage Three: Elements of the Monetary Policy Strategy in the ESCB, European Monetary Institute, Frankfurt. Federal Reserve Bank of Kansas City (1996) Achieving Price Stability, Kansas City, Federal Reserve Bank of Kansas City. Feldstein, M. (1997) `The Costs and Bene®ts of Going from Low In¯ation to Price Stability', in C.D. Romer and D.H. Romer (eds), Reducing In¯ation: Motivation and Strategy, Chicago, University of Chicago Press. Feldstein, Martin (ed.) (1999) Costs and Bene®ts of Price Stability, Chicago, Chicago University Press.
De®ning and Maintaining Price Stability 99 Fischer, S. (1994) `Modern Central Banking', in F. Capie, C. Goodhart, S. Fischer and N. Schnadt, The Future of Central Banking, Cambridge, Cambridge University Press. ÐÐÐÐ (1996) `Why Are Central Banks Pursuing Long-Run Price Stability?', in Federal Reserve Bank of Kansas City (1996). Fisher, I. (1934) Stable Money; published 1935 in Britain under the title of Stable Money: A History of the Movement, London, Allen & Unwin. Friedman, B.M. (1975) `Targets, Instruments, and Indicators of Monetary Policy', Journal of Monetary Economics, 1, 443±73. Gerlach, S. and G. Schnabel (1998) `The Taylor Rule and Average Interest Rates in the EMU-11 Area: A Note', Bank for International Settlements, Working Paper. Gerlach, S. and L.E.O. Svensson (1999) `Money and In¯ation in the Euro Area: A Case for Monetary Indicators?', Working Paper. Gordon, R.J. (1996) `Comments and Discussion', Brookings Papers on Economic Activity 1, 60±6. ÐÐÐÐ (1997) `The Time-Varying NAIRU and its Implication for Economic Policy', Journal of Economic Perspectives, 11(1), 11±32. Greenspan, A. (1994) `Discussion', in F. Capie, C. Goodhart, S. Fischer and N. Schnadt, The Future of Central Banking, Cambridge, Cambridge University Press. Hall, R.E. (1984) `Monetary Strategy with an Elastic Price Standard', in Price Stability and Public Policy, Kansas City, Federal Reserve Bank of Kansas City, 137±59. Hallman, J.J., R.D. Porter and D.H. Small (1991) `Is the Price Level Tied to the M2 Monetary Aggregate in the Long Run?', American Economic Review, 81, 841±58. Heikensten, L. and A. Vredin (1998) `In¯ation Targeting and Swedish Monetary Policy ± Experience and Problems', Sveriges Riksbank Quarterly Review, 4/1998, 5±33. HM Treasury (1997) `Remit for the Monetary Policy Committee', News Release, 12 June. Isard, P. and D. Laxton (1996) `Strategic Choice in Phillips Curve Speci®cation: What if Bob Gordon is Wrong?', Paper prepared for a conference on European unemployment in Florence, Italy. Issing, O. (1998) `The European Central Bank at the Eve of EMU', speech in London, 26 November. Kalchbrenner, J.H. and P.A. Tinsley (1975) `On the Use of Optimal Control in the Design of Monetary Policy', Federal Reserve Board, Special Studies Paper, 76. Kareken, J.H., T. Muench and N. Wallace (1973) `Optimal Open Market Strategy: The Use of Information Variables', American Economic Review, 63, 156±72. Kiley, M.T. (1998) `Monetary Policy under Neoclassical and New-Keynesian Phillips Curves, with an Application to Price Level and In¯ation Targeting', Federal Reserve Board, Working Paper. King, M.A. (1994) `Monetary Policy in the UK', Fiscal Studies, 15, (3), 109±28. ÐÐÐÐ (1996) `How Should Central Banks Reduce In¯ation? ± Conceptual Issues', in Federal Reserve Bank of Kansas City (1996). ÐÐÐÐ (1997) `Changes in UK Monetary Policy: Rules and Discretion in Practice', Journal of Monetary Economics, 39, 81±97. Laubach, T. and A.S. Posen (1997) Disciplined Discretion: Monetary Targeting in Germany and Switzerland, Essays in International Finance, 206, Princeton University. Leeper, E.M., and T. Zha (1999) `Modest Policy Intervention', Indiana University, Bloomington, Working Paper. Leiderman, L. and L.E.O. Svensson (eds) (1995) In¯ation Targets, London, Centre for Economic Policy Research. Levin, A., A.V. Wieland and J.C. Williams (1999) `Robustness of Simple Monetary Policy Rules under Model Uncertainty', in Taylor (1999).
100 Lars E. O. Svensson LeRoy, S.F. and R.N. Waud (1977) `Applications of the Kalman Filter in Short-run Monetary Control', International Economic Review, 18, 195±207. Lowe, P. (ed.) (1997) Monetary Policy and In¯ation Targeting, Sydney: Reserve Bank of Australia. McCallum, B.T. (1997) `Issues in the Design of Monetary Policy Rules', NBER Working Paper, 6016. McCallum, B.T. and E. Nelson (1999) `Nominal Income Targeting in an Open-economy Optimizing Model', Journal of Monetary Economics, No. 43, 3. Neumann, M. (1997) `Monetary Targeting in Germany', in K. Iwao (ed.), Towards More Effective Monetary Policy, New York, St. Martin's Press. Onatski, A. and J.H. Stock (1998) `Robust Monetary Policy under Model Uncertainty in a Small Model of the US Economy', Harvard University, Working Paper. Orphanides, A. (1998) `Monetary Policy Evaluation with Noisy Information', Federal Reserve Board, Working Paper. Orphanides, A. and V. Wieland (1998) `Price Stability and Monetary Policy Effectiveness when Nominal Interest Rates Are Bounded at Zero', Federal Reserve Board, Working Paper. Pearlman, J.G. (1992) `Reputational and Nonreputational Policies under Partial Information', Journal of Economic Dynamics and Control, 16, 339±57. Pearlman, J.G., D. Currie and P. Levine (1986) `Rational Expectations Models with Partial Information', Economic Modelling, 3, 90±105. Peersman, G. and F. Smets (1998) `Uncertainty and the Taylor Rule in a Simple Model of the Euro-Area Economy', Working Paper, presented at a conference on Monetary Policy and Monetary Institutions held on March 5, 1999 under the joint sponsorship of the Federal Reserve Bank of San Fransciso and the Stanford Institute for Economic Policy Research. Reifschneider, D.L., D.J. Stockton and D.W. Wilcox (1997) `Econometric Models and the Monetary Policy Process', Carnegie±Rochester Conference Series on Public Policy, 47, 1±37. Rudebusch, G.D. (1998) `Is the Fed Too Timid? Monetary Policy in an Uncertain World', Federal Reserve Bank of San Francisco, Working Paper. Rudebusch, G.D., and L.E.O. Svensson (1999a) `Policy Rules for In¯ation Targeting', in Taylor (1999). ÐÐÐÐ (1999b) `Eurosystem Monetary Targeting: Lessons from U.S. Data', NBER Working Paper, 7179. Sack, B. (1988) `Does the Fed Act Gradually? A VAR Analysis', Federal Reserve Board, Discussion Paper, 98±17. Sargent, T.J. (1999) `Discussion of ``Policy Rules for Open Economies'' by Laurence Ball', in Taylor (1999). Smets, F. (1999) `Output Gap Uncertainty: Does it Matter for the Taylor Rule?' Bank for International Settlements, Working Paper. So È derlind, P. (1998) `Solution and Estimation of RE Macromodels with Optimal Policy', Stockholm School of Economics, Working Paper. So È m, U. (1999a) `Monetary Policy with Uncertain Parameters', Stockholm School È derstro of Economics, Working Paper. ÐÐÐÐ (1999b) `Should Central Banks Be More Aggressive?' Stockholm School of Economics, Working Paper. Stock, J.H. (1999) `Making Policies Robust to Model Uncertainty: Comment on ``Policy Rules for In¯ation Targeting'' by Glenn Rudebusch and Lars E.O. Svensson', in Taylor (1999). Svensson, Lars E.O. (1996) 'Commentary: How Should Monetary Policy Respond to Shocks while Maintaining Long-run Price Stability?±Conceptual Issues', in Federal Reserve Bank of Kansas City (1996).
De®ning and Maintaining Price Stability 101 ÐÐÐÐ (1997a) `In¯ation Forecast Targeting: Implementing and Monitoring In¯ation Targets', European Economic Review, 41, 1111±46. ÐÐÐÐ (1997b) `Price Level Targeting vs. In¯ation Targeting', Working Paper version. ÐÐÐÐ (1998a) `In¯ation Targeting as a Monetary Policy Rule', NBER Working Paper, 6790. ÐÐÐÐ (1998b) `In¯ation Targeting in an Open Economy: Strict vs. Flexible In¯ation Targeting', Victoria Economic Commentaries, 15-1, Victoria University of Wellington. ÐÐÐÐ (1999a) `Better to Respond to Determinants of Targets than to Targets Themselves', mimeo. ÐÐÐÐ (1999b) `In¯ation Targeting as a Monetary Policy Rule', Journal of Monetary Economics, 43, 607±54. ÐÐÐÐ (1999c) `In¯ation Targeting: Some Extensions', Scandinavian Journal of Economics, 101(3), 337±61. ÐÐÐÐ (1999d) `Monetary Policy Issues for the Eurosystem', Carnegie±Rochester Conference Series on Public Policy, 51±1, 79±136. ÐÐÐÐ (1999e) `Price Level Targeting vs. In¯ation Targeting', Journal of Money, Credit and Banking, 31, 277±95. ÐÐÐÐ (2000a) `Does the P Model Provide Any Rationale for Monetary Targeting?', Working Paper, German Economic Review 1, 69±81. ÐÐÐÐ (2000b) `Open-Economy In¯ation Targeting', Journal of International Economics, Vol. no. 50, 155±83. Svensson, Lars E.O. and M. Woodford (2000) `Indicators for Optimal Policy under Asymmetric Information', mimeo. Sveriges Riksbank (1997), In¯ation Report, September 1997, Stockholm: Sveriges Riksbank. Taylor, John B. (1996) `How Should Monetary Policy Respond to Shocks while Maintaining Long-run Price Stability ± Conceptual Issues', in Federal Reserve Bank of Kansas City (1996). ÐÐÐÐ (1998) `The Robustness and Ef®ciency of Monetary Policy Rules as Guidelines for Interest Rate Setting by the European Central Bank', IIES Seminar Paper, 649. Taylor, John B. (ed.) (1999) Monetary Policy Rules, Chicago, Chicago University Press. Tinsley, P.A. (1975) `On Proximate Exploitation of Intermediate Information in Macroeconomic Forecasting', Federal Reserve Board, Special Studies Paper, 59. Todter, K.-H., and H.-E. Reimers (1994) `P-Star as a Link between Money and Prices in Germany', Weltwirtschaftliches Archiv, 130, 273±89. Vickers, J. (1998) `In¯ation Targeting in Practice: the UK Experience', Bank of England Quarterly Bulletin, 38, 368±75. von Hagen, Jurgen (1995) `In¯ation and Monetary Targeting in Germany', in Leiderman and Svensson (1995). Wallis, K.F. (1999) `Asymmetric Density Forecasts of In¯ation and Bank of England's Fan Chart', National Institute Economic Review, 1/99, 106±12. Wieland, V. (1996) `Monetary Policy, Parameter Uncertainty and Optimal Learning', Federal Reserve Board, Working Paper. ÐÐÐÐ (1998) `Monetary Policy and Uncertainty about the Natural Unemployment Rate', Federal Reserve Board, Finance and Economics Discussion Paper, 22. Williams, J.C. (1997) `Simple Rules for Monetary Policy', Federal Reserve Board, Working Paper. Wolman, A.L. (1998a) `Staggered Price Setting and the Zero Bound on Nominal Interest Rates', Federal Reserve Bank of Richmond Economic Quarterly, 84(4), 1±24. ÐÐÐÐ (1998b) `Real Implications of the Zero Bound on Nominal Interest Rates', Federal Reserve Bank of Richmond, Working Paper.
102 Lars E. O. Svensson Woodford, M. (1994) `Nonstandard Indicators for Monetary Policy: Can Their Usefulness Be Judged from Forecasting Regressions?' in G.N. Mankiw, (ed.), Monetary Policy, Chicago, University of Chicago Press.
Discussion
Mervyn King
It has been said about a great newspaper that `all human life is there'. It could equally be said about this chapter that `all Lars Svensson is there'. What Chapter 2 does is to bring together many of Svensson's contributions to the analysis of in¯ation targets, together with some new ideas, and ask the question ± what can the European Central Bank (ECB) learn from this work? It is a good question, and I hope that even if the ECB does not agree with everything that Lars says, it will nevertheless bene®t from a serious discussion of the points which he makes. I should like at the outset to acknowledge the personal contribution which Lars Svensson has made to the theoretical analysis of in¯ation targets. Those of us working in central banks with in¯ation targets know only too well the debt we owe him for the intellectual development which he has made. And, most important, this has been made in a spirit of intellectual open-mindedness. This is important. No central bank can claim a monopoly of wisdom, and in¯ation targets share much in common with more traditional monetary policy frameworks such as monetary targets. Both approaches can learn from the other. But it is clear that the 1990s have been the decade of in¯ation targets, and Svensson has made a major contribution to that achievement. The central point in Svensson's chapter is the need for `a systematic operational framework for policy decisions by central banks'. Svensson sees forecast targeting as the most sensible way forward. Among the class of policy frameworks in this category in¯ation targeting appears superior. Svensson contrasts the systematic approach of forecast targeting with two alternatives. These are, ®rst, simple instrument rules and, second, intermediate targets. I agree with Svensson's ranking of these three alternative approaches to monetary policy. But one must be a little cautious before drawing hard and fast distinctions between them. No advocate of instrument rules would suggest that policy be put on autopilot and that the rule be used in a completely mechanical way to set policy. Equally, advocates of intermediate monetary targets would wish, in certain circumstances, to deviate from the policy implied by a mechanical application of those targets, as the Bundesbank 103
104 Mervyn King
did with money targets for many years. Of course, if the deviation from the rule or the target is too great then its use either as a means of discipline or a form of communication becomes low. But there is equally no simple mechanical link between any particular summary statistic of the in¯ation forecast and the choice of the policy instrument (usually the short-term interest rate). There is always a judgement about what policy setting is appropriate given the outlook for in¯ation. The reason I prefer in¯ation targeting is that it makes clear that the discretion used is in the mapping from the outlook for in¯ation to the choice of instrument. That enables the central bank to divide its work into two parts. The ®rst is to think through and then explain its assessments of the outlook for in¯ation and output implied by the current state of the economy. The second is to explain why, given that outlook, a particular policy choice was appropriate. That seems to me to correspond to what central banks do in practice. In contrast, simple instrument rules can be useful cross-checks on choices made but are not very useful as communication strategies. And intermediate targets likewise pose the problem of how one explains a deviation of policy from that implied by the intermediate target. Surely that would end up being an explanation in terms of the outlook for in¯ation. Hence I like Svensson's phrase `distribution forecast targeting'. But it is important to note that this is a framework for policy and not an optimal rule for interest rates. And one factor which is striking in practice is that the choice between the three approaches identi®ed by Svensson implies nothing about views concerning the transmission mechanism of monetary policy. Indeed, I could quite easily envisage my `optimal monetary policy council', whether for the ECB or any other area, comprising individuals who differed in their belief about the choice of policy framework. What matters most is their ability to think of the transmission mechanism and to form judgements about the current state of the economy. Let me turn now to four speci®c points on Svensson's chapter. First, Svensson advocates moving gradually from in¯ation targets to price level targets. I have some sympathy with his arguments. There is still the important question of how long is the horizon over which the price level is brought back to its target path. It would be useful to carry out simulations to judge how far an instrument rule, such as the Taylor rule, would imply a different interest rate path if an error-correction term were added in order to bring the price level back to a predetermined path within some reasonable time horizon. How much greater would the volatility of interest rates be? But there is also a deeper question of whether the costs of in¯ation are more to do with instability of the expected price level over long horizons or of in¯ation itself. More work is required on that question which, understandably, is not the focus of Svensson's chapter. Second, Svensson criticises the ECB for announcing a range for their de®nition of price stability (close to an in¯ation target) rather than a point
Discussion 105
target. Again, I have some sympathy with this point. The symmetry of the Bank of England's in¯ation target has been an immensely helpful aspect of our remit because we have been able to show that it can be necessary to increase or decrease interest rates according to circumstances. It is important to avoid giving the impression that central banks are reluctant to cut interest rates and quick to raise them. That would be the opposite of the politician's preference for raising interest rates `when necessary' and lowering them `when possible'. Symmetry should be an important part of the approach to setting interest rates. The problem with a target range is that there is a lack of clarity about the objective of monetary policy. Of course one might argue that the central bank would do best by aiming for the centre of the range, and economic agents would know that. The analogy is that, faced with an open goal and with some uncertainty about your ability to kick a football, the best advice is to aim for the point between the middle of the goalposts. Sadly, many footballers of my acquaintance seem to be confused by an open goal, hesitate, and too often miss. A point target has the virtue of concentrating the mind. It is also particularly useful in the context of a policy making committee, to demonstrate that the concept of `hawks' and `doves' on the same committee has no meaning when the committee has a common point in¯ation target. Third, Svensson suggests that the use of the reference value for money by the ECB is misguided. The announcement of a money target might confuse agents rather than `anchor' in¯ationary expectations. I have more sympathy with the idea that money is special. There is still much that we do not fully understand about the transmission mechanism, but it is dif®cult to talk about in¯ation or monetary policy without according a special role for money. And many of the models which play down the role of money and give a prominent place to such concepts as the neutral real interest rate or the equilibrium exchange rate presume more knowledge about the empirical values of these concepts than is available in practice to central banks. So I think there are two reasons for allowing a special role to money. The ®rst is that given uncertainty about the appropriate model for the transmission mechanism, money may have a robust property with respect to nominal outcomes. Second, there is a real danger that focusing solely on traditional econometric models results in trying to explain in¯ation solely in terms of real variables. So it is extremely important that a central bank sees as one of its ®rst responsibilities the need to explain movements in money and why it thinks that its policy stance is consistent with the likely behaviour of money. Fourth, Svensson considers the choice between targeting the mean, median or modal future in¯ation rate. He points out that different loss functions might imply the need to target different summary statistics of the distribution of future in¯ation. More generally, one of the aspects of central bank discretion is to decide how it should react to the entire distribution of in¯ation prospects rather than commit itself to targeting a particular summary statistic in all circumstances. Hence I much prefer Svensson's approach of considering
106 Mervyn King
the distribution of in¯ation as a whole and not just summary statistics such as the mean, median or mode. That is also why in the `fan charts' for the in¯ation outlook which are published in the Bank of England's In¯ation Report, we stress that there can be no mechanical link between any particular line on that chart and the policy decision. In some circumstances it may make sense to focus very much on the mean outlook, and there is no doubt that over a long period any deviation from that must cancel out if the target is to be met on average. But there may be circumstances in which the mean is affected by a very small probability of a signi®cant upside or downside risk. And it may not make sense to adjust policy today to that eventuality. The fan chart should be constructed in order to present information about the future outlook for in¯ation ± the entire distribution ± in an informative way. The policy decision is separate. In conclusion, I would stress that all central banks, whatever policy framework they use, face essentially the same problem. The lags in monetary policy mean that they must look forward to assess the impact of policy decisions on the outlook for in¯ation and output. There are important differences in the way in which the policy decision is presented, and the choice of framework is not unimportant. As Svensson, I prefer the in¯ation target approach. But I do believe that central banks can learn from each other, and that just as the Bank of England learnt a great deal from the experience of the Bundesbank in the period when it was developing its new approach and acquiring the credibility which was necessary for the granting of independence in 1997, I hope that our experience will offer some lessons for the ECB.
Discussion
Jose Vin Äals
Chapter 2 is a work in the best Svensson tradition, in the sense that it summarises, completes and builds on previous work by the author in the ®eld of monetary policy; it is written in a very clear, organised and pedagogical manner; and it makes use of economic theory to obtain results which are of direct policy relevance in general and also for the Eurosystem. The chapter addresses three blocks of issues related, respectively, to the de®nition of price stability, the maintenance of price stability and the lessons to be drawn for the Eurosystem. Although I shall say something on each of these issues, my comments will focus mainly on the second issue, which I consider to be the most important part of the chapter. On the de®nition of price stability, I believe that the main question addressed by the author is whether price stability should be understood as constancy of the price level or as a suf®ciently low rate of in¯ation. From a conceptual viewpoint, the debate is far from settled, and it is presented as such in the chapter. Still, I think it is interesting to consider Svensson's idea that constancy of the price level may not only be preferable on long-term grounds ± since it leads to a lower variance of the price level ± but, under certain circumstances, also on short-term grounds. In particular, in those cases where there is a serious risk of de¯ation or when de¯ation actually occurs, having a constant price-level objective would be preferable to having a low in¯ation objective as long as it is credible before the eyes of the public. This would make in¯ationary expectations higher, as a result of the expected return of the price level to its initial value, and real interest rates considerably lower. More generally, however, as the author himself recognises, the issue of whether a constant price level or a suf®ciently low rate of in¯ation is a preferable de®nition of price stability is far from being conceptually settled. For this reason, he resorts to a pragmatic solution to my liking: for the time being, achieving and maintaining low in¯ation is a suf®ciently ambitious objective for price stability. This is, moreover, the solution chosen by the Eurosystem and other central banks. The discussion of the issues concerning the maintenance of price stability is, in my view, the core of the chapter and where the most substantive points are 107
108 Jose Vin Äals
raised, both from a theoretical and a practical perspective. Svensson compares the relative merits of three strategies for monetary policy: commitment to a simple instrument rule (e.g. a Taylor rule or a money base rule), forecast targeting (e.g. in¯ation-forecast targeting) and intermediate targeting (e.g. monetary targeting). Before going any further, I would like to say that if the author had written his paper ten years ago, he would have had an even more dif®cult task than at present in trying to ®nd out which is the best way of conducting monetary policy ± that is, which is the best monetary policy practice. The reason why this task is made somewhat easier nowadays ± without having at all become trivial ± is that in recent years a broad consensus has emerged both among economists and central bankers about what should be the most appropriate institutional framework for monetary policy. Consequently, whatever the strategy chosen, monetary policy decisions are nowadays taken within an institutional framework where the maintenance of price stability is the main, the primary or, at least, an important goal of monetary policy, and where central banks enjoy a considerable degree of independence in the pursuit of such goal. Without going as far as saying that this makes which strategic framework is ®nally chosen to conduct monetary policy irrelevant, it nevertheless cannot be denied that those central banks which pursue price stability with a considerable degree of independence, in spite of adhering to formally very different monetary policy strategies tend, in practice, to look at broadly comparable sets of economic variables ± albeit in more or less formal ways ± and to make interest rate decisions that are rather similar when faced with similar circumstances. Yet, it is of paramount importance to realise that alternative strategies are not equally effective in reaching the established policy objectives. Thus, it becomes indispensable to carry out a comparative analysis of the available strategies, such as that performed by Svensson in the chapter 2. Before critically appraising the analysis, let me just point out what I would personally view as desirable features of what constitutes 'best monetary policy practice'. First, it should combine rigour and discipline in the medium-term policy stance with some room for ¯exibility to respond to certain macroeconomic disturbances in the shorter term. Secondly, it should be forwardlooking, given the lags with which monetary policy affects price developments. Thirdly, it should not be run mechanistically, but rather make an ef®cient use of the available information, both model-based and extra-model. Fourthly, it should ensure that the decision-making process performs reasonably well across different models of the economy, given how imperfect our knowledge is about the transmission mechanism of monetary policy. And ®nally, both the strategy and the decisions made on the basis of it should be clearly and transparently communicated to the public, so as to favourably in¯uence the private sector's expectations.
Discussion 109
In principle, such 'best monetary policy practice' ought to be able to solve the problems frequently encountered in the rules vs. discretion debate. On the one hand, it would avoid the dynamic consistency problems arising from an excessively discretionary management of monetary policy, which make it impossible to maintain price stability. And, on the other, it would overcome the drawbacks associated with ®xed or rigid rules which, by predetermining the course of monetary variables or instruments, may unduly constrain, in actual practice, policy-makers' scope to react to useful information. The comparison of alternative strategic frameworks for maintaining price stability made by the author is based on a model where the central bank optimises its loss function subject to the restrictions implied by the structure of the economy. His analysis yields the result that the preferable monetary policy strategy is what he terms 'forecast targeting', since it is considered to be superior both to what he calls 'simple instrument rules' and 'intermediate targeting'. While I share the thrust of the author's basic conclusions, I would like to discuss his analysis from the standpoint of the considerations I made earlier on. Indeed, I believe that part of the superiority of 'forecast targeting' is much more due to easiness of communication within the central bank and also with the public than Svensson actually acknowledges. Concerning the comparison between instrument rules and forecast targeting, a crucial distinction is drawn between the `optimal' instrument rule and a `simple' instrument rule. I ®nd the arguments in the chapter against the commitment to a `simple' instrument rule fairly convincing. While it has been argued by some authors that simple rules may not be optimal in any given model but nevertheless can be superior in the sense of delivering reasonably good results across models, I agree with Svensson that, in general, simple instrument rules are too rigid to be valid in a real world monetary policy decision framework. The reason is that they prevent judgmental analysis and extra-model information from being properly taken into account, not to mention problems of incentive compatibility. Nevertheless, as pointed out in the chapter, there is also an `optimal' instrument rule, which is the solution to the general optimisation problem faced by the central bank. Theoretically, this is neither better nor worse than forecast targeting since both are alternative ways of characterising the optimal monetary policy. Consequently, provided that any preferred alternative to an optimal instrument rule must also ful®l the requirement of being a solution to the general optimisation problem, the reasons why the optimal instrument rule is not considered by the author to be appropriate merit careful attention. Interestingly, he disregards the optimal instrument rule because, on the one hand, it is very sensitive to the underlying model of the economy ± which in practice is not very precisely known ±, and, on the other, because it is overly complex as a policy guide. As concerns the ®rst reason, I wonder whether model uncertainty does not have the same unfavourable consequences for forecast targeting as for instrument rules, and whether such complications
110 Jose Vin Äals
could be taken into account in an analogous manner as with forecast targeting, the latter being explained in detail in the chapter. Insofar as this could be done ± and I think it could ±, then the only remaining argument against the optimal instrument rule would be its complexity and thus the dif®culty in fostering ready communication with the public. If, because of such reasons, the optimal instrument rule is discarded in favour of a simpler rule, then forecast targeting would be superior. The obvious reason is that the simpler rule will not be exploiting appropriately the available information, in addition to the other problems of incentive-compatibility mentioned earlier on. Regarding the comparison between forecast targeting and intermediate targeting, insofar as we restrict ourselves to the conceptual setting used by the author, it follows that only under very restrictive and well known conditions ± i.e. a recursive macroeconomic structure ± will intermediate targeting be as good as forecast targeting. In fact, even those central banks adhering formally to intermediate targeting ± like monetary targeting ± have pursued their strategies in a rather pragmatic manner, thus perhaps proving how unlikely it is that the conditions for pure intermediate targeting to be the 'best policy practice' are satis®ed in the real world. Overall, I think that Svensson's defence of forecast targeting as constituting the `best monetary policy practice' can be also rationalised in terms of the ®ve desirable features that such practice should contain and which I alluded to earlier in my discussion. In this regard, forecast targeting would yield a good mix between medium-term rigour and short-term ¯exibility; it would constitute an organised framework for systematic decision making inside the central bank which makes the best possible use of all the available information; it would be forward-looking and nonmechanistic in the sense of comparing forecasts across various models of the economy and using extra-model information to arrive at the best possible overall judgement on prospective price developments; and ®nally, it would make it easier for the public to understand monetary policy decisions by placing the focus of attention on the key variable, namely price developments. Indeed, any central banker knows that, in practice, monetary policy decisions are taken in a forward-looking manner, something that becomes necessary given the lags with which monetary policy impacts the economy and, in particular, prices. Therefore, decisions are made, explicitly or implicitly, on the basis of the best appraisal of the price outlook, something which is nothing else but Lars Svensson's `best conditional forecast'. But if so, where do the differences lie among monetary policy strategies? In my view, in the following three domains. First, in the extent to which the forecast is or not an explicit feature of the strategy, which serves to determine whether policy is formally conducted according to a forecast targeting scheme or not. Secondly, in the extent to which the ± explicit or implicit ± forecast is
Discussion 111
based on macroeconometric models or other less formal tools. And ®nally, in the extent to which the conditional forecast is published and thus serves as an explicit guide for policy action in the eyes of the public. To conclude, as Alan Greenspan has stated 'There is no way to avoid making a forecast, explicit or implicit, when making monetary policy'. The issue is whether, as Lars Svensson argues, forecast targeting is the best way for central banks to make use of the available information in the most ef®cient possible way to maintain price stability. The experience of those central banks which have adopted such a framework in the past, although such experience is still relatively short compared to other longer-standing strategies, suggests that forecast targeting is worth considering.
3
The Transmission Process1 Allan H. Meltzer
I must confess some vested interests in this topic. I ®rst discussed it in a paper with Karl Brunner, more than thirty-®ve years ago (Brunner and Meltzer, 1963). I have returned to the topic many times, most recently in a published symposium (Meltzer, 1995). I will refrain from reviewing these earlier studies, although I will refer in passing to some of the main ideas. I will concentrate on two topics. They do not exhaust the subject, but they raise issues that I believe are central. First, I raise some issues about a current class of models of monetary transmission in which a short-term interest rate represents the transmission process. The class of models is so widely accepted that the conclusions I challenge have become part of the canon. A different class of models ± a more useful one, I believe ± does more than give different answers. Some issues do not arise; they are no longer relevant. And some issues remain relevant but receive a different answer. The role of money is one such issue. Second, I discuss some of the evidence I have gathered from my study ± A History of the Federal Reserve ± the work that has been my main occupation for the past four years. The two pieces are related, as I hope to show. The evidence from history shows that the transmission process cannot be summarised by a single interest rate. In the ®nal section, I present some econometric evidence to supplement the historical data.
1
Standard models of monetary transmission
The core or central issue about the transmission of monetary policy is: How does a monetary impulse affect relative prices and real demands as it moves through the economy from its ®rst appearance to its ultimate effect on the main determinants of economic welfare? The standard answer to this question is, now as for some considerable time in the past, that monetary injections change an interest rate in the money market. Because some prices, money wages, or anticipations do not adjust instantly, there are effects on output and employment that, though temporary, may be large and costly. Eventually, 112
The Transmission Process 113
relative prices are restored, so the price level changes equi-proportionally. Money is neutral in the long run.2 Most of us accept large parts of this story. Real business cycle theorists would both add and omit, as would other economists. Loan or credit market models make other changes. I shall not dwell on these quali®cations, or the relevance of these analyses. Instead, I will use as illustrative a model that represents some of the best recent work on monetary policy to give speci®city where it is needed.3 Most of my comments apply more generally. The class of models has two equations, aggregate demand and supply. Aggregate supply is a Phillips-type curve. Important for current purposes is that monetary policy is represented only by the interest rate in the aggregate demand function. An expansive monetary action lowers the nominal interest rate; a contractive action raises it. The action is unanticipated, at least as to timing, so the real rate of interest changes with the nominal rate absolutely and relative to the anticipated real rate. In Svensson's (1998) model, that I have used as a good, clear example, aggregate demand is replaced by the output gap, and the aggregate supply function is adjusted accordingly. This difference is not consequential for my discussion. The story about transmission embedded in this model is that a change in interest rates changes aggregate demand in the opposite direction. If the interest rate increases, output falls, the output gap increases, so in¯ation declines, raising the real rate of interest and continuing the process of mutual adjustment of aggregate demand and supply. When the target rate of in¯ation is reached, the central bank rests, and the economy adjusts to the new equilibrium. The money stock and the demand for money remain in the background. The central bank sets the interest rate and provides the stock of money required to satisfy demand. The price level is determined with the rate of in¯ation and output, and so too is the demand for real money balances. This stripped-down model is useful for illustrating the interaction of aggregate demand and supply. As a model of monetary transmission, I believe it is de®cient for two main reasons. First, it has implications that are either misleading or wrong. Second, it omits signi®cant parts of the adjustment of relative prices and real wealth to monetary and other shocks and the responses of aggregate demand and output to changes in relative prices and real wealth. These omissions are major parts of the transmission process. Two false implications Many economists have used a model of this kind to discuss monetary policy in the 1930s depression and the long Japanese recession of the 1990s. In both cases, short-term nominal interest rates approached zero. The implication drawn from these experiences is that the monetary authority was powerless because the short-term interest rate was close to zero. The economy was said to be in a liquidity trap. (For the Japanese case, see Krugman, 1998, and Ito, 1998).
114 Allan H. Meltzer
A `liquidity' trap is de®ned as a condition in which the demand for nominal money balances equals the stock of nominal balances at unchanged prices and interest rates, for all values of the nominal stock. Additions to the nominal stock cannot be transmitted to the real sector or the price level because the interest rate has reached a ¯oor at or near zero. As is customary in this class of models, assume that all assets are gross substitutes in portfolios. Ignoring transaction and information costs, it does not matter which asset the central bank buys or sells. Whichever asset it buys, the market adjusts the relative prices of all assets to re¯ect the change in relative supplies. This is a powerful and useful argument with strong implications. For example, economists use it to show that sterilised foreign exchange market intervention has no effect or to show why unsterilised exchange rate intervention differs from sterilised intervention. Suppose, now, that with a short-term interest at zero, the Bank of Japan announces that it wants the dollar exchange rate to fall by 50 per cent and that it is prepared to print yen to buy dollars until that occurs. Is there any doubt that the yen would depreciate or that the depreciation would affect spending, output and prices in Japan? Suppose, instead, that the Bank of Japan makes no announcement but buys dollars with the intention of depreciating the yen by 50 per cent. There may be differences in the timing of responses, but the ultimate effect would be the same: monetary expansion would affect the economy. There would be no liquidity trap whatever the short-term interest rate in the market in which the central bank usually operates. Two questions occur. First, how can we reconcile our standard assumption of gross substitution with this obvious contradiction. Second, does this argument imply that a liquidity trap is impossible in a multi-asset world? The liquidity trap, by assumption, makes short-term Treasury bills (or similar securities) a perfect substitute for base money or bank reserves. Exchanging one for the other does nothing of interest. Exchanging either money or Treasury bills for some other asset such as foreign money, domestic or foreign long-term bonds, equities, or commodities changes relative prices and real wealth. In this hypothetical case, base money plus bills is a composite good. The composite good is a gross substitute for other assets; increasing either component, or both, is expansive. For a full liquidity trap to be effective, the composite asset ± money plus bills ± must be a perfect substitute for all other assets. When the marginal rate of substitution of money for bonds goes to zero, all marginal rates of substitution must go to zero.4 All assets are parts of a single composite good. If assets other than bills and money remain gross substitutes, a liquidity trap means only that one row and one column in the matrix of marginal rates of substitution has been eliminated. All other marginal rates of substitution remain. Monetary policy remains effective. The standard class of models gives the wrong answer about policy. It implies that a liquidity trap is possible and,
The Transmission Process 115
for some, is a reality, (Krugman, 1988, Ito, 1998). The alternative denies that a liquidity trap is possible except in the limit when all prices are zero. A closely related proposition has received considerable attention as in¯ation rates fell and remained low in the 1990s. Summers (1991) revived the argument that a zero in¯ation target is socially costly because it sets a lower bound for nominal interest rates. Monetary policy becomes weak or powerless; it cannot lower the short-term nominal rate or prevent falling prices from raising the real rate of interest. With money wages in¯exible downward, unemployment rises. Akerlof, Dickens and Perry (1996) perform the remarkable feat of ®nding evidence for this proposition using data for a period in which in¯ation never remained close to zero. And Benhabib, Schmitt-Grohe and Uribe (1998) argue that it is perilous to use a Taylor rule when in¯ation is near zero. A more sophisticated version of Summers' argument uses a stochastic model with non-linearity in the transmission process when in¯ation is below 2 per cent. Orphanides and Wieland (1998) ®nd that there is no evidence of an operative lower bound in US postwar data. They claim that the lower bound was in effect during the 1930s, so monetary policy was in¯exible for part of that decade. For this claim to be true, the short-term interest rate must be the principal or only means by which monetary actions are transmitted from the central bank, through the market, to the economy. As my old friend Karl Brunner often said: we know this is false. Monetary actions are effective and powerful in the less developed countries of Africa, Latin America, or Asia where there is no money market. Relative prices respond to monetary impulses in countries without central banks, and without money markets. There is more to the transmission process than the models recognise. Do these additional channels operate in countries with active money markets? In writing the History of the Federal Reserve, I have found three relevant examples from the years 1914 to 1951 that I have researched to date.
2 Historical evidence Each of the examples I consider concerns a period of de¯ation. Prices fell, raising real interest rates through most or all of the recession. Expansive ®scal actions in each episode were either modest or absent. Two of the recessions are considered severe according to rankings by the National Bureau of Economic Research. In each case the economy recovered, and two of the three recessions were of average length. The common feature that is relevant for the monetary transmission process is that real money balances and real interest rates rose together. In each case there was a common cause: prices fell. In some cases gold in¯ows or Federal Reserve actions increased the monetary base; in other cases, the monetary authorities were passive or restrictive through most or all of the recession.
116 Allan H. Meltzer
Differences of this kind are of secondary importance in the three examples (but not in the Great Depression). The dominant, common impulse in the three examples was de¯ation. Two of the three episodes share a second relevant feature: the interest rate on short-term Treasury bills was historically low. During the 1948±9 recession, rates on Treasury bills were about one per cent. In 1937±8, bill rates were close to zero. In the third case, 1920±1, short-term rates remained well above zero, but the de¯ation was sharp and severe, so real interest rates and real money balances rose together. 1937±8 The National Bureau ranks the 1937±8 recession as the third most severe recession in the years after the First World War. Real GNP fell 18 per cent and industrial production 32 per cent in the thirteen months from May 1937 to June 1938. Unemployment reached a peak of 20 per cent, not very different from the 25 per cent peak in 1932. The probable causes of the recession include both ®scal and monetary actions. There is a very large reduction in the government de®cit in 1937 and a very large reduction in growth of the monetary base. The main ®scal actions are the end of the soldiers' bonus payment, the enactment of an excess pro®ts tax to pay for part of the bonuses in ®scal 1937, and the start of social security tax collections in ®scal 1936. The soldiers' bonus is the largest item, $1.7 billion of current spending. It was paid in June 1936, in time for the election later that year. The bonus was paid in bonds, but the bonds could be sold for cash. By December 1936, $1.4 billion had been cashed. Balke and Gordon's (1996) quarterly data show an 18 per cent average rate of increase in real GNP for the last three quarters of 1936. The most important monetary actions are the beginning of gold sterilisation at the end of 1936 and the second and third increase in reserve requirement ratios in March and May 1937. These increases completed the doubling of reserve requirement ratios between August 1936 and May 1937. During the entire period December 1936 to December 1938 that brackets the recession, interest rates on Treasury bills remained between 0.03 per cent and 0.56 per cent. Long-term nominal rates on Treasury bonds were modestly higher during the recession than before or after, but the difference is small; the range is 2.55 per cent ± 2.83 per cent. Annualised monthly rates of price change are consistently negative from October 1937 to February 1938 and intermittently negative for the rest of 1938. To smooth the data, I used moving twelve month averages of rates of price change. Figure 3.1 compares the real interest rate to the annual growth of the monetary base. The common element in the two series is the twelve-month moving average of the rate of price change. The divergence between the two series re¯ects some release of sterilised gold into the monetary base in September 1937 and a small
per cent
+7.5
+25
+6.0
+20
+4.5
+15
+3.0 +1.5
+5
0
0
–1.5
–5
–3.0
–10
Year-over-year real base growth
–4.5
–15
6.0 –6.0
–20
Q1
Q2
Q3 1936
Q4
Q1
Q2
Q3 1937
Q4
Q1
Q2
Q3
Q4
1938
Year-over-year real base growth versus real long-term interest rate, January 1936±December 1938
117
Figure 3.1
+10
Real long-term interest rate
Year-over-year real base growth
Real long-term interest rate
per cent
118 Allan H. Meltzer
volume ($38 million) of open market purchases in November, principally for seasonal reasons. Not until February 1938, after nine months of a deep recession, did the Federal Reserve propose countercyclical action, the release of additional gold from sterilisation. In April, the Roosevelt Administration announced $2 billion of additional government spending for construction and relief. As part of this programme, the Treasury released another $1.4 billion from sterilisation and the Federal Reserve released $750 million of reserves by lowering reserve requirement ratios. Figure 3.1 shows the sustained rapid increase in the real value of the monetary base beginning in February 1938. Real ®nal sales rose in the following quarter, but inventories fell, so real GNP did not increase until the third quarter. What does this episode suggest about the transmission of monetary policy? In the months preceding recovery, and in the early months of expansion, the real interest rate rose from 2.9 per cent in January 1938 to more than 6 per cent in September±November 1938. Although nominal rates remained historically low, real rates were relatively high. In contrast, real money balances accelerate ®ve months before the end of the recession: between February and June, growth of real balances rises from ±7.6 per cent to 17.6 per cent. By the end of 1938, growth of real balances reached an almost 25 per cent annual rate. I draw three conclusions for the transmission process. First, low nominal interest rates misled the Federal Reserve, on this occasion as on others, into believing that monetary policy was expansive. Second, although short-term interest rates stayed at or near zero, monetary policy was not powerless. Desterilising gold to increase the monetary base raised nominal and real money balances and increased spending. Third, the ®nancial system was not in a liquidity trap. Channels other than the short-term interest rate transmitted monetary expansion to output and the price level. 1948±9 The 1948±9 recession provides a second example refuting the liquidity trap and the small or vanishing effect of monetary policy at low nominal interest rates. The Federal Reserve pegged nominal long-term interest rates below the 2.5 per cent ceiling in effect from 1942 to 1951. Despite the pegging policy, the monetary base fell through most of 1948. The principal reason is that the Treasury used its budget surplus to retire debt held by the Reserve banks. The monetary base fell as a consequence of the Treasury's actions. Although the Federal Reserve complained about being an engine of in¯ation, prices fell in half the months of 1948 and 1949. The National Bureau dates the end of the expansion in November 1948 and the recession trough in October 1949. The twelve-month moving average rate of in¯ation fell from above 9 per cent in June and July 1948 to negative values in May 1949. It remained negative for the rest of that year.
The Transmission Process 119
During most of the recession the Federal Reserve was more concerned about a return of in¯ation than about the recession. The nominal rate on Treasury bills remained between 1.02 per cent and 1.17 per cent throughout the recession. Figure 3.2 compares annual growth of the real monetary base to the real interest rate in the two years that include the recession. Data are computed as in Figure 3.1. As before, the high positive correlation re¯ects the common effect of the rate of price change on the two series. The high correlation and parallel movement show that until late in 1949, when the recession was almost over, the Federal Reserve took few actions to increase base growth. Real base growth fell to ±11 per cent in September 1948, two months before the cyclical peak. Thereafter base growth rose, but did not become positive until April 1949, six months before the trough. The peak rate of base growth is close to 6 per cent in August 1949, two months before the end of the recession. At that time the real long-term interest rate was above 5 per cent. Once again, the movement of real base growth is consistent with the beginning and end of recession; the movement of real interest rates is not. Once again, low nominal short-term interest rates do not appear to have weakened the effects of monetary policy. And once again there appears to be more to the transmission process than is contained in standard models. 1920±1 The third episode is the recession from January 1920 to July 1921.5 The National Bureau ends the expansion in January 1920 and puts the last month of recession in June 1921. The Federal Reserve undertook larger policy actions, so nominal interest rates and nominal base growth re¯ect these actions. In¯ationary policies in much of Europe and restrictive policies in the United States brought an in¯ow of gold. The base and interest rate changes re¯ect these in¯uences also. Nevertheless, real base growth and real interest rates are positively correlated during the recession. Both are negative at the start of the recession, turn positive about a year later and reach a peak at the end of the recession. Judged by base growth, monetary actions are countercyclical in the ®rst half of 1921; judged by real interest rates, these actions are procyclical. Figure 3.3 shows these data. The long-term nominal rate remains within a narrow range but is higher at the trough of the recession than at the previous peak. The dominant in¯uence on real rates and real base growth during the recession is the decline in in¯ation followed by de¯ation. As in the previous two episodes, interest rates give a misleading signal about the thrust of policy. Real base growth gives a more correct signal. In this recession, the de¯ation is severe; the peak annualised rate reached 17 per cent, and it was above 10 per cent for ten consecutive months. The real long-term interest rate, (i ± )/(1 ), is above 25 per cent at the end of the recession. The economy recovered despite, not because of, the level of real interest rates.
120
per cent per cent +4.5
Real long-term interest rate
+3.0
+4
Real long-term interest rate
+1.5
+2
0
0
–1.5
–2
–3.0
–4
–4.5
–6 Twelve-month moving avarage real base growth Twelve month moving avarage real base growth
–6.0
–8
–7.5
–10
–9.0
–12
Q1
Q2
Q3 1948
Q4
Q1
Q2
Q3
Q4
1949
Figure 3.2 Twelve-month moving average real base growth versus real long-term interest rate, December 1947±December 1949
Twelve month moving avarage real base growth Twelve-month
+6
per cent
per cent
Real long-term interest rate +10
+15
+5
0
0
–15
Year-over-year real base growth
–5
–30
–10
Q1
Q2 1920
Q3
Q4
Q1
Q2 1921
Q3
Q4
Q1
Q2
Q3
Q4
1922
Year-over-year real base growth versus real long-term interest rate, January 1920±December 1922
121
Figure 3.3
Year- over- year real base growth
Real long-term interest rate
+30
122 Allan H. Meltzer
The three historical periods raise doubts about the central role assigned to interest rates in the transmission process. They suggest an important role for real balances. I return to these issues below. The Great Depression The only other period of US de¯ation after 1914 is during 1929±33, the Great Depression. The real interest rate rose from 5 per cent to 15 per cent and remained near 15 per cent through the last two years of the recession. Real base growth rose once bank runs began late in 1930, but this is, of course, misleading. As Figure 3.4 shows, growth of real balances ± measured here by M1 ± is very different in this period than in other de¯ationary periods. A principal difference is that monetary contraction was strong enough to offset the effects of de¯ation on real balances. An additional problem Although the central bank's objective function is not directly part of the transmission process, the three US recessions considered here raise doubts about the choice of objective function used in current analyses. The usual function has two arguments: (1) the difference between the in¯ation target and the actual or expected in¯ation rate and (2) the output gap, the difference between actual output and output consistent with the natural rate. Yet, we have seen that faced with de¯ation and a deep recession (in three of the four episodes) the Federal Reserve was slow to act. And it did not act effectively to end de¯ation and recession. The objective function fails in these cases as a positive statement of central bank objectives.6 My concerns about the standard objective function are not limited to its empirical support, important as that is. I believe that using the output gap as an objective of the central bank is problematic. This gap can arise for reasons unrelated to monetary policy actions ± for example, an oil shock, reductions in employment and output in the European Union resulting from provisions of the welfare state, or other real events. My colleague Bennett McCallum suggests that the problem can be overcome by rede®ning the natural rate to take account of non-monetary effects (Chapter 1 in this volume). In principle, this can be done; in practice it is dif®cult to do accurately. The natural rate is not like the gravity constant. We neither measure it precisely nor agree on its value. Opening the objective function to the pull and tug of opinions about the size of the gap carries a risk to economic stability and the apolitical position of an independent central bank. Recent discussion in Europe and the United States shows that there can be differences of opinion about measurement of both potential output and the natural rate of unemployment. A second problem with the now standard objective function is that it neglects several issues of concern to central bankers. There is no role for a lender of last resort. A run to currency, bank failures, a drain of foreign
per cent
per cent Real interest rate
+15
+15
+10
+10
+5
+5
0
0
–5
Real M1 growth
–5
–10
–10
Q3
Q4 1929
Q1
Q2
Q3
1930
Q4
Q1
Q2 1931
Q4
Q1
Q2
Q3
1932
Q4
Q1 1933
123
Figure 3.4 Real M1 growth and real interest rate, August 1929±March 1993
Q3
124 Allan H. Meltzer
exchange reserves, a major change in the exchange rate, or a so-called `credit crunch' enters the bank's objective function only by changing the output gap or in¯ation. This misstates what central banks do. Properly functioning central banks respond to the increased demand for money before output and prices fall. Of®cials of the new European Central Bank (ECB) emphasise that in¯ation is their sole objective. Their insistence on this point does not, and should not, preclude concern for the cost of achieving the zero or target rate of in¯ation. This cost should appear in the objective function, so that the objective is not just to achieve zero in¯ation but to do so at minimum social cost. With cost replacing the output gap in the objective function, the way is open to treat other issues of concern. The implicit cost function used by central banks includes costs of maintaining or restoring ®nancial safety or solvency, avoiding a credit crunch, or increasing unemployment. In practice, even central banks most concerned about in¯ation do not ignore the social cost of reducing in¯ation. They proceed gradually, over a period of months or years. And they, properly, do not ignore ®nancial fragility, ®nancial failures and lender of last resort responsibilities when making decisions. In fact, the opposite seems to be true: some central banks have recently been overly sensitive to ®nancial fragility. The Federal Reserve's decisions to reduce interest rates three times in the autumn of 1998 suggests that ®nancial fragility, or its prospective costs, in¯uences what central banks do. Since these changes were made at a time of high employment and rapid money growth, the experience suggests that, at times, fragility may enter lexicographically in the central bank objective function. The size and direction of change in real money balances may be useful measures of some of the costs of systemic ®nancial fragility.
3
Econometric evidence and interpretation
The three historical episodes discussed earlier suggest that an interest rate does not fully represent the monetary transmission process. Changes in real money balances appear at times to dominate changes in real interest rates as indicators of the direction of change induced by monetary actions. One interpretation of this evidence follows from the Haberler±Pigou± Patinkin wealth effect. Falling prices raise private sector real money balances, increasing real wealth. Some of the increase is spent on consumption. A second interpretation views the change in real balances as a mixture of a pure wealth effect and the changes in wealth induced by changes in the relative prices of assets and output (Brunner and Meltzer 1963, 1968). On either interpretation, we lose the simplicity of the two-equation, aggregate demand± aggregate supply, class of models. If real balances affect aggregate demand, the demand for money and perhaps other parts of the ®nancial structure become relevant for the transmission process. The single interest rate is no longer suf®cient to represent monetary policy and ®nancial markets.
The Transmission Process 125
Let me turn the issue around. The requirements of the standard model are strong and are, as we have seen, easily falsi®ed empirically using historical episodes. Analytically, the model is very demanding. Not only must the wealth effect remain small and inconsequential for spending, but a single interest rate must represent all relative price adjustment. An appeal to rational expectations puts additional burdens on the model. Market makers and participants must rationally anticipate how much output prices will change to restore equilibrium at the natural rate. One data point, one observation on the interest rate, is not suf®cient to determine whether the change is permanent or temporary, a change in level or in growth rate, or an error that will be reversed. Even if one believes that adjustment is rapid in markets for actively traded assets, many assets are traded infrequently or not at all. The transmission of monetary policy involves adjustment of prices of existing hotels to newly produced hotels, existing plants and equipment to newly constructed plants and equipment, new cars to used cars, new houses to existing houses ± and, of course, money wages to prices. I do not propose adding equations for each of these variables. That would take us toward large-scale econometric models, a technology that has not contributed greatly to our understanding of monetary policy. But there is also considerable evidence that the expectations theory of the term structure or interest arbitrage theories of the exchange rate are far from adequate. This latter evidence suggests, again, that a single interest rate does not summarise adequately responses on asset and output markets. Recent analyses of the credit channel exploit differences in information to develop an additional transmission channel for monetary actions (Bernanke and Gertler, 1995). The credit channel operates parallel to the interest rate (or monetary) channel, and the two interact. Empirical support for this channel has not been persuasive; the hypothesised effect is probably swamped by the endogenous response of bank loans to the monetary base and other determinants of aggregate demand. The use of costly information to separate parts of the transmission process is a step forward, however. It may ®nd a better use distinguishing markets for bonds and real capital or distinguishing different types of capital. How well do real money balances capture the many channels of monetary transmission? Koenig (1990) tested a two-stage model of changes in consumption with changes in real money balances, real interest rates, income and other variables as arguments of the consumption function. Consumption includes only purchases of non-durables and services. The interest rate is adjusted for in¯ation. All variables other than the interest rate and a dummy variable are in logarithms. The dummy variable takes a value of unity in 1980 and after to represent ®nancial deregulation. All economic variables are used as ®rst differences. Using data for the period from 1951:2 through 1986:1, Koenig found that changes in real money balances had a signi®cant positive effect on changes in
126 Allan H. Meltzer
consumption. To test his ®nding, Koenig introduced lagged values of changes in durable goods, income, consumption, and stock prices sequentially as additional explanatory variables. None was signi®cant at the 5 per cent level; in all cases the change in real money balances remained positive and signi®cant.7 Furthermore, he ®nds evidence of a relatively strong real balance effect. A 10 per cent change in real balances, ceteris paribus, results in a 2.5±3 per cent change in spending on consumption. This result is in addition to any effect of real balances on the real interest rate and, thus, indirectly on consumption. Koenig found the latter effect relatively small, however. He used a short-term interest rate. In the three historical episodes discussed earlier, real balances rose in each recession. In two of the recessions, the increase was more than 13.5 per cent from peak to trough. In 1948±9, the increase was less than 1 per cent from peak to trough but three±four times larger if measured one month before the trough. If Koenig's estimates apply to these out of sample cycles, they imply a 3.5±4 per cent increase in consumption spending in the two earlier recessions. This result is economically meaningful. To test this hypothesis, I attempted ®rst to replicate Koenig's results. Then I extended his sample period by 30 per cent, an addition of forty-two quarterly observations. Most of the attempts at replication supported his hypothesis. However, the change in the real interest rate was usually signi®cant and positive, contrary to the hypothesis. Koenig presents several different two-stage least square estimates. I was able to replicate many of his ®ndings. Table 3.1 reports only a small part of the results, but many are of interest principally because the main results are robust to changes in the arguments included in the estimation. All results are twostage least squares estimates. Table 3.1 Response of consumption to real balances and other variables, 1951:2±1986:1 Koenig (1990) Constant Dummy r ln m ln c(t ± 1) R2 DW
0.0058* (0.0005) ±0.0048* (0.0012) 0.1948 (0.1247) 0.2851* (0.0588) 0.37 1.94
Notes: Standard errors in parentheses. *Signi®cant at the 5 per cent level.
0.0056* (0.0006) ±0.0047* (0.0012) 0.1949 (0.1235) 0.2738* (0.0624) 0.0307 (0.0754) 0.37 1.99
Replication 0.0064* (0.0004) ±0.0056* (0.0011) 0.4490* (0.1740) 0.3046* (0.0851) 0.37 1.99
0.0061* (0.0007) ±0.0055* (0.0012) 0.4458* (0.1738) 0.2907* (0.0910) 0.0258 (0.0778) 0.37 2.02
The Transmission Process 127 Table 3.2
Response of consumption to real balances: extended sample, 1950:4±1995:4 m Real M1
Constant Dummy r ln m ln c(t ± 1) R2 DW
0.0064* (0.0005) ±0.0044* (0.0009) 0.3212* (0.1293) 0.2688* (0.0583) 0.13 1.91
m Monetary base 0.0060* (0.0007) ±0.0042* (0.0010) 0.3090* (0.1333) 0.2542* (0.0590) 0.0419 (0.0747) 0.15 1.99
0.0060* (0.0005) ±0.0049* (0.0010) 0.3105* (0.1369) 0.3193* (0.0925) 0.14 1.98
Notes:
Standard errors in parentheses.
* Signi®cant at the 5 per cent level.
Table 3.2 shows some results for the extended sample. The ®rst two columns use changes in real M1 balances, as in Koenig's work. The last two columns replace M1 with the St Louis monetary base, the variable I used in the historical data after 1935. Other estimates using lagged values of money, real income, stock prices and other variables give similar results. The only change that made a substantial difference was omission of the dummy variable used to shift the intercept after 1980. These ®ndings suggest, again, that the transmission of monetary impulses involves more than is found in the standard class of models. Changes in real money balances, operating as a wealth effect or as a proxy for changes in both relative prices and real wealth, have positive and signi®cant effects on the change in consumption.
4 Conclusion Viewed one way, the point of this chapter is an old one, one that is well-known as the Haberler±Pigou±Patinkin real wealth effect. The conclusion in that case would be that the wealth effect is more relevant in a dynamic context than current, standard analysis recognises. An alternative interpretation is that the dynamic real wealth effect includes more than the direct effect of changes in real money balances. If anticipated returns and anticipated in¯ation affect prices of assets with low transaction costs more rapidly than the prices of new production, many changes in relative prices of assets to output are part of the transmission process. As home prices change relative to foreign prices, the real exchange rate changes. This
128 Allan H. Meltzer
change is supplemented by a change in the nominal exchange rate in a ¯uctuating rate regime. The standard model includes a type of Phillips curve, so prices do not adjust instantly to changes in money. Hence real balances change cyclically. The consumption equation in the text suggests that the effects of monetary change are not fully expressed by a change in a single short-term interest rate. A study of forecast performance in Germany suggests that this ®nding is not limited to the United States (Kirchga ssner and Savioz, 1998). Rather than rely on a single regression equation, I have used data from periods of de¯ation to show that changes in real interest rates cannot explain some major episodes in monetary history. In 1920±1, nominal interest rates were higher at the trough than at the preceding peak. Prices fell sharply, so real interest rates were higher still. Postwar ®scal policy was contractive. Why did the economy recover? Falling prices raised existing real wealth and increased the gold ¯ow to the United States. Real money balances rose rapidly. The same process worked to end the 1937±8 recession and, more modestly, the 1948±9 recession. Even in the 1929±33 depression, real balances played a role. This time, the money stock fell in most periods by more than the price level, so the real wealth effect did not sustain expansion. Evidence from periods of de¯ation is useful because de¯ation raises both real interest rates and real money balances. Positive changes in interest rates and money have different implications: one points to expansion and an end to recession, the other to continued, possibly deeper recession. In periods of in¯ation, real interest rates are often low in recession, while real money balances may fall. Monetary expansion reduces interest rates and raises real balances. Both transmission processes are at work, so it is more dif®cult to reliably separate their relative importance. Econometric work of the past quarter-century testi®es to the dif®culty of distinguishing the two effects when prices are rising. Secular and permanent changes in real balances supplement the cyclical changes that I have stressed. The introduction of the Euro is one such permanent change. The breakdown of Bretton Woods, ®nancial intermediation and other long-term changes affected relative demands for real balances. Such changes have persistent effects that, at times, dominate short-term changes. Currency crises also are transmitted through changes in demand for real balances as well as through interest rates and exchange rates. The current standard model neglects these secular changes in the demand for real balances. Central bankers choose to conduct their operations by changing an interest rate, usually a short-term rate. This need not be harmful. The central bank must adjust the interest rate enough to prevent de¯ation and in¯ation, having regard to the cost of maintaining zero expected in¯ation. Whether making a discretionary judgment or following a rule, history suggests that the bank will avoid large, costly errors if it does not ignore the role of changes in nominal money and in real balances when making its decisions.
The Transmission Process 129
Notes * This chapter is a revised version of a dinner speech given by Allan Meltzer at the conference. 1. My thanks to Randolph Stempski for his excellent assistance and to Bennett McCallum for helpful discussion. 2. A similar but more complicated story is told about changes in money growth. Issues about super neutrality are put aside. 3. I base my discussion on Svensson (1998). Other references could be used ± for example, Taylor (1993, 1999). 4. A more complicated proof of this argument is in Brunner and Meltzer (1968). 5. For this period, the monetary base is high-powered money from Friedman and Schwartz (1963). The price index is not seasonally adjusted. 6. Some may object that studies of Taylor's (1993) rule support use of the objective function as currently relevant. Orphanides (1998) argues that this evidence is much less compelling if we restrict the Federal Reserve to data available at decision time. 7. De®nition of variables is in Koenig (1990, pp. 422±4). We have used his de®nitions wherever possible, but we are uncertain about his measurement of interest rates ± end of quarter, quarterly average, etc. Interest rates are after tax rates, where tax rates are marginal rates from Barro and Sahasakul (1983). We assumed constant marginal tax rates after 1980.
References Akerlof, G. A. Dickens W. T. and G. L. Perry (1996) `The Macroeconomics of Low In¯ation', Brookings Paper on Economic Activity, 1, 1±76. Balke, Nathan S. and R.J. Gordon (1996). `Appendix B: Historical Data' in R.J. Gordon (ed.), The American Business Cycle: Continuity and Change. Chicago: University of Chicago Press for the National Bureau of Economic Research, 789±842. Barro, R. J. and C. Sahasakul (1983) `Measuring the Average Marginal Tax Rates from the Individual Income Tax', Journal of Business, 56, 419±52. Benhabib, J., S. Schmitt-Grohe and M. Uribe (1998) `The Perils of Taylor Rules', C. W. Starr Center for Applied Economics, New York University, November, unpublished. Bernanke, B. S. and M. Gertler (1995) `Inside the Black Box: The Credit Channel of Monetary Policy Transmission', Journal of Economic Perspectives, 9, 27±48. Brunner, K. and A. H. Meltzer (1963) `The Place of Financial Intermediaries in the Transmission of Monetary Policy', American Economic Review, 53, 372±82. ÐÐÐÐ (1968) `Liquidity Traps for Money, Bank Credit and Interest Rates', Journal of Political Economy, 76, 1±37. Friedman M. and A. Schwartz (1963) A Monetary History of the United States, 1867±1960, Princeton: Princeton University Press for the National Bureau of Economic Research. Ito, T. (1998) `Japan and the Asian Financial Crisis: The Role of Financial Supervision in Restoring Growth', paper presented at a conference on the `The Japanese Financial System', New York, Columbia University, 1±3 October. Kirchga ssner, G. and M. Savioz (1998) `Monetary Policy and Forecasts for Real GDP Growth: An Empirical Investigation for the Federal Republic of Germany', University of St Gallen, unpublished. Koenig, E. F. (1990) `Real Money Balances and the Timing of Consumption', Quarterly Journal of Economics, May, 399±425. Krugman, P. (1998) `Japan's Trap', Krugman web page, MIT, May. Meltzer, A. H. (1995) `Monetary, Credit, and (other) Transmission Processes', Journal of Economic Perspectives, 9, 49±73.
130 Allan H. Meltzer Orphanides, A. (1998) `Monetary Policy Evaluation with Noisy Information', Board of Governors of the Federal Reserve System, Finance and Discussion Series, 1998±50, unpublished. Orphanides, A. and V. Wieland (1998) `Price Stability and Monetary Policy Effectiveness when Nominal Interest Rates are Bounded at Zero', Board of Governors, Finance and Discussion Series, 1998±35, unpublished. Summers, L. (1991) `How Should Long-Term Monetary Policy Be Determined?', Journal of Money, Credit and Banking, 23, 625±31. Svensson, L. (1998) `Monetary Policy Issues for the ESCB', paper presented at the Carnegie ± Rochester Conference, November. Taylor, J. B. (1993) `Discretion versus Policy Rules in Practice', Carnegie ± Rochester Conference Series on Public Policy, 39, 195±214. ÐÐÐÐ (1999) Monetary Policy Rules, Chicago: University of Chicago Press.
4
Asymmetric Interest Rate Policy in Europe: Causes and Consequences Axel A. Weber
1 Introduction The recent theoretical and empirical literature on monetary policy rules has increasingly focused on short-term interest rates rather than monetary aggregates for studying European monetary policy issues. There are several reasons for this: ®rst, as in the United States, monetary aggregates in Europe have displayed a less obvious link to real economic activity and in¯ation during the 1980s and 1990s as opposed to the 1960s and 1970s. Second, many central banks have de-emphasised the role of monetary aggregates and have moved to operating procedures that focus more on interest rates (i.e. the Fed funds target rate in the United States) or in¯ation rates (i.e. the in¯ation targets in the United Kingdom, Canada, or New Zealand). Following the paper by Taylor (1993) and more recent applications by Clarida and Gertler (1997), Clarida, Gali and Gertler (1997, 1998), Gerlach and Smets (1998), Kuttner and Posen (1998) and Rudebusch and Svensson (1998) there is now a growing literature on so-called `interest rate smoothing' rules for Europe.1 These papers use a simple policy reaction function in which interest rate adjustment towards equilibrium depends on the deviations of in¯ation and output from their respective target values. It is shown that such policy reaction functions ®t the data quite well. The present chapter takes a critical look at this new literature and discusses two major problems. The ®rst of these problems is related to the interpretation of such policy rules. It is argued that it is impossible to disentangle the causes and consequences of policy actions in any policy reaction function, be it a monetary policy rule or an interest rate policy rule, without imposing serious prior restrictions. Unfortunately, such restrictions are typically dif®cult to justify and to test. Think of the Clarida, Gali and Gertler (1998) paper, which uses the level of short-term interest rates as the policy instrument and the twelve-month ahead expected ( actual) in¯ation rate as well as the output gap as the exogenous variables. If a restrictive monetary policy of increasing interest rates in response to expected in¯ation is effective in easing 131
132 Axel A. Weber
in¯ationary pressure, then future actual in¯ation and hence in¯ation expectations will decline as a consequence of the policy. Clarida, Gali and Gertler (1998) realise this, since they state that `the expectations [of in¯ation] on the right hand side [should] be based on all relevant information available to policymakers ..., which includes the current interest rate itself, since expected in¯ation and output will not be invariant to it'. They take account of this endogeneity by using an instrumental variable2 technique. But singleequation models can only indirectly capture such endogeneity, and the risk of a serious bias of the estimated coef®cients remains. As an alternative we propose to use the bivariate VAR model developed by King and Watson (1997) in order to identify the in¯ationary causes and consequences of interest rate smoothing policies more directly. In particular, we take a closer look at the sensitivity of the estimated in¯ation response coef®cient ( ) with respect to the exogeneity assumption.3 This is of key importance since the magnitude of the in¯ation response coef®cient provides an important yardstick for the evaluation of such policy rules: if > 1, the target real rate adjusts to stabilise in¯ation, whilst for < 1 it instead moves to accommodate changes in in¯ation. The long-run Fisher effect is the critical threshold. As Clarida, Gali and Gertler (1998) state, the central bank in this latter case raises the nominal rate in response to an expected rise in in¯ation, but it does not do it suf®ciently to keep the real rate from declining. Bernanke and Woodford (1996) and Clarida, Gali and Gertler (1998) have shown that in such an accommodative regime self-ful®lling bursts of in¯ation may be possible. There is a second problem with interest rate smoothing rules which will be brie¯y addressed here. We argue that it is impossible to discuss European interest rate policy rules in isolation. Even if the short-run (or high frequency) crosscountry correlations in interest rates are small, there are clearly common longrun factors behind European interest rate movements. In a dynamic bivariate VAR model these long-run correlations are identi®able, and it will be possible to distinguish between long-run and short-run correlations. It will be shown below that the long-run cross-country correlations are high throughout the ERM period and have increased over time owing to nominal convergence. This ®nding is already obvious from Figures 4.1a±4.1d, which display the evolution of shortterm money market rates, discount rates, long-term government bond rates and in¯ation rates for Germany and a number of selected European countries. In the pre-1983 sample interest rates and in¯ation rates show little co-movement, but during the 1990s there is an increasing synchronisation of European interest and in¯ation rates owing to the convergence pressure from the Maastricht treaty. Nominal convergence is almost complete in late 1998 shortly before the move to EMU. Such cross-country co-movements of European interest and in¯ation rates have largely been ignored when it comes to estimating policy reaction functions. If the ERM was really an asymmetric exchange rate arrangement in which Germany has played the role of the anchor country, then estimating countryspeci®c policy reaction functions is of interest only for Germany. The policy rules
(a) Short-term money market interest rates
18
20
Asymmetric Interest Rate Policy In Europe 133
14
16
France
12
France
10
a
4
6
8
Belgium
Germany
0
2
Netherlands
75
79
(b) Discount rates
83
87
91
95
99
(b) Discount rates
12
14
16
1971
10
Belgium
6
8
b
4
Netherlands
Austria Belgium
2
Germany
1971
75
79
83
87
Figure 4.1 Interest rates and in¯ation, 1972:1±1998:12 (a) Short-term money market interest rates (b) Discount rates
91
95
99
134 Axel A. Weber
24
(c) Long-term government bond interest rates
20
Italy
c 16
France
12
Netherlands Italy
8
France Germany
4
Germany
1971
75
79
83
87
91
95
99
22
26
Inflationtrates rates (d) Inflation
France
10
d
14
18
Italy
6
Netherlands
2
Italy
–2
Germany
1971
75
79
83
Figure 4.1 continued (c) Long-term government bond interest rates (d) In¯ation rates
87
91
95
99
Asymmetric Interest Rate Policy In Europe 135
for France, Italy and the remaining ERM countries have to be speci®ed differently and this is why Clarida, Gali and Gertler (1998) include the German interest rate as an important determinant of interest rates in the other ERM countries. But provided that German interest rate movements can be well explained in terms of a Bundesbank policy reaction function with respect to German in¯ation and German output growth, these two factors will also explain some proportion of interest rate movements in France. Figures 4.1a±4.1c show that German and French interest rates are correlated, and Figures 4.1d±4.2a show that the same is true for both in¯ation and output growth, but less so for money growth in Figure 4.2b. From the data's point of view it therefore seems to be unclear whether interest rate smoothing rules in the ERM member countries like France are best estimated on domestic or German data or any combination of the two. This is why we aim at studying the degree of interest rate interaction between European countries rather than looking at these countries in isolation. The remainder of the chapter is organised as follows: Section 2 brie¯y surveys the existing evidence about interest rate smoothing rules for Europe while Section 3 deals with the ability of interest rate smoothing policies to achieve the ®nal objective of monetary policy, which is low in¯ation. To get a handle on this we use a bivariate VAR with in¯ation and interest rates. Orphanides (1999) refers to this model as a simple prudent policy rule. We study the short-run and long-run interaction of both variables in this prudent rule under the very ¯exible identi®cation scheme proposed by King and Watson (1997). Section 4 deals with the interaction of German and other European interest rates, and we will examine the degree of long-run and shortrun interest rate co-movements. Section 5 concludes.
2 Empirical evidence about interest rates smoothing rules for Europe In this section we brie¯y review some of the existing evidence regarding interest rate smoothing rules. It is important to note that such rules may be viewed as a generalisation of the simple interest rate rule proposed by Taylor (1993), which received widespread attention. In the typical Taylor rule the central bank responds with the nominal interest rate
it in a prescribed manner to some known target deviations of in¯ation
t � and output
yt � y : it r t 0:5
yt � y 0:5
t � "t
4:1
Gathering all constant terms in (4.1) results in the typical Taylor rule in which nominal interest rates
it respond to output
yt and in¯ation
t : it r � 0:5
y 0:5yt 1:5t "t |{z} r~
4:2
16
136 Axel A. Weber
(a) Output (industrial production) growth in Germany and France
12
Germany
–4
0
a
4
8
Germany
–8
France
1971
75
79
83
87
91
95
99
(b) Money (M1) growth in Germany and France
14
18
22
–16
–12
France
–6
–2
2
6
10
b
1971
75
79
83
87
91
Figure 4.2 Output and money growth, 1972:1±1998:12 (a) Output (industrial production) growth in Germany and France (b) Money (M1) growth in Germany and France
95
99
Asymmetric Interest Rate Policy In Europe 137
whereby the in¯ation response coef®cient has a numerical value of 1.5, and hence real interest rates rise in response to in¯ation. Clarida, Gali and Gertler (1997, 1998) and Rudebusch and Svensson (1998) introduce a partial adjustment mechanism for interest rates owing to adjustment costs and derive the dynamic interest rate smoothing rule: it
1 �
1 � t
1 � yt it�1 "t
4:3
whereby 0 < < 1 holds for the partial adjustment coef®cient. The estimates obtained by these authors are reproduced in Table 4.1. The key point about Table 4.1 is that the long-run in¯ation response coef®cient is larger than unity for Germany, Japan and the United States, but not for the United Kingdom, France and Italy. In Clarida, Gali and Gertler (1998) the same result also applies in many of the auxiliary regressions which include other potential policy target variables. Rudebusch and Svensson (1998) use a modi®ed interest rate smoothing rule in their optimal policy simulations, which are based on: it 0:88t 0:30t�1 0:38t�2 0:13t�3 1:30yt 0:33yt�1 0:47it�1 0:06it�2 0:03it�3 "t
4:4
In contrast to the simple Taylor rule their optimal rule involves a long-run in¯ation coef®cient of larger than 2, whilst the long-run output coef®cient in the optimal rule is greater than 1.4 Kuttner and Posen (1998), on the other hand, estimate the interest rate rule in ®rst differences. One of the rules they estimate is the simple prudent rule: it1 it1 � it c
t � t�1 "t
4:5
Table 4.1 Selected coef®cient estimates of Clarida, Gali and Gertler (1998), baseline regressions Country
Period
In¯ation
Output
Lagged interest rate Constant
AR (p) model
Germany Japan USA USA UK France Italy
79.03±93.12 79.04±94.12 79.10±94.12 82.10±94.12 79.06±90.10 83.05±89.12 81.06±89.12
1.31 (0.09) 2.04 (0.19) 1.79 (0.18) 1.83 (0.45) 0.98 (0.09) 1.13 (0.07) 0.90 (0.04)
0.25 (0.04) 0.08 (0.03) 0.07 (0.06) 0.56 (0.16) 0.19 (0.04) 0.88 (0.10) 0.22 (0.08)
0.91 (0.01) 0.93 (0.01) 0.92 (0.03) 0.97 (0.03) 0.92 (0.01) 0.95 (0.01) 0.95 (0.01)
3.14 (0.28) 1.21 (0.44) 0.36 (0.85) ±0.1 (1.54) 5.76 (0.69) 5.75 (0.28) 7.14 (0.37)
AR(1) AR(1) AR(2) AR(2) AR(1) AR(1) AR(1)
Coef®cient range
0.90±2.04
0.08±0.88
0.92±0.97
±0.1±7.14
Simple Taylor rule
1.5
0.5
Source: Clarida, Gali and Gertler (1998).
138 Axel A. Weber Table 4.2 Selected in¯ation coef®cient estimates of Kuttner and Posen (1998) Country
Pre-in¯ation targeting period
In¯ation targeting period
Source in Kuttner and Posen (1998)
UK Canada New Zealand
1.38 (0.62) 2.84 (1.35) 3.67 (0.30)
2.01 (0.76) ±0.67 (0.43) 10.95 (1.41)
Table 7, Col. (a) Table 4, Col. (a) Table 9, Col. (a)
Source: Kuttner and Posen (1998).
which corresponds closely to the type of interest rate policy rule we will analyse below. The corresponding -estimates are summarised in Table 4.2. Note that the in¯ation response estimates are typically larger than unity and that in¯ation is assumed to be exogenous in Kuttner and Posen (1998). In the analysis below we will see the potential impact of the possible endogeneity of in¯ation on these coef®cient estimates.
3
Are interest rates and in¯ation linked in the long run?
Let us now turn to this endogeneity issue. Our key argument is that estimating policy reaction functions is uninformative owing to the problem of observational equivalence. Take the link between interest rates and in¯ation, with the latter typically being considered as the monetary policy objective: without detailed knowledge about the time series properties of the in¯ation process, reduced form econometric methods are unable to discriminate empirically between an interest rate policy reaction function (an anti-in¯ation feedback rule) on the one side and the in¯ationary consequences of the interest rate policy on the other side. However, as McCallum (1984) has shown, such tests can be constructed using cross-equation restrictions in a bivariate vector-autoregressive (VAR) approach. Like much of the recent literature, we study a VAR representation of policy and the economy, and we address an important criticism of VAR methods concerning the possible non-robustness of identi®cation. To be precise, we aim at relaxing the parameter restrictions required for identi®cation. In particular, instead of focusing attention on the implications of a single set of point estimates, we consider a sequence of identi®cation schemes satisfying economically motivated inequality constraints. This estimation technique is in the spirit of robustness checks exempli®ed by the studies of Blanchard (1989), King and Watson (1997) and Bernanke and Mihov (1998). We limit ourselves to the analysis of bivariate VARs, since the nature of this simple structural model allows us to relate the long-run comovements of interest rates and in¯ation to interpretable parameters describing the short-run dynamics of both series in a bivariate setting. In particular, we specify the space of
Asymmetric Interest Rate Policy In Europe 139
admissible structural parameters and examine how our tested propositions fare as the parameters vary over that space. In what follows we will focus on the long-run link between in¯ation and interest rates. We argue that this is what matters for a central bank's interest rate policy strategies. Consider the long-run Fisher effect: although many controversies surround the question of how, and to what extent monetary policy can affect the economy in the short run, the long-run neutrality (LRN) of money is widely accepted among policy-makers and monetary economists. LRN means that nominal shocks do not have signi®cant effects on real quantities such as output, employment and real interest rates in the long run. It is further important to note that LRN implies the long-run homogeneity (LRH) of nominal variables: nominal interest rates and in¯ation must change proportionally in response to a nominal shock, which is the well known prediction of the so-called Fisher effect. Previous empirical evidence about the Fisher effect Although LRN and LRH propositions appear to be commonly accepted (as illustrated, for example, by their prominence in undergraduate macroeconomics textbooks), formal empirical tests have yielded surprisingly mixed results. One set of studies looks at the long-run Fisher effect from a crosscountry perspective. Using restricted least-squares regression techniques for data averaged over long periods in a cross-section of countries, Lothian (1985) tests the Fisher relation for twenty OECD countries and Duck (1988, 1993) for a set of thirty-three countries (sixteen industrialised and seventeen developing countries). While Duck (1993) ®nds cross-country evidence in favour of a long-run Fisher effect, Lothian (1985) reports a less than proportional effect of in¯ation on nominal interest rates, thereby rejecting the Fisher hypothesis. Another set of studies, such as Lucas (1980), Mills (1982), Geweke (1982, 1986) and Summers (1983), have attempted to test long-run economic relationships by means of frequency-domain time series techniques. Summers (1983) uses band-spectrum regression techniques to obtain low-frequency estimates of the effects of in¯ation on real interest rates in a Fisher relation and rejects the Fisher relation for post-war United States data. In a third class of papers, Geweke (1986), Stock and Watson (1988), Fisher and Seater (1993), Weber (1994) and King and Watson (1997) draw inference about the long-run Fisher proposition based on explicit tests of coef®cient restrictions in bivariate vector-autoregressive models, while a single equation autoregressive moving average representation is estimated in Mishkin's (1984) analysis of the Fisher effect. The evidence reported in Weber (1994) and King and Watson (1997) rejects the existence of a Fisher effect for United States data, while the results reported in Mishkin (1984) are in favour of a Fisher effect for the United States, the United Kingdom and Canada, but not for Germany, France, the Netherlands or Switzerland.
140 Axel A. Weber
The present chapter follows Geweke (1986), Stock and Watson (1988), Fisher and Seater (1993), Weber (1994) and King and Watson (1997) and bases inference about long-run economic propositions on explicit tests of coef®cient restrictions in bivariate vector-autoregressive models.5 But in order to be able to test for long-run neutrality (homogeneity) it has been shown that meaningful tests can be constructed only if both monetary and real variables satisfy certain non-stationarity conditions, which are spelled out in detail in Fisher and Seater (1993) and King and Watson (1997). These studies demonstrate that straightforward neutrality tests, such as imposing the restriction that the coef®cients of current and lagged monetary impulses in a regression on real economic variables sum to zero, can be conducted only if the order of integration of both series is at least one and equal for both series.6 In addition, both series should not be co-integrated.7 Fisher and Seater (1993) further show that much of the evidence in the older literature on long-run neutrality or homogeneity violates these non-stationarity requirements, and hence has to be disregarded.8 Before we present any structural estimates it is therefore key to discuss brie¯y the unit root and cointegration properties of the data. The time series properties of interest rates and in¯ation As mentioned above, our testing strategy critically depends on the relative order of integration of the data. But these data properties also matter for the approach of Clarida, Gali and Gertler (1998), who explicitly state that their econometric approach relies on the assumption that within their short sample both short-term interest rates and in¯ation are I(0).9 They ®nd that standard Dickey±Fuller tests reject the null hypothesis that in¯ation is I(1) in G-3 economies in favour of the alternative of stationarity. For Germany they reject that the short-term interest rate is I(1), while for the United States and Japan there is less evidence against an I(1) process in interest rates. This would preclude most of our tests. Thus, before presenting any of our results of the Fisher hypothesis, it is important to discuss at some length the unit root properties of the data. The assessment in Tables 4.3 regarding in¯ation rates and nominal as well as real interest rates is based on augmented Dickey±Fuller (ADF) `t-statistics', but we also report Stock's (1991) 95 per cent con®dence intervals for the largest unit root (). Table 4.3 shows that nominal interest rates are non-stationary (I(1)) in Germany, Italy, the Netherlands, Belgium, the United Kingdom and the United States, and trendstationary (I(0)) in France, Denmark and Japan. In¯ation rates, on the other hand, are non-stationary in all countries under study. This is in contrast to the above assessment of Clarida, Gali and Gertler (1998). But it is well known that non-stationarity tests have notoriously low power in discriminating a unit root from an autoregressive coef®cient close to unity, especially in the short samples we are studying. Nevertheless, our
Asymmetric Interest Rate Policy In Europe 141
®nding of unit roots in in¯ation and interest rates is consistent with previous non-stationarity ®ndings in Gali (1992), Weber (1994) and King and Watson (1997), who also ®nd indications of unit-roots in US in¯ation and interest data during the postwar period and use both variables in ®rst difference form in their VARs.
Table 4.3 Unit-root test statistics Country (1)
ADF Test (3)
Drift, Stock's Trend Stock's intervals (4) (5) (6)
Decision (7)
79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11
±2.43 ±3.66* ±3.19 ±1.51 ±2.86 ±4.09** ±2.49 ±2.87 ±3.61*
T T T ± T D,T T T T
0.98 0.93 0.95 0.97 0.90 0.83 0.92 0.94 0.96
( ±, ± ) ( ±, ± ) ( ±, ± ) (0.96,1.02) ( ±, ± ) ( ±, ± ) ( ±, ± ) ( ±, ± ) ( ±, ± )
I(1) I(0) I(1) I(1) I(1) I(0) I(1) I(1) I(0)
79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11
±1.70 ±0.88 ±1.03 ±1.69 ±2.36 ±2.75 ±1.88 ±1.63 ±1.82
± ± ± ± T T ± ± ±
0.98 1.00 1.00 0.98 0.97 0.95 0.98 0.99 0.97
(0.95,1.02) (0.98,1.02) (0.97,1.02) (0.95,1.02) ( ±, ± ) ( ±, ± ) (0.94,1.02) (0.95,1.02) (0.94,1.02)
I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1) I(1)
±3.04* ±3.20* ±2.31 ±2.13 ±2.69 ±2.99* ±2.03 ±2.80 ±3.11
D D D D D D D D T
0.93 0.94 0.97 0.95 0.90 0.88 0.94 0.94 0.91
(0.88,1.00) (0.87,1.00) (0.92,1.01) (0.93,1.01) (0.90,1.01) (0.88,1.00) (0.93,1.01) (0.90,1.01) ( ±, ± )
I(0) I(0) I(1) I(1) I(1) I(0) I(1) I(1) I(1)
Period (2)
(A) Nominal interest rates Germany France Italy Netherlands Belgium Denmark UK USA Japan (B) In¯ation rates Germany France Italy Netherlands Belgium Denmark UK USA Japan
(C) Real interest rates Germany France Italy Netherlands Belgium Denmark UK USA Japan
79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11
142 Axel A. Weber Table 4.3 (continued) Country (1)
Period (2)
ADF Test (3)
Drift, Stock's Trend Stock's intervals (4) (5) (6)
Decision (7)
(D) Nominal interest rate differentials (relative to Germany) France Italy Netherlands Belgium Denmark UK USA Japan
79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11 79:03±98:11
±3.00 ±2.23 ±4.91** ±3.69* ±4.91** ±2.40 ±1.95 ±3.72*
T T ± D,T D,T D ± D,T
0.91 0.96 0.82 0.77 0.72 0.93 0.97 0.94
( ±, ± ) ( ±, ± ) ( ±,0.89) ( ±, ± ) ( ±, ± ) (0.91,1.01) (0.94,1.01) ( ±, ± )
I(1) I(1) I(0) I(0) I(0) I(1) I(1) I(1)
Notes:
Column 3 reports the augmented Dickey±Fuller tests for detrended data or demeaned data,
respectively. Their signi®cance levels are taken from Table 8.5.2. of Fuller (1976), p. 373.
A rejection of the null hypothesis of a unit root at the 1 per cent signi®cance level is marked with **,
at the 5 per cent level with *.
Stock's (1991) 95 per cent con®dence intervals for the largest unit root were calculated from the
ADF statistics using Stock's Tables A1 and A2 and the proceedure described in Appendix B of his
(1991) paper.
In addition to the con®dence belts for the estimated roots are displayed.
All ADF statistics are based on regressions including six lagged differences of the variable.
A second condition necessary for neutrality tests to be meaningful, as pointed out by King and Watson (1992), is that nominal interest rates and in¯ation must not be simultaneously non-stationary and cointegrated. The neutrality hypothesis implied by the Fisher relation is that in¯ation has a zero long-run effect on real interest rates and for this to be testable real interest rates must be integrated of order one if in¯ation follows an I(1) process. This in turn requires that nominal interest rates (rt ) and in¯ation (t ) must not be cointegrated, since otherwise their linear combination is stationary. The unit root tests reported for real interest rates in Table 4.3 suggest that except for Germany, France and Denmark real interest rates also appear to be integrated of order one from the ADF tests. Based on this assessment we decided to carry out the neutrality tests conditional on integrated processes, but we accept that in our short sample some caution should be exercised in interpreting these results, since a misspeci®cation might arise from the potential stationarity or cointegration in the data for Germany, France and Denmark. However, our results also show that a spurious regression problem may apply to the estimates of interest rate smoothing rules speci®ed in levels of interest rates, in¯ation and output.10
Asymmetric Interest Rate Policy In Europe 143
An econometric framework for studying the long run link between interest rates and in¯ation McCallum (1984) points out that valid tests of long-run neutrality may be constructed only by using cross-equation restrictions in a bivariate approach. The present chapter follows King and Watson (1997) and bases inference about long-run economic propositions on explicit tests of coef®cient restrictions in bivariate vector-autoregressive models. In order to brie¯y illustrate their approach, consider the hypothesis that a permanent change in the in¯ation rate (pt t ) has no long-run consequences for the level of real interest rates (rt ). Since real interest rates are typically measured as the difference between nominal interest rates and in¯ation rates, this neutrality proposition implies the homogeneity restriction that nominal interest rates move one-to-one with in¯ation in the long run. If we allow in¯ation to affect nominal interest rates (i.e. owing to the Fisher effect) and at the same time assume that in¯ation reacts endogenously to movements in interest rates rates (i.e. as a result of the effects of an interest rate policy) we obtain a bivariate vector-autoregressive (VAR) model in both in¯ation and interest rates: t i it
p X j1
it i t
j
i it�j
p X j1
j
ii it�j
p X j1 p X j1
j t�j "m t
j
i t�j "t
4:6
4:7
where i and i represent the contemporaneous effect of interest rate policy on in¯ation and the contemporaneous response of interest rate policy to in¯ation, respectively. A more convenient representation of this bivariate VAR system is:
Lt i
Lit "m t
4:8
ii
Lit i
Lt "t
4:9
whereby Pp Pp Pp j j j j j ii
L 1 � j1 ii Lj ; i
L i j1 i L ;
L 1 � j1 L as well Pp j j as i
L i j1 i L applies. In stacked form this may be re-written as:
LXt "t where
L
Pp
j Lj with m " t 1 Xt ; "t t ; 0 "t it �i
j0
4:10 �i ; 1
144 Axel A. Weber
"
j
and � j i j
# j i ; j 1; 2; :::; p j ii
In the above notation the long-run multipliers are i and i; where
i i
1=ii
1 measures the long-run response of nominal interest rates to a one unit permanent increase in in¯ation, while i i
1=
1 measures the long-run response of in¯ation to a permanent unit increase in interest rates. The long-run Fisher effect implies the restriction i 1. As noted by King and Watson (1997), (4.10) is econometrically unidenti®ed and the restrictions implied by the Fisher effect are no longer testable when in¯ation is endogenous. Thus, even if the hypothesis that "m t and "t are uncorrelated is maintained, one additional restriction is required in order to identify the linear simultaneous equation model. Common identifying assumptions are that in¯ation is exogenous ( i 0) or predetermined i 0. Alternatively, the long-run Fisher effect with i 1 may be imposed in order to identify the system and estimate the remaining parameters. In principle it is possible to identify the above model by specifying a value of any one of the four parameters i ; i ; i or i and then ®nding the implied estimates for the other three. Empirical evidence regarding the long-run link between interest rates and in¯ation Table 4.4 reports the evidence for the ERM period. As compared to Clarida, Gali and Gertler (1998) we analyse the slightly longer sample 1979:03 to 1998:10 and consider a broader set of European countries. The estimated short-run correlation (column (5)) between nominal interest rates and in¯ation for Germany is zero, and the long-run correlation (column (8)) of 0.79 is much higher but still substantially smaller than one. For France the short-run correlations are somewhat higher (0.15) and the long-run correlations somewhat lower (0.62) than for Germany, and the same holds for Italy, the Netherlands, Belgium, and Denmark. The United Kingdom, and Japan have higher long-run correlations than Germany, whilst the United States displays one of the lowest long-run correlations. Imposing a long-run Fisher effect
i 1 furthermore results in some interesting estimates of the remaining coef®cients (columns (12)±(14)). For example, under a long-run Fisher effect in¯ation is found to be exogenous in the long run in all countries. For Germany, France, Italy, the Netherlands, the United Kingdom, the United States and Japan a signi®cant negative shortrun impact of domestic interest rate policy on in¯ation is found, but not for Belgium and Denmark. For the small ERM countries this result is not too surprising, given the high degree of openness and the importance of non domestically determined traded goods prices in their consumer price indices.
Table 4.4
Long-run link between interest rate policy and in¯ation in European and selected G7-countries
Country (1)
Period (2)
Germany France Italy Netherlands Belgium Denmark UK USA Japan
79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10
VAR estimates 2 cori 2i (3) (4) (5) 0.31 0.27 0.31 0.27 0.32 0.45 0.49 0.27 0.44
0.30 0.01 0.62 0.15 0.55 0.04 0.61 ±0.01 0.95 0.08 1.37 0.10 1.08 0.17 0.67 0.16 0.29 ±0.01
Structural model estimates 2i 2 cori (6) (7) (8) 0.49 0.47 0.71 0.40 0.39 0.49 1.14 0.55 0.54
0.75 0.77 0.73 0.55 0.63 0.87 0.86 0.68 0.70
0.79 0.62 0.66 0.48 0.47 0.50 0.86 0.49 0.88
i 1 in 95 per cent con®dence interval i i
i (9) (10) (11) ±1.8,0 ±0.6,0 ±0.04 ±0.2 ±0.3,0.8 ±0.1,0.2 ±0.2 ±1.1,0 0.02
0,1.5 0.44 0.26 0.86 ±0.29 0.85 1.16 0.84 0
±4 ±2.5 0.16 0.17 0.2 n.a n.a. 0.1 ±
Estimates imposing i 1 i (12) 0.18 0.01 ±0.86 ±0.29 ±0.14 ±0.05 4.26 ±0.43 0.25
i (13)
i (14)
(0.21) 0.48 (0.26) ±0.50 (0.28) (0.19) 1.24 (0.51) ±0.18 (0.12) (1.02) 1.13 (0.51) ±0.35 (0.18) (0.32) 2.15 (0.86) ±0.43 (0.18) (0.21) 0.80 (0.55) ±0.06 (0.07) (0.15) ±0.11 (0.48) 0.05 (0.06) (4.02) 2.27 (0.72) ±0.48 (0.22) (0.40) 1.89 (0.63) ±0.27 (0.14) (0.22) 0.28 (0.16) ±0.67 (0.38)
Notes: All results for the second moments are based on VARs with six lags.
2i denotes the variance estimate for variable i, corij indicates the correlation between variables i and j.
Variances and correlations are calculated for the residuals of the unrestricted VARs and the shocks implied by the long-run covariance matrix of the estimated
VAR (the spectral density matrix of the variables at frequency zero).
The coef®cient ranges in columns (9)±(11) are those for which the long-run homogeneity proposition cannot be rejected at the 95 per cent level.
The point estimates of the coef®cients and their 95 per cent con®dence regions (2 standard errors in parentheses) implied by long-run homogeneity are
reported in columns (12)±(14).
145
146 Axel A. Weber
Another interesting ®nding of Table 4.4 concerns the link between the longrun and short-run in¯ation response coef®cients. Both the Taylor rule and the interest rate smoothing rules of Clarida, Gali and Gertler (1998) have a built-in long-run Fisher effect as the neutral policy stance under which the central bank's interest rate policy neither actively stabilises nor accommodates in¯ation. The Fisher effect is therefore useful as a benchmark case. By imposing the Fisher effect ( i 1), we obtain estimates of the short-run in¯ation response coef®cient (i ) which are signi®cantly larger than one for the Netherlands, the United Kingdom and the United States, while for Germany, France, Italy, Belgium and Japan they are signi®cantly greater than zero but not larger than one. For the latter countries a less than proportional interest rate adjustment to in¯ation appears to be consistent with a long-run neutral policy stance, whilst for the Netherlands, the United Kingdom and the United States the data point towards some degree of short-run interest rate smoothing policies. Since the short-run in¯ation response coef®cient for Germany is signi®cantly smaller than unity, does this imply that the Bundesbank has accommodated in¯ation? The answer is no. Consider Figure 4.3, which displays the coef®cient sensitivity analysis graphically. Economically meaningful policy reaction functions clearly imply i 1. Panel b shows that for Germany even moderately positive in¯ation response coef®cients (0 < i < 1:5) are compatible with a long-run Fisher effect ( i 1). According to Panel c the data also indicate that German in¯ation is in the long run exogenous with respect to interest rates across a wide range of possible identi®cation schemes, whilst from Panel a it is obvious that i 1 implies a strongly negative immediate impact of interest rates changes on in¯ation (i ). On the other hand, if we assume in¯ation to be predetermined (i 0), this is consistent with i 1 only for an initial policy response to in¯ation which is close to zero (�0:1
Asymmetric Interest Rate Policy In Europe 147
Figure 4.3 The link between interest rates policy and in¯ation, Germany, 1979:3± 1998:10
monetary policy stance. This empirical result may simply be a re¯ection of the fact that the United States actually conducted its monetary policy through an interest rate smoothing rule, whilst the Bundesbank did not. For the remaining countries the results in Table 4.4 may be summarised as follows: Overall, our results are quite similar for all countries, and Figure 4.5 demonstrates this for the con®dence ellipses under the null hypothesis of a long-run Fisher effect. More speci®cally, our ®ndings for most ERM countries resemble the results obtained for Germany, while the United Kingdom is more like the United States. With the exception of the United States, the closest long-run correlations between in¯ation and interest rates are typically found for the large economies (Germany, United Kingdom, Japan), while smaller economies display a lower long-run correlation. This may indicate that during the ERM German interest rates rather than domestic in¯ation may have been the dominant driving force behind domestic interest rates in many countries. We will examine this proposition in more detail below.
148 Axel A. Weber
Figure 4.4 The link between interest rates policy and in¯ation, United States, 1979:3±1998:10
4
Are European interest rates linked in the long run?
Since January 1999 European interest rate policy was placed under the authority of the European Central Bank (ECB) and of®cial interest rates in EMU member countries are, by de®nition, moving in a perfectly symmetric fashion. Prior to the creation of the monetary union this was not the case, but owing to nominal convergence there has been an increasing degree of co-movements between European rates. In this context it has frequently been argued that Germany has played the role of the anchor country in the ERM, and that interest rate changes originating in Germany would ultimately cause symmetric changes in the interest rates of the remaining ERM countries. To examine this proposition, we ®rst analyse the data properties and then apply our estimation technique. The time series properties of interest rate differentials As already mentioned above, our testing strategy depends on the relative order of integration of the data, which need to be I(1) and not cointegrated. The time series properties of nominal interest rates were discussed on p. 140, and it was shown that nominal interest rates are non-stationary, I(1), in Germany, Italy, the Netherlands, Belgium, the United Kingdom and the United States but trendstationary, I(0), in France, Denmark and Japan. For the latter countries
Asymmetric Interest Rate Policy In Europe 149
5
95% confidence ellipse when γyi,π=1 Netherlands
3
4
UK
USA
1
Germany
0
λi,π
2
Italy
–1
Belgium
–2
Denmark –1.2
–0.8
–0.4
–0.0
0.4
λπ,i
Figure 4.5 The estimated tradeoff between policy response and the policy effects for interest rate policy and in¯ation, 1979:03±1998:10 Note: The coef®cients ;i and i; represent the contemporaneous effect of interest rate policy on in¯ation and the contemporaneous in¯ation response of interest rate policy respectively. The con®dence epilses indicate the parameter space for the pair
;i ; i; over which the data do not reject the long-run Fisher-effect at the 5 signi®cance level.
the tests below have to be interpreted with caution. But in addition to nonstationarity, the absence of cointegration and hence the non-stationarity of Germany is required. Based on augmented interest rate differentials vis-a-vis Dickey±Fuller tests we cannot reject the non-stationarity of interest rate differentials for France, Italy, the United Kingdom and the United States, while trend-stationary processes are found for Belgium, Denmark and Japan and a mean-stationary process is identi®ed for the Netherlands. We will discuss our ®ndings in detail only for the ®rst set of countries. A VAR model of international interest rate linkages Formally, the long-run cross-country proportionality of interest rates may be studied by relating interest rate movements abroad (it ) to changes in the German interest rates (it ): it i i it "t
4:11
with a coef®cient ii 1 implying the perfect immediate transmission of German interest rate innovations. But typically estimation of (4.11) is
150 Axel A. Weber
complicated by two facts. First, the interest rate policy may be transmitted with lags, and in practice it therefore makes sense to distinguish between the immediate and the long-run transmission of interest rate policy by using the dynamic speci®cation: p X
it i i it
j1
j
i i it�j
p X j1
j
i i it�j "t
4:12
which allows the concept of long-run versus short-run policy transmission to be implemented empirically. A second important complication in the estimation of (4.11) or (4.12) arises from the fact that the German interest rate it cannot safely be considered as an exogenous variable. For example, if the German central bank is concerned about exchange rate movements and therefore responds to interest rate movements abroad, German rates would change in line with interest rates abroad. This endogeneity problem may explicitly be taken into account by estimating the simultaneous two equation system: it i i it
p X
it i i it
j1
j
i i it�j
p X j1
j
ii it�j
p X j1 p X j1
j
i i it�j "m t
j
i i it�j "t
4:13
4:14
using the 2SLS instrumental variable techniques developed in King and Watson (1997). This approach has two advantages over the previous literature: ®rst, the estimates explicitly allow German interest rate movements to be both predetermined (i i 0) and/or exogenous in the long-run [ i i i i
1=i i
1 Pp Pp j 0 with lag polynomials i i
L i i j1 i i L j and i i
L 1 ± j1 ii i Lj ] without necessarily imposing these restrictions onto the data. Second, equations (4.13) and (4.14) provide a natural framework within which the concept of immediate interest rate policy transmission (ii 1) and long-run Pp i j transmission [ ii ii
1=ii
1 1 with ii
L ii j1 ii L and ii
L Pp j j 1 � j1 ii L ] can be formalised and tested empirically. As noted above, the bivariate VAR system in (4.13) and (4.14) is econometrically unidenti®ed. In the previous literature various identifying restrictions are to be found: it is common practice in the older literature to invoke the timing assumption and restrict German interest rate policy signals to be long-run exogenous, so that i i 0 holds. Another approach is to assume that the model is recursive, so that German interest rates are predetermined i i 0. Alternatively we could impose ii 0. As before, we will identify the model by pre-specifying a value for any one of the four
Asymmetric Interest Rate Policy In Europe 151
parameters ii ; i i ; i i or ii and then ®nding the implied estimates for the other three parameters. Empirical evidence about long-run international interest rate linkages Before discussing the results it is important to note that equations like (4.13) have previously been estimated in the context of the EMS asymmetry literature by applying Granger-causality tests to European interest rates. Such Granger-causality tests have been conducted by Cohen and Wyplosz (1989), DeGrauwe (1989), Fratianni and von Hagen (1990a, 1990b), and Weber (1990) in order to demonstrate the asymmetrical functioning of the EMS as a de facto `DM zone'. However, the evidence from this literature has revealed a surprisingly symmetrical working of the EMS, with causality typically running both ways. Wyplosz (1991) has argued that this ®nding is hardly surprising, given that the strategic (game-theoretic) policy-making literature suggests that it is sub-optimal for policy makers to follow a predetermined or exogenous policy without taking the other central bank's past or current policy reactions into account. Equation (4.14) captures precisely this fact. The two-equation system estimated here may thus be viewed as a more consistent approach to testing for asymmetric interest rate policies if policy makers are known to behave strategically and to care about each other's policy settings. Some non-formal evidence about the high degree of European interest rate policy interdependence was already visible in Figure 4.1. Discount rate policies appear to be relatively closely co-ordinated between Germany, the Netherlands ± and, to a slightly lesser extent Belgium. The same three countries display close co-movements of short-term money market rates, long-term government bond rates and in¯ation rates. Figure 4.1 also showed that there were short-run destabilising effects of the speculative attacks between September 1992 and August 1993 on European money market interest rates ± in particular, for Italy, Ireland, Denmark and France. Overall, the impression from Figure 4.1 was that interest rate co-movements within Europe were quite pronounced. In order to obtain formal evidence on the degree of asymmetry in interest rate policies in Europe we have estimated the simultaneous two-equation system (4.13) and (4.14). Table 4.5 reports the empirical results. All the long-run correlations (column (8)) of interest rate innovations are positive and fairly high for France (0.81) and the Netherlands (0.75), while relatively low correlations are found for Italy (0.46) and the United Kingdom (0.31). This is caused by the speculative attacks of September 1992, because the exit of both countries from the ERM allowed their interest rates to move more freely and more independently from the rest of Europe. The highest short-run correlation is found for the Dutch±German interest rate, and thus points towards a high degree of short-run policy credibility. Columns (9)±(14) show our results when we impose a perfect long-run comovement of interest rates, which theoretically should be the case under
152
Table 4.5
Long-run cross-country in¯uence of German interest rates policy on interest rates in European and selected G7-countries
Country (1)
Period (2)
VAR Structural model estimates estimates 2i corii i2 2i corii 2i (3) (4) (5) (6) (7) (8)
France Italy Netherlands Belgium Denmark UK USA Japan
79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10 79.03±98.10
0.31 0.31 0.30 0.31 0.31 0.31 0.29 0.31
0.59 0.18 0.58 0.87 0.55 0.12 0.55 0.79 0.56 0.35 0.52 0.56 0.91 ±0.01 0.53 0.64 1.44 0.13 0.55 1.33 1.15 0.11 0.54 0.83 0.70 0.04 0.57 0.87 0.31 0.20 0.55 0.71
0.81 0.46 0.75 0.59 0.53 0.31 0.54 0.37
ii 1 in 95 per cent con®dence interval ii ii
ii (9) (10) (11)
±0.2,0.2 ±0.4,0.1 ±0.2,0.1 ±0.3,0.0 ±0.1,0.1 ±0.3,0.0 ±0.8,±0.1 ±0.2
Estimates imposing ii 1 ii (12)
ii (13)
ii (14)
±0.5,0.9 ±0.3 0.21 (0.14) 0.20 (0.31) 0.04 (0.09) 0,1.3 0.1 ±0.29 (0.24) 0.56 (0.29) ±0.11 (0.11) 0.1,1.2 0.2 ±0.48 (0.53) 0.71 (0.25) ±0.02 (0.09) 0.2,2.8 0.1 ±0.39 (0.33) 1.12 (0.54) ±0.13 (0.06) ±1.0,1.8 ± 0.1 0.06 (0.09) 0.53 (0.73) 0.01 (0.04) 0.9,3.6 0.0 ±0.27 (0.20) 1.91 (0.64) ±0.12 (0.06) 0.8,4.0 0.2 ±0.12 (0.20) 1.82 (0.64) ±0.30 (0.12) 0.4 0.6 ±0.44 (0.36) 0.80 (0.23) ±0.74 (0.35)
Notes:
All results for the second moments are based on VARs with three lags.
2i denotes the variance estimate for variable i, corij indicates the correlation between variables i and j.
Variances and correlations are calculated for the residuals of the unrestricted VARs and the shocks implied by the long-run covariance matrix of the estimated VAR (the spectral density matrix of the variables at frequency zero). The coef®cient ranges in columns (9)±(11) are those for which the long-run homogeneity proposition cannot be rejected at the 95 per cent level. The point estimates of the coef®cients and their 95 per cent con®dence regions (2 standard errors in parentheses) implied by long-run homogeneity are reported in columns (12)±(14).
Asymmetric Interest Rate Policy In Europe 153
EMU. Let me focus on the ERM countries here and disregard the free ¯oaters, because for them this is an inadequate restriction.12 But even among the ERM countries this restriction is inconsistent with a predetermined or a long-run exogenous German interest rates policy. Except for France the estimated coef®cients in columns (12) and (14) are predominantly negative and frequently signi®cant. For the data not to reject the imposed restriction, the Bundesbank would have had to increase German interest rates in response to falling interest rates in the rest of Europe. During German uni®cation this was in part the case, but the restrictive German monetary policy was motivated exclusively by domestic objectives related to the tax versus credit ®nancing of uni®cation and not in response to European developments. Furthermore, the above restriction would imply a relatively high immediate impact of German interest rate policy on interest rates abroad, ranging from 0.2 for France to 1.12 for Belgium. Except for France this by far exceeds the short-run correlations in the data and we therefore reject this restriction. When comparing Tables 4.4 and 4.5, another interesting result becomes obvious. For the smaller ERM countries, such as the Netherlands, Belgium or Denmark, the long-run correlations of domestic interest rates with German interest rates are actually higher than the corresponding correlations with domestic in¯ation. In this case the obvious interest rate policy rule appears to be that these countries have predominantly followed interest rate movements in Germany. But our estimates in Table 4.5 show that whilst interest rates appear to be interconnected, there is no simple long-run homogeneous pattern of co-movements during the ERM period.
5 Summary and conclusions In this chapter we have taken a closer look at interest rate behaviour in industrialised countries with special emphasis on Europe. We have argued that it is inherently dif®cult to interpret interest rate movements simply as re¯ecting policy actions in response to key economic variables such as in¯ation or output growth. Rather than following the standard approach of the current literature and estimating interest rate smoothing policy rules in order to discuss their degree of asymmetry, we have analysed the joint behaviour of interest rates and in¯ation under a variety of identifying restrictions. For simplicity and clarity of the argument we have chosen a bivariate setting with interest rates and in¯ation only. Orphanides (1999) refers to this as a `simple prudent policy rule'. In our analysis we inquire whether or not the data are consistent with a long-run Fisher effect under endogenous interest rate policy choices. In this case it is obvious that no simple interpretation of the joint behaviour of interest rates and in¯ation is possible owing to the problem of observational equivalence. This is a well known and old problem of single-equation econometric policy evaluation, dating back to the early studies of Lucas (1972,
154 Axel A. Weber
1976) Sargent (1973) and McCallum (1984). What is surprising is the consistency and persistency with which this problem is being ignored in empirical work. The main reason for this is that no easily interpretable econometric results follow from the type of sensitivity check which we have performed in this chapter with respect to different identi®cation schemes. We ®nd that it is dif®cult to disentangle the causes and consequences of interest rate policy even in our simple prudent policy rule which only incorporates two variables. But imposing certain long-run neutrality characteristics (i.e. the Fisher effect) helps us to distinguish between the causes and the consequences of interest rate policy within a bounded set of parameter values. Our ®ndings and conclusions are, of course, subject to some reservations: ®rst, to analyse `typical' policy rules our bivariate prudent policy rule would have to be extended to a multivariate setting to avoid the danger of a serious misspeci®cation. Second, the short sample of monthly observations for the ERM period conveys only limited `long-run' information, and this seriously undermines the power of both the unit root and the long-run neutrality tests conducted in this chapter. But moving to longer samples involves the risk of mixing different policy regimes. Both are serious drawbacks, and there is no obvious solution. Third, the application of the neutrality tests developed by King and Watson (1997) critically depends on the order of integration in the data, which is subject to a large degree of uncertainty owing to the low power of unit root tests. It is found that the sub-set of time series which possess the postulated unit-root properties may be small, so that in many cases long-run neutrality or homogeneity tests are applicable only to a limited extent. The present chapter also raises another critical point concerning the standard literature on interest rate smoothing policies, which has to do with the degree of interdependence of interest rate policies across Europe. Again, it is well known that interest rates are linked through the uncovered interest rate parity (UIP) condition and it is hard so see how under an exchange rate pegging arrangement like the ERM independent national interest rate policies can be conducted. This is different under a free ¯oat, and our estimates con®rm this. But since European interest rates are shown to move closely together, estimating independent interest rate smoothing rules for various ERM countries (such as Germany, France and Italy) in Clarida, Gali and Gertler (1997) is problematic. For these countries movements in German rates need to be accounted for when estimating interest rate policy reaction functions. For the smaller ERM countries, such as the Netherlands, Belgium or Denmark the long-run correlations of domestic interest rates with German interest rates are actually higher than the corresponding correlations with domestic in¯ation. But in spite of all the above mentioned limitations, we think that the King and Watson (1997) methodology provides a powerful tool for analysing the long-run predictions of economic theories within a fairly general framework. In future research, we plan to extend our model to a three-variable VAR including output, in¯ation and interest rates. However, as pointed out by King
Asymmetric Interest Rate Policy In Europe 155
and Watson (1997), the identi®cation problem becomes much more dif®cult in this case, since the number of identifying parameter restrictions increases with the square of the variables in the model. This may pose computational problems, given that each identifying restriction is iterated hundreds of times over a wide range of plausible values.
Appendix
Time series and data sources
The chapter uses monthly data. The time series and data sources used were: . Interest rates: Money market rates, International Monetary Fund (IMF), International Financial Statistics, various issues, line 60b: ± Discount rates, IMF, International Financial Statistics, various issues, line 60. ± Treasury bill rates, IMF, International Financial Statistics, various issues, line 60c. ± Government bond rates, IMF, International Financial Statistics, various issues, line 61. . Consumer prices: IMF, International Financial Statistics, various issues, line 64. . Money stock (M1): IMF, International Financial Statistics, various issues, line 34. . Output, industrial production: IMF, International Financial Statistics, various issues, (line 66b)
Notes 1. Gerlach and Smets (1998) apply such a model to aggregate data constructed for the Euro area, whilst the other studies focus on various countries separately. 2. Clarida, Gali and Gertler (1998) try to control for the endogeneity of expected in¯ation by using past interest and in¯ation rates as well as other variables (output, commodity prices and real exchange rates) as instruments in a two-step nonlinear two-stage least squares estimator. For details see n. 11 of their paper. 3. Obviously, our choice to employ a bivariate VAR instead of a multivariate VAR approach involves the risk of a potential misspeci®cation. However, the alternative approach also suffers from a potential spurious or unbalanced regression problem since output is typically not found to be a trendstationary process, as proposed by Clarida, Gali and Gertler (1998). We ®nd that using output growth instead results in at the best only marginally signi®cant coef®cient estimates. Thus, under a more appropriate speci®cation the potential misspeci®cation problem largely reduces. In our view this justi®es the use of a bivariate VAR, but we agree that extending our model to a multivariate VAR is the obvious next step for future research. 4. See Rudebusch and Svensson (1998), p. 25. 5. Long-run neutrality here implies a zero-restriction on the sum of coef®cients of the contemporaneous and lagged in¯ation variables in a regression on real interest rates. Tests of short-run neutrality, such as those conducted in Sims (1972), on the other hand, impose zero-restrictions on the individual coef®cients of the monetary impulses. Long-run neutrality is thus a weaker test, and short-run non-neutralities may well be compatible with long-run neutrality.
156 Axel A. Weber 6. In the present context unit root tests are important since they indicate those cases for which the neutrality results may be a statistical artefact. This has been shown to be the case if the two variables under study do not have the same relative order of integration. There is no implication whatsoever with respect to the absolute order of integration in the data, given that it is at least one. The absolute order of integration does determine the appropriate minimum degree of differencing of the data before conducting neutrality tests, but overdifferencing does not affect the neutrality propositions. See Fisher and Seater (1993) for details. 7. If it and t are non-stationary and cointegrated, then a ®nite VAR in ®rst differences does not exist because there would be a unit root in a moving average polynomial, which is not invertable. Fisher and Seater (1993) brie¯y discuss cointegration in the context of long-run neutrality restrictions in an Appendix. They show that cointegration per se does not affect the long-run neutrality restrictions derived. However, cointegration typically is suf®cient for rejecting long-run neutrality. 8. Long-run neutrality test results will be presented even if unit root tests suggest that they may be uninformative. The key point is that the unit root tests may indicate in which case the neutrality results are a statistical artifact. 9. See Clarida, Gali and Gertler (1998), n. 9. 10. Similar unit root tests for output (not reported here) show that in none of the countries considered here is output (industrial production) driven by a trendstationary process, as is assumed in Clarida, Gali and Gertler (1998) for their measure of the output gap. In line with much of the literature we ®nd output to be an I(1) process. 11. Note that the corresponding estimates of i in Clarida, Gali and Gertler (1998) for Germany are also very low (i 0:118) owing to the large AR(1) coef®cient in the interest rate equation. 12. For the free-¯oating countries the data are clearly inconsistent with this restriction: in most cases the contemporaneous effect of German interest rate changes on foreign interest rates would be relatively high, taking minimum values of 0.4 for Japan to 0.8 for the United States and 0.9 for the United Kingdom. This by far exceeds the short-run correlations in the data. German interest rates would also be endogenous and react strongly to interest rate settings abroad.
Asymmetric Interest Rate Policy In Europe 157
References Bernanke, B. and I. Mihov (1998), `The Liquidity Effect and Long-run Neutrality', National Bureau for Economic Research Discussion Paper, 6608. Bernanke, B. and M. Woodford, (1996), `In¯ation Forecasts and Monetary Policy', Princeton University, October, mimeo. Blanchard, O., (1989), `A Traditional Interpretation of Macroeconomic Fluctuations', American Economic Review, 79, 1146±64. Clarida, R. and M. Gertler (1997) `How the Bundesbank Conducts Monetary Policy', in C.Romer and D. Romer (eds), Reducing In¯ation, Chicago, Chicago University Press. Clarida, R., J. Gali and M. Gertler (1997) `Monetary Policy Rules and Macroeconomic Stability: Theories and Some Evidence', New York University, May, mimeo. ÐÐÐÐ (1998) `Monetary Policy Rules in Practice', European Economic Review, 42, 1033± 68. Cohen, D. and C. Wyplosz, (1989) `The European Monetary Union: An Agnostic Evaluation', Centre for Economic Policy Research Discussion Paper, 306. DeGrauwe, P. (1989) `Is the European Monetary System a DM-Zone?', Centre for European Policy Studies Working Document, 39. Duck, N. W. (1988) `Money, Output and Prices: An Empirical Study Using Long-term Cross Country Data' European Economic Review, 32, 1603±19. ÐÐÐÐ (1993) `Some International Evidence on the Quantity Theory of Money', Journal of Money, Credit and Banking, 25, 1±12. Fisher, M. E. and Seater, J. J. (1993) `Long-Run Neutrality and Superneutrality in an ARIMA Framework', American Economic Review, 83, 402±15. Fuller, W. A. (1976) Introduction to Statistical Time Series, New York: John Wiley. Fratianni, M. and J. von Hagen (1990a) `German Dominance in the EMS: Evidence from Interest Rates', Journal of International Money and Finance, Vol. 9, pp. 358±375 ÐÐÐÐ (1990b) `The European Monetary System: Ten Years After', in A.H.Meltzer and C. Plosser (eds), Carnegie±Rochester Conference Series on Public Policy, 32. Gali, J. (1992) `How Well Does the IS-LM Model Fit the Postwar US Data?', Quarterly Journal of Economics, may, 709Ð38. Gerlach, S. and F. Smets (1998) `Output Gaps and Monetary Policy in the EMU Area', Bank for International Settlements, September, mimeo. Geweke, J. (1982) `Measurement of Linear Dependence and Feedback Between Multiple Time Series', Journal of the American Statistical Association, 77, 304±13. ÐÐÐÐ (1986) `The Superneutrality of Money in the United States: An Interpretation of the Evidence', Econometrica 54, 1±21. King, R. G. and M. W. Watson (1997) `Testing Long Run Neutrality' Federal Reserve Bank of Richmond, Economic Quarterly, 83, 69±101. Kuttner, K. N. and A. S. Posen (1999) `Does Talk Matter After All? In¯ation Targeting and Central Bank Behaviour', mimeo. CFS Working paper 99/04. Lothian, J. R. (1985) `Equilibrium Relationships Between Money and Other Economic Variables' American Economic Review, 75, 828±35. Lucas, R. E. Jr, (1972) `Econometric Testing of the Natural Rate Hypothesis', In O. Eckstein (ed.), `The Economics of Price Determination', Washington, DC, Board of Governors of the Federal Reserve System. ÐÐÐÐ (1976) `Econometric Policy Evaluation: A Critique. The Phillips Curve and Labour Markets', Carnegie±Rochester Conference Series on Public Policy, 1, K. Brunner and A. H. Meltzer (eds), Amsterdam: North-Holland.
158 Axel A. Weber ÐÐÐÐ (1980) `Two Illustrations of the Quantity Theory of Money', American Economic Review, 70, 1005±14. McCallum, B. T. (1984) `On Low-frequency Estimates of Long-run Relationships in Macroeconomics', Journal of Monetary Economics, 14, 3±14. Mills, T. C. (1982) `Signal Extraction and Two Illustrations of the Quantity Theory', American Economic Review, 72, 1162±8. Mishkin, F. S. (1984) `The Real Interest Rate: A Multi-country Empirical Study', Canadian Journal of Economics, 1, 283±311. Orphanides, A. (1999) `The Quest for Prosperity Without In¯ation', Board of Governors of the Federal Reserve System, May, mimeo. Rudebusch, G. and L. Svensson (1998) `Policy Rules for In¯ation Targeting', Centre for Economic Policy Research Discussion Paper, 1999, October. Sargent, T. J. (1973) `Interest Rates and Prices in the Long Run: A Study of the Gibson Paradox', Journal of Money, Credit and Banking, 5, 383±449. Sims, C. A. (1972) Money, Income and Causality', Papers and Proceedings of the American Economic Association, 62, 540±52. Stock, J. H. (1991) `Con®dence Intervals for the Largest Unit Root in US Macroeconomic Time Series', Journal of Monetary Economics, 28, 435±59. Summers, L. H. (1983) `The Non-adjustment of Nominal Interest Rates: A Study of the Fisher Effect in J.Tobin (ed.) Macroeconomics, Prices and Quantities: Essays in Memory of Arthur Okun, Washington DC, Brookings Institution. Taylor, J. B. (1993) `Discretion versus Rules in Monetary Policy Making', in A.H.Meltzer, and Plosser, (eds), Carnegie Rochester Conference Series on Public Policy, 39, December, 195±214. Weber, A. A. (1990) `European Economic and Monetary Union and Asymmetries and Adjustment Problems in the European Monetary System: Some Empirical Evidence', Centre for Economic Policy Research Discussion Paper 448. ÐÐÐÐ (1991) `Credibility, Reputation and the European Monetary System', Economic Policy, 12, 57±102. ÐÐÐÐ (1994) `Testing Long-run Neutrality: Empirical Evidence from the G7Countries with Special Emphasis on Germany', in A. H. Meltzer and Plosser (eds), Carnegie±Rochester Conference Series on Public Policy, 41, 67±117. Wyplosz, C. (1991) `EMS Puzzles', Revista Espanola de Economia, 7(1), 33±66.
Discussion
Carlo A. Favero
This Comment offers some criticisms of the empirical literature on monetary policy rules by focusing on two main issues: interpretation of estimated parameters and the effects of misspeci®cation. On the issue of identi®cation it is argued that it is impossible to disentangle the causes and consequences of policy actions in any policy reaction function. Reduced form econometric methods are unable to discriminate empirically between an interest rate policy reaction function (an antiin¯ation feedback rule) on the one hand and the in¯ationary consequences of the interest rate policy on the other. The potential impact of simultaneity is addressed by estimating bivariate VARs in ®rst differences for in¯ation and interest rates to apply a methodology proposed by King and Watson(1997). Such a methodology allows us to evaluate the results of all possible identi®cation schemes in bivariate VAR models. Imposing that monetary policy has long and variable outside lags ± i.e. in¯ation initially responds moderately to interest rate changes ± and that monetary policy is neutral in the long-run ± i.e. a long-run Fisher effect exists ± it is found that monetary policy of the Bundesbank is unlikely to be adequately described by an interest rate smoothing rule. In fact, a long-run unit effect of in¯ation on interest rates for Germany is compatible with no short-run policy reaction of interest rates to in¯ation at all. The same prior beliefs leads to the conclusion that for the United States only a very large immediate response of interest rates to in¯ation would be consistent with a long-run neutral monetary policy stance. The author concludes that these results point clearly towards a problem of lack of exogeneity in single-equation speci®cations for monetary policy rules. On the issue of misspeci®cation, it is argued that it is impossible to discuss European interest rate policy rules in isolation. I shall introduce the discussion by proposing ®rst a general framework to understand identi®cation and speci®cation of monetary policy rules. Within such a general framework, I shall address both general and speci®c issues. 159
160 Carlo A. Favero
1 A general framework to understand monetary policy rules: the simplest model of the monetary transmission mechanism Identi®cation and speci®cation problems for monetary policy rules are best understood by brie¯y outlining their derivation. I shall consider here one of the simplest possible models, taken by Svensson (1998). The central bank faces the following intertemporal problem: Et
1 X
i Lti
i0
L
1
t � 2 2
D 4:1
where Et denotes expectations conditional upon the information set available at time t , is the relevant discount factor, t is in¯ation at time t and is the target level of in¯ation. The central bank is de®ned as a strict in¯ation targeter. As the monetary instrument is the policy rate, it , the structure of the economy must be described to obtain an explicit form for the policy rule. We consider the following speci®cation for aggregate supply and demand in a closed economy: xt1 x xt � r
it � Et t1 � r z zt udt1 t1 t x xt z zt ust 1
D 4:2
D 4:3
where x represents deviations of output from its natural level and is the rate of in¯ation; zt is a vector of exogenous variables. The following interest rate rule can be derived from the ®rst-order conditions for optimality: 1 x r
D 4:4 it r
Et t1 � x r
x z z xt zt Et zt1 r r x r
A number of comments on
D4:4 are in order:
. Fitted parameters are convolutions of the parameters describing central
banks' preferences( ) and those describing the structure of the economy
x ; r ; x ; r : . The economy is best described by a three-equation model (under the assumption of valid exogeneity of the variables included in zt : Identi®cation of parameters in the monetary policy rule is achieved by the blockrecursivity assumption: interest rates react simultaneously to all the variables
Discussion 161
in the economy, but the reaction of in¯ation and output to monetary policy takes at least a one-period lag . In order to make
D4:4 consistent with the data, sluggish adjustment of actual to desired policy rate has to be imposed (see Clarida, Gali and Gertler, 1998). . Only one empirical implication can be confronted with the data by estimating the monetary policy rule as a single equation ± namely, whether the parameter describing the reaction of policy rates to a gap between current and target in¯ation is larger than one.
2 Some considerations on identi®cation and speci®cation problems of monetary policy rules The two issues raised by Weber in Chapter 4 are easily understood within the framework proposed in Section 1. The identi®cation problem can be referred to the estimation of the monetary policy rule as a single equation, when the data are generated by a multi-equation model. A related issue is the validity of the block-recursivity assumption. In general, it is absolutely true that the estimation of the full model is much more informative than a single-equation speci®cation of the monetary policy rule. In fact, tests of the compatibility of actual behaviour of central banks with different type of in¯ation targeting models can be conducted only by checking cross-equation restrictions on parameters of multi-equation models. However, the two-equation speci®cation for in¯ation and interest rates adopted in Chapter 4 is also a misrepresentation of the full model, in that it leaves at least one equation out of the picture. Therefore, the validity of the Weber critiques on identi®cation might be very limited if the omission of the equation for output leads to invalid estimation and inference in the twoequation model. The point of misspeci®cation as a consequence of interdependence of European interest rates is easily interpreted as a misspeci®cation of the vector zt , in the case where the German interest rates are exogenous. If this does not happen, then one more dimension has to be added to the problem by considering in¯ation targeting in open economies. As a consequence, the full model of the economy includes at least four equations, and misspeci®cation problems induced by considering just two of them are likely to become more severe. Some speci®c issues on the econometric approach Weber estimates bivariate VARs in ®rst differences for in¯ation and interest rates using the methodology in King and Watson(1997) to assess the impact of different identifying assumptions on tests of the relevant hypotheses. We have already pointed out a potential misspeci®cation induced by considering bivariate models. As well explained in Chapter 4, this choice is
162 Carlo A. Favero
driven by the `curse of dimensionality problem', which affects the methodology. Leaving aside this problem, it is probably worth considering a technical issue related to the speci®cation of VARs in ®rst differences. In general, the application of the King and Watson (1997) methodology requires the ability to compute ®nite cumulative impulse responses on which the long-run restrictions are imposed, and therefore the speci®cation of stationary VARs. Stationarity is achieved in Chapter 4 by ®rst-differencing. However, if series are cointegrated, then VAR in ®rst differences delivers biased estimates of the parameters of interest. Interestingly, the long-run restrictions considered in the two main sections of the Chapter are the Fisher effect and the uncovered interest parity (UIP). Both relationships deliver cointegrating vectors, between nominal interest rate and in¯ation and between domestic and foreign interest rates, respectively. It might then be more appropriate to achieve stationarity by considering an error-correction speci®cation. � Consider, for simplicity, the case of a bivariate model yt yt ; xt ; in which variables are nonstationary I(1) but cointegrated with cointegrating vector
1; �1: yt v1t 11 yt�1 b11 b12
1 � 1
D 4:5 xt 21 xt�1 b21 b22 v2t A VAR in ®rst differences is clearly misspeci®ed because it omits the variable capturing the long-run relationship between x and y. However, model
D4:5 can be rewritten as follows: �1 1 11 0
yt�1 � xt�1 1� L 0
yt � xt 0 1 xt 21 0 xt�1 0 1 v1t b12 b 11
D 4:6 b21 b22 v2t The two representation are absolute identical (same residuals). The cointegrating properties of the system suggests the presence of two types of shocks: a permanent one (to be related to the single common trend shared by the two variables) and a transitory one (to be related to the cointegrating relation). It seems therefore natural to identify one shock as permanent, the other as transitory. Given that we have a stationary system, the identi®cation of shocks is obtained by deriving long-run responses of the variables of interest to relevant shocks. From
D4:6 we have: �1 1 1� L 0
yt � xt 11 L 0 � 0 1 0 1 xt 21 L 0 b11 b12 v1t
D 4:7 b21 b22 v2t
Discussion 163
From which long-run responses are obtained by setting L 1 and by inverting the matrix, premultiplying variables in the stationary representation of VAR:
yt � xt xt
yt � xt xt
�11 �21
1 1
�1
�b11 b21 11 �21 �21 b11 11 b21 11 �21
b11 b21
b12 b22
�b22 -b12 11 �21 -21 b12 11 b22 11 �21
!
v1t v2t v1t v2 t
D 4:8
D 4:9
so v2t can be identi®ed as the transitory shock by imposing the following restriction: �21 b12 11 b22 0 which, given knowledge of the -parameters from cointegration analysis, provides the just-identifying restriction for the parameters in B1 . This little example illustrates that speci®cation in ®rst differences might be a source of mis-speci®cation and that the presence of cointegration between the series of interest leads naturally to a set of speci®c identifying restrictions, which could be exploited to limit the effects of the curse of dimensionality. To conclude, we agree with the main message of this chapter, which invites us to consider monetary policy rules in the context of macromodels rather than in isolation. Bivariate models speci®ed in ®rst differences are a ®rst step in this direction; it would be desirable to extend such framework to achieve a closer relation between theoretical and applied models and to avoid misspeci®cation of applied models owing to the omission of long-run equilibria. Note 1. Interestingly there is one case in which such structural identi®cation is achieved thorugh the standard Choleski decomposition: this is the case in which yt is weakly exogenous for the estimation of b21 :
References Clarida R., J. Gali and M. Gertler (1998) `Monetary Policy Rules in Practice:some International Evidence', European Economic Review, 42, 1033±68. Svensson L. E. O. (1998) `Monetary Issues for the ESCB', paper presented at the Carnegie± Rochester Conference; available at < http://www.iies.su.se/leosven/ >. King R. G. and M. W. Watson (1997) `Testing Long-run Neutrality', Federal Reserve Bank of Richmond Economic Quarterly, 83, 69±101.
Discussion
Philippe Moutot
Over the past few years, the estimation of policy reaction functions has become quite popular. Nevertheless, studying empirical literature often seems much like a `®shing expedition'.2 The Weber's study in Chapter 4 adds some interesting results to an important issue, namely the issue of robustness. The chapter argues that, on the one hand, it is not possible to disentangle the determinants and consequences of monetary policy without prior restrictions, a fact which is well known in econometrics as the `identi®cation problem'. On the other hand, it is stated that it is impossible to analyse the interest rate policy rules of the European countries in isolation, owing to the interest rate linkages via the ERM. Both problems are not new and ± as far as I can see ± unsolved in literature so far. My comments will be threefold. I will ®rst argue that the kind of issues tackled in this chapter and the techniques used for this purpose are increasingly relevant, at least from the point of view of a central banker. I will then discuss the results of the chapter to see what lessons might be learned from it with regard to the identi®cation of rules, submitting that this chapter should only be the ®rst step of a long-term research project. I would, however, caution against the ultimate hope that rules could fully replace discretion in the context of central banking.
1
Both the issues raised and the techniques used are relevant
The issues In today's empirical literature, VAR and Vector Error Correction (VEC) models have become increasingly popular tools. Unfortunately, and as is also well known, such reduced form models per se do not provide us with any knowledge about causality. Following this line of reasoning, traditional estimates of the monetary policy reaction ± e.g. those referred to as `Taylor Rules' ± cannot in most cases discriminate between the reaction function ± e.g. monetary policy changes in response to in¯ation, and the transmission process running from the interest rate to in¯ation, owing 164
Discussion 165
to the joint interaction of all variables being under investigation. Therefore, the very use of such reaction functions by central bankers either as actual rules ± or, more realistically, as possible references for action ± is questionable. This is all the more true given that, ex ante, the general public may have held different views of the action of monetary authorities, implying that, once implemented, the rule could actually lose its signi®cance, another case of `Goodhart's Law'. The issue is relevant not only in the case of VARs and VECs, but also in the case of large econometric models, where the estimation of structural parameters is often very much dependent on the underlying assumptions concerning the reaction functions of authorities. Therefore, determining a rule by optimising a loss function on the basis of a structural model, estimated without adequately facing the `observational equivalence' issue, is also problematic. Finally, the European case is even more complex than the cases of other countries in such respects. Owing to monetary union, the monetary regime changed drastically at the start of 1999, so that the past reaction functions of monetary authorities are an uncertain guide. Moreover, owing to the asymmetric functioning of the EMS, an average of these reaction functions is likely to be of uncertain interest, given that even the nature of variables within these putative reaction functions may have varied across countries. Taking into account the past nature of reaction functions in the EMS countries is therefore advisable for any econometric analysis of Euro Area data which attempts to draw inference of relevance to the European Central Bank (ECB). This is certainly an issue which the Eurosystem itself considers today in the construction of its econometric models. In this context, a thorough investigation of the leadership role of the Bundesbank in the determination of European interest rates before monetary union is certainly relevant, given that this constitutes a necessary ®rst step to model these monetary regimes, although the conclusions of the analysis have been clear for a long time, at least to European central bankers. In the next step of his research, I would encourage A. Weber to extend his analysis towards an investigation of the relationships among in¯ation, the short-term interest rate and other key economic variables such as output or the slope of the yield curve in a simultaneous framework. The techniques One way of dealing with the issue is to apply the Choleski decomposition which is widely used in the older literature and which requires an ordering of the variables. Owing to its rather mechanical nature, this procedure is increasingly being viewed as unsatisfactory. Another possibility for overcoming this problem is to be found in a structural approach ± i.e. in imposing restrictions either on the short-run or on the long-run coef®cients in the variance±covariance matrix.3 Inference can then be carried out in two ways: one way could be to choose an identi®cation
166 Philippe Moutot
scheme (in fact re¯ecting the researcher's a priori economic assumption about the nature of the shocks and therefore also about causality) and subsequently to derive the corresponding point estimates. Another way could be to impose several identi®cation schemes and to deduce whether the values of the parameters of interests are in line with them. The second alternative has been used in the well-known study by King and Watson (1997),4 that provides a rather ¯exible and elegant scheme for imposing and testing identifying restrictions and which incidentally also coincides with the spirit of Edward Leamer's early work.5 This is also the alternative chosen by Weber. In the end, Weber estimates its parameters by imposing a restriction that is compatible both with economic theory and with the data. Given the generally low power of the stationarity tests used by Weber to test for co-integration in small samples, the use of VEC rather than VAR models might have been considered.
2
Interpreting the results from a central banker's point of view
Following this line of reasoning, Weber investigates a number of interesting hypotheses related to interest rate functions. The long-run neutrality of money In the ®rst part of Chapter 4 Weber addresses the questions of the long-run neutrality of money (e.g. the hypothesis that nominal shocks have no in¯uence on real variables) and the long-run homogeneity of prices (e.g. the hypothesis that nominal interest rates and in¯ation must change proportionately in response to a nominal shock, the `Fisher effect') for a number of countries. As a consequence of his setup, it turns out that, once in¯ation is endogenised in the model and the usual identi®cation procedure is carried out, the model is no longer identi®ed. In other words: an additional restriction is needed to estimate the model. Weber therefore imposes a long-run correlation of one between nominal rates and in¯ation and shows in the next step that the Fisher effect is supported by data only under the assumption that the short-term impact of in¯ation on the interest rate and vice versa belong to a bounded set (see Figure 4.3). However, since the resulting parameter values are in line with the ones given by the simple (unrestricted) VAR, the hypothesis seems to be in line with the data.6 All in all, Weber does not reject the Fisher effect for the United States, Germany, Canada and Japan. Let me turn back, however, to the more general lessons that can be drawn from this exercise. First, indeed, the Fisher effect seems to be borne out by the data. However, the estimation of the rule and of the corresponding structural model cannot be made with great precision. In most cases, the hypothesis of a rule combined with a well functioning economic structure is reasonable. But it cannot be
Discussion 167
identi®ed with full precision. For each assumption on the short-term impact of in¯ation on interest rates, there is a corresponding separate estimate of the interest rate rule. Second, the hypothesis of a rule combined with a well functioning economic structure cannot always be con®rmed. This is obviously the case with the United Kingdom where some more detailed analysis needs to be carried out. One may wonder if the time series is not too long in that case, mixing too many different monetary regimes and therefore being misleading. Somehow, con®rming a rule also depends on the actual period on which it can be observed. Third ± and this is meant to be a more philosophical conclusion ± it is not at all sure whether it is possible for a new central bank like the ECB to ®nd optimal rules `ex ante'. Owing to this issue of `observational equivalence', some `learning by doing' is unavoidable. Credibility is of the essence in such a process. And the way credibility in¯uences the functioning of the economy is likely to depend on the nature of the rule itself. This is a point which becomes even more clearly visible when considering the second type of issue tackled by Weber. The leadership of German interest rates in the EMS Weber then tests for the long-run interactions between German interest rates and the interest rates of other European as well as some G-7 countries by taking the difference between the immediate and long-run transmission explicitly into account once again. In sum, he ®nds evidence in favour of an asymmetrical EMS with Germany as the anchor country. This can be seen from the long-run interest rate effects (see Table 4.5, p. 152). Concerning the case of France, however, I would like to add some considerations that might complement the interpretation slightly. Owing to the EMS constraint, French short-term interest rate changes have indeed generally followed German ones in the short run. However, as the credibility of the exchange rate peg improved, it was in some occasions possible for the Banque de France to avoid or only partially follow some upward movements of German rates or to exceed some of their downwards movements. Thus the French short-term rate would converge gradually to the German one. However, this was possible only when the economic situation ± and, in particular, the growth of output ± made it clear that the French authorities would not be faced with any serious dilemma in the case of a speculative attack on the EMS. Moreover, when growth was expected to dip, speculative attacks could ¯are up leading the short-term rates and, at the start of the 1980s, also the long-term rates to higher levels. Therefore, it might be expected that the slope of the French yield curve would show a more direct link to the output gap than the German one. As mentioned above, extending the system to additional variables like output and the yield curve could provide us with additional insights. Again, the credibility of monetary policy seems to have been of the essence.
168 Philippe Moutot
3
Rules can limit, but not replace discretion
To conclude, progress can be made in a number of areas to better identify rules useful to monetary authorities. In Chapter 4 Weber has initiated a useful research programme and is trying to tackle essential issues. But the hope that rules can, as a result of such analysis, replace discretion should not be overestimated. Going beyond the problems of `observational equivalence' mentioned above, it should be recognised that the rules estimated on the basis of central banks' past behaviour usually contain an auto-regressive pattern. In other words: without adding, though usually without a strong economic justi®cation, an auto-regressive part to the in¯ation and output gap variables, a `bad ®t' would result. Simultaneously, rules deduced from the optimisation of a loss function applied to a structural model usually yield optimal reaction functions, which create an excessive degree of interest rate volatility. Therefore, a good explanation to the inertial behaviour of central banks is needed. A possible explanation is given by Woodford (1998), who indicates that an optimal central bank should be able to commit itself over the long term to best in¯uence long-term rates, thereby needing inertia. This is in line with earlier remarks by Goodfriend (1991). Moreover, it is consistent with the conclusion I drew earlier, according to which credibility and its evolution are essential variables which must be taken into account. However, credibility is always the combination of two elements. Credibility entails, ®rst, doing what one promised to do. But it also implies that what is done is generally viewed as reasonable by the public. Ful®lling these two conditions in a changing world, the economic model of which is both changing and only partially known, implies the following: while making good on its commitments, the central bank needs to adapt its view of the economy and its explanation of the future, whenever needed. While the analysis of commitments is increasingly well studied by academic literature, the analysis of this discretionary need to adapt models and views is usually ignored. However, while continuously listening to new ideas developed by academics and market people, the central bank needs to pass judgement on whether such new views should be integrated in its analysis. This is essential to offer the best monetary policy possible to the public. The central bank therefore needs discretion to adapt its analysis to new developments. In view of current uncertainty on the way both the global economy and the Euro Area economy function, such a need for discretion should not be overlooked. A ®xed monetary policy rule is therefore not reasonable. However, the relevant issue in this context is how to combine the need for some limited the public. discretion with the need for transparency vis-a-vis The best solution for a central bank facing this issue is, in my view, to use one or several rules as reference points for its action and to endeavour to explain to the public the deviations from such rules in a systematic manner. In this view, describing the particular changes in indicators or in views that motivate such
Discussion 169
deviations seems to be at least as important for transparency as providing the overall view of the economic prospects that synthesise the decisions made. This is what the ECB has chosen to do. By mentioning that the ®rst pillar of its strategy is a reference value for a monetary aggregate, it made clear its strong prior conviction that in¯ation is always, in the end, a monetary phenomenon. Simultaneously, by indicating that its strategy has a second pillar ± the outlook for price developments ± it also showed that the model of the economy it has in mind may evolve over time and that it would also use a wide set of analyses and indicators, including forecasts, to motivate its decisions. Further, its desire to ®nd the basis for this outlook on both indicators and forecasts without giving an excessive role to such forecasts is consistent with the need to give the best possible justi®cation to and the best possible account of the unavoidable discretion entailed by good monetary policy making.
Notes 1. Prepared by Philippe Moutot, with the help of Dieter Gerdesmeier. 2. Shiller (1990), p. 667. Shiller was actually referring to the estimation of the determinants of the term structure, but his statement seems to be valid for our considerations as well. 3. See Sims (1986) and Bernanke (1986) for example. 4. King and Watson have also used the framework for the estimation of the long-run Phillips Curve (i.e. the relation between in¯ation and unemployment) which is of particular interest for a central bank. 5. See Leamer (1978) and (1983) for a more detailed overview. 6. It should be stressed that a simple correlation between the residuals in the simple VAR can easily be calculated. However this does not say anything about the causality. This question cannot be answered without imposing identifying restrictions and testing for them.
References Bernanke, B. (1986) `Alternative Explanations of the Money-Income Correlation', Carnegie±Rochester Conference Series on Public Policy, 25, 49±99. Goodfriend, M. S. (1991) `Interest Rates and the Conduct of Monetary Policy', (Carnegie± Rochester Conference Series on Public Policy, 34, 7±30. King, R. G. and M. W. Watson (1997) `Testing Long-run Neutrality,' Federal Reserve Bank of Richmond Quarterly, 83(3), 69±101. Leamer, E. E. (1978) Speci®cation Searches: Ad hoc Inference with Non-experimental Data, New York: John Wiley. ÐÐÐÐ`Let's Take the Con out of Econometrics', American Economic Review, 73, 31±43. Shiller, R. J. (1990) The Term Structure of Interest Rates, in [Friedman, B. M., and Hahn, F. H. (eds.), Handbook of Monetary Economics, vol. I, Elsevier Science Publishers B.V., pp. 627±722., 1990]. Sims, C. (1986) `Are Forecasting Models usable for Policy Analysis?', Federal Reserve Bank of Minneapolis Quarterly Review, 10, 2±16. Woodford, M. (1998) `Optimal Monetary Policy Inertia', paper presented at the CFS conference in Frankfurt, October.
5
Legal Structure, Financial Structure and the Monetary Policy Transmission Mechanism1 Stephen G. Cecchetti
1
Introduction
Over the past decade, the countries of central Europe have become more alike in many ways. As the new members of the European Monetary Union (EMU) prepared for the birth of the Euro on 1 January 1999, their economic policies became substantially more uniform. All eleven countries in the new Euro Area had virtually eliminated in¯ation and taken serious steps toward ®scal consolidation.2 As their monetary and ®scal policies have adjusted to meet these common goals, the countries' business cycle ¯uctuations appear to have become more synchronised as well.3 While this makes the job of the European System of Central Banks (ESCB) easier, numerous dif®cult challenges remain. Primary among these is the making of policy in the face of the possibility that it will have differential impacts across the countries of the Euro Area. The task facing the ESCB is even more complex than that facing countries with stable monetary regimes, where the measurement of the national and regional impact of policy has already proved to be extremely dif®cult. The creation of the ESCB constitutes a `regime shift' in virtually every sense of the term. The introduction of the Euro seems sure to prompt adjustments in the economies of the member countries, and these adjustments will probably alter the relationship between the actions of the central bank and the real economy. That is, the monetary transmission mechanism of the countries in the Euro Area will change, making the job of the new European Central Bank (ECB) even more dif®cult than it is already. But how quickly will it change, and what will it become? To answer these questions, we must understand the fundamental determinants of the impact of policy actions on output and in¯ation. For insight into these determinants, I turn to the modern views of the monetary transmission mechanism, which assign a central role to ®nancial structure. Kashyap and Stein (1997) provide a starting point; they focus on the importance of the banking system and go on to emphasise the distributional effects of monetary policy changes. The conventional wisdom has always been that some industries are more sensitive to interest rate changes than others, and so 170
Legal Structure, Financial Structure and Monetary Policy Transmission 171
changes in policy-controlled interest rates have differential effects across industries. The view based on ®nancial structure both formalises this reasoning and takes it one step further by noting that some ®rms are more dependent on banks for ®nancing than others, and that this is true both across and within industries. According to this `lending view' of the transmission mechanism, monetary policy actions change the reserves available to the banking system, thereby affecting the willingness of banks to lend ± and, ultimately, the supply of loans. How this mechanism will affect individual ®rms depends on the ®nancing methods available to them. Monetary policy has a bigger impact on ®rms that are reliant on banks for their ®nancing. Furthermore, healthier banks will be able to adjust to the policy-induced reserve changes more easily than other banks will. The distributional effects implied by the lending view of monetary policy transmission have clear implications for the Euro Area and the ESCB. Countries in which ®rms are more bank dependent and banking systems are less healthy will be more sensitive to the Eurosystem's decisions to change interest rates. This brings me to the ®rst question I will address in this chapter: is there evidence that the impact of monetary policy innovations varies across countries with the strength and scope of the banking system? With this in mind, I examine differences in the size, concentration and health of national banking systems, as well as in the availability of non-bank sources of ®nance. I ®nd, consistent with the most casual observation, that banking system characteristics vary dramatically across the countries of the European Union (EU). Furthermore, these differences do seem to be related to estimated differences in the impact of interest rate changes on output and in¯ation. Countries with many small banks, less healthy banking systems and poorer direct capital access display a greater sensitivity to policy changes than do countries with big, healthy banks and deep, well developed capital markets. But this is just the ®rst question. The more important issue facing the ESCB is whether the national banking systems, and the implied sensitivity of each country's real economy to monetary policy shocks, will change now that there is monetary union. It is easy to assert that European banks will soon look like US banks, exhibiting a ®nancial structure and transmission mechanism similar to American models. After all, the Euro Area does resemble the United States, at least super®cially. It has a slightly larger population ± 292 million for the eleven members of the monetary union relative to 270 million for the United States ± and nearly as high a level of GDP ± $6.8 trillion compared with $8.1 trillion in 1997. The Euro Area also has a similar degree of openness to trade, with imports accounting for slightly more than 10 per cent of GDP. These parallels, along with the fact that ®nancial technology is easily transferable across national boundaries, have led a number of observers to conclude that the introduction of the Euro may act as a catalyst, speeding the rate at which ®nancial relationships in Europe become like those in the United States. For
172 Stephen G. Cecchetti
example, while Dornbusch, Favero and Giavazzi (1998, pp. 48±9) do note the possibility for EU-wide asymmetries resulting from differences in ®nancial structure, they assert that `the euro will change the way ®nancial markets work, inducing corresponding changes in the monetary mechanism. In addition to pervasive deregulation already under way and innovation, the introduction of the euro will revolutionize the ®nancial structure of Europe. Europe will in a short period become more nearly like the USA.' McCauley and White (1997, p. 17) suggest that there may be an acceleration in the rate at which securities replace loans on the asset side of bank balance sheets and commercial paper replaces deposits on the liability side. They point to a `dramatic potential for assets to be stripped out of the banking system' and for securities markets to absorb as much as one-third of the corporate loans now originated in European banks.4 Overall, these commentators are speculating that the increased liquidity of European ®nancial markets brought about by monetary union will lead to signi®cant consolidation of banks, with mergers at both the national and the international level, as well as a direct substitution of traded equities and bonds for bank loans. Why should we believe that the European ®nancial structure will quickly be transformed into one that mirrors the one in the US model? Without an explanation for the evolution of these countries' national ®nancial structures that is based on their existing differences, such claims are unconvincing. What accounts for the variation in the ®nancial intermediation systems across countries? Traditionally, we look to taxes and regulation for an explanation, and Dornbusch, Favero and Giavazzi (1998) as well as White (1998) do mention these. Danthine et al. (1999) identify a number of barriers to change in national ®nancial structures and note the importance of the historical path that has brought each country's banks to their current state. Danthine et al. (p. 45) then go on to assert that `legal differences between EU states, in particular the lack of some form of ``European corporate law'', also remain important and constitute an additional factor of market segmentation'. Such disparities in legal structure can explain important economic patterns, and they can be maintained for long periods of time, signi®cantly delaying the harmonisation of national banking systems.5 It is my main contention that the differences in ®nancial structure across the countries of Europe are a consequence of their dissimilar legal structures. My argument draws on the work of La Porta et al. (1997, 1998), who focus on the relationship between legal structures and ®nance. They argue that the structure of ®nance in a country depends on the rights accorded shareholders and creditors by the laws of that country, as well as on the degree to which these laws are enforced. The nature of the laws is, in turn, a product of the legal tradition on which the civil codes of a country are based. La Porta et al. establish that the character of a country's ®nancial markets depends on the country's legal structure. Putting their arguments together with the lending view of the monetary transmission mechanism leads to the possibility that it is
Legal Structure, Financial Structure and Monetary Policy Transmission 173 Table 5.1 Effectiveness of 14cy and the origins of the legal system
Legal family English Scandinavian French German
Index of effectiveness of monetary policy 1.1 1.8 2.1 2.4
Impact of policy On output
on in¯ation
±0.45 ±0.52 ±0.70 ±1.25
±0.21 ±0.22 ±0.20 ±0.49
Notes: Index of effectiveness of monetary policy, from Table 5.5, is based on ®nancial structure variables described in Section 3 of the text, with higher values implying a larger expected impact of interest rate changes on output and prices. The impact of policy on output and in¯ation, from Table 5.6, is a measure of the maximum response, in percentage points, to an interest rate movement of 100 basis points, estimated using a small-scale structural model. Countries are classi®ed by the origin, or family, of their legal structure, and group means are reported based on data for Ireland, the United Kingdom and the United States (English common law); Denmark and Sweden (Scandinavian common law); Belgium, France, Italy, Portugal, and Spain (French civil law); and Germany (German civil law).
the legal system in a country that forms the basis for the structure of ®nancial intermediation and, hence, for the impact of monetary policy on output and prices. Table 5.1 reports the empirical ®ndings that support the basic conclusion of the chapter. After classifying countries by the origin, or `family' of their legal structure, I calculate for each family the average level of an index of monetary policy's likely effectiveness (based on banking system size, concentration and health, with a higher value implying greater effectiveness) and the estimated impact of an interest rate change on output and in¯ation (from a small-scale structural model). The results suggest that a country's legal structure, ®nancial structure and monetary transmission mechanism are interconnected. The clear pattern is that the predicted effectiveness and its measured impact vary systematically based on the origin of a country's legal system. Countries with better legal protection for shareholders and debtors (countries with a legal structure based on English common law) have ®nancial structures in which the lending channel of monetary transmission is expected to be less potent; for these countries, the measured impact of an interest rate change on output and in¯ation is lower. The implication is that unless the laws governing shareholder and creditor rights and the enforcement of those laws are harmonised across the members of EMU, monetary policy will continue to have a differential impact. Put slightly differently, it is my belief that the ®nancial structures in the countries of the Euro Area will not converge into one large US-style system unless there are dramatic legislative changes. If such legal harmonisation occurs ± that is, if the civil codes protecting shareholders and creditors are made uniform across
174 Stephen G. Cecchetti
the countries that have entered the monetary union ± then the regional variation in the impact of interest rate changes on output and in¯ation should decrease.6 But if legal convergence does not occur, ®nancial structure will remain heterogeneous, and so will the monetary transmission mechanism, and the job of the Eurosystem will be to construct an appropriate policy that takes these asymmetries into account.7 The remainder of this chapter provides the building blocks for this argument. In Section 2, I provide a brief survey of the theories of the monetary transmission mechanism, focusing on the importance of ®nancial structure to an understanding of monetary transmission. Section 3 assesses the national banking systems, including measures of overall size, concentration, health and the relative importance of non-bank ®nance. Overall, this analysis allows me to evaluate the likely strength of the lending channel across countries. Section 4 reports estimates, for a set of ten countries, of the impact of an interest rate increase on output and in¯ation. These estimates follow the pattern that is expected: countries where ®nancial structure data suggest that the lending channel should be strong exhibit more sensitivity to monetary policy movements. In Section 5, I present the data and arguments from La Porta et al. (1997, 1998) on the relationship between legal and ®nancial structures. This allows me to test the prediction that countries with poor shareholder and creditor protections and poor law enforcement will have less developed ®nancial systems and greater sensitivity of output and in¯ation to interest rate changes. While far from being de®nitive, the results are consistent with my main hypothesis: differences in legal systems give rise to variations in national ®nancial structures, and these variations in turn lead to divergences in monetary transmission mechanisms. So long as the legal systems of the Euro Area countries remain distinct, the impact of interest rate changes across these countries will differ.
2
Theories of the transmission mechanism
A number of comprehensive surveys of the theories of the monetary transmission mechanism have appeared in recent years. These include Bernanke (1993), Gertler and Gilchrist (1993), Kashyap and Stein (1994, 1997), Hubbard (1995), and my own survey, Cecchetti (1995). As a result, I will be brief. All theories of how interest rate changes affect the real economy share a common starting point. A monetary policy action begins with a change in the level of bank reserves. For this to have any real effects at all, there must be nominal rigidities in the economy. Otherwise, a change in the nominal quantity of outside money cannot have any impact on the real interest rate. While the ability of the central bank to change the level of bank reserves is not in question, the source of the nominal rigidity that allows the change in reserves to alter short-run real rates of return has been under debate for decades. The current state of this discussion is well summarised by Christiano,
Legal Structure, Financial Structure and Monetary Policy Transmission 175
Eichenbaum and Evans (1997). They distinguish three sets of theories: one set based on sticky wages, a second based on sticky prices and a third built on the idea of limited participation. The sticky-wage and sticky-price models, which are the most familiar, rest on the idea that there are costs to nominal price and wage changes, and so adjustments are infrequent. In limited participation models, introduced in Rotemberg (1984), individuals (households) are unable to adjust their cash balances suf®ciently rapidly in response to changes in the environment ± that is, households have a limited ability to participate in ®nancial markets, and so must commit themselves to certain portfolio holdings for relatively long periods of time.8 The sources of nominal rigidities are relatively unimportant for the discussion of the mechanism by which interest rate changes have short-run effects on output and prices, and so I will move directly to a discussion of the current theories of the transmission mechanism.9 Our current views are based on the work of Bernanke (1983), Bernanke and Blinder (1992) and Bernanke and Gertler (1989, 1990). These authors distinguish between the traditional money view, in which interest rate movements affect the level of investment and exchange rates directly, and the lending view, in which ®nancial intermediaries play a prominent role in transmitting monetary impulses to output and prices. I will describe each of these in turn. The traditional view, which is largely the foundation for the textbook IS ± LM model, is based on the notion that reductions in the quantity of outside money raise real rates of return. This outcome has two effects, the ®rst directly from interest rates to investment and the second through exchange rates. An interest rate increase reduces investment, as there are fewer pro®table projects available at higher required rates of return. A policy action induces a movement along a ®xed marginal-ef®ciency-of-investment schedule. This interest rate channel will be more powerful the less substitutable outside money is for other assets. The exchange rate channel is also familiar from textbook models. Here, an interest rate increase results in a real appreciation of the domestic currency, reducing the foreign demand for domestically produced goods. Regardless of whether the transmission mechanism occurs through the interest rate channel or the exchange rate channel, there is no real need to discuss banks. In fact, there is no reason to distinguish any of the `other' assets in investors' portfolios. This is a simple two-asset model. An important implication of this traditional model of the transmission mechanism concerns the incidence of the investment decline. Since there are no externalities or market imperfections, only the least socially productive projects ± those with the lowest rates of return ± go unfunded. As a result, the capital stock is marginally lower, but, given that a decline is going to occur, the allocation of the decline across sectors is socially ef®cient. As most of the surveys cited earlier emphasise, the lending view has two parts ± one that focuses on the impact of policy changes on borrower balance sheets and another that focuses on bank loans. In both, the effectiveness of
176 Stephen G. Cecchetti
policy depends on capital market imperfections that make it easier for some ®rms to obtain ®nancing than others. Information asymmetries and moral hazard problems, together with bankruptcy laws, mean that the state of a ®rm's balance sheet has implications for its ability to obtain external ®nance.10 By reducing expected future sales and by increasing the cost of rolling over a given level of nominal debt, policy-induced increases in interest rates (which are both real and nominal) cause a deterioration in the ®rm's net worth. Furthermore, there is an asymmetry of information in that borrowers (®rms) have better information about the potential pro®tability of investment projects than do creditors (banks). As a result, as the ®rm's net worth declines, the ®rm becomes less creditworthy because it has an increased incentive to misrepresent the riskiness of potential projects ± an outcome that will lead potential lenders to increase the risk premium they require when making a loan. The asymmetry of information makes internal ®nance of new investment projects cheaper than external ®nance. More important for the transmission mechanism per se is that some ®rms are dependent on banks for ®nance and that monetary policy affects bank loan supply. A reduction in the quantity of reserves forces a reduction in the level of deposits, which must be matched by a fall in loans. Nevertheless, lower levels of bank loans will have an impact on the real economy only insofar as there are ®rms without an alternative source of investment funds. Substantial empirical evidence supports the importance of both capital market imperfections and ®rm dependence on bank ®nancing. Kashyap and Stein (1997) provide a summary of two types of studies. The ®rst type suggests that banks rely to a large extent on reservable-deposit ®nancing and that, for this reason, a contraction in reserves will prompt banks to contract their balance sheets, reducing the supply of loans. The second type establishes that there are a signi®cant number of bank-dependent ®rms that are unable to mitigate the shortfall in bank lending with other sources of ®nance. Overall, recent research does imply the existence of a lending channel.11 Models of monetary policy transmission based on ®nancial structure suggest a natural place to begin looking for sources of cross-country differences in the monetary transmission mechanism. The prediction is that, overall, the transmission mechanism will be stronger in those countries where ®rms are more bank dependent, and where the banking system is less healthy and less concentrated. In the ®rst instance, ®rms that have less direct access to capital markets are unable to blunt the effect of a contraction in bank loans. In the second, banks themselves have restricted access to non-reservable deposits and are forced to contract their balance sheets by more for a given change in policy. In Section 3, I examine data on national ®nancial structure and try to rank countries based on the likely strength of the transmission mechanism. To the extent that these cross-country differences are present, then the lending view implies that they will persist until the ®nancial structures become more uniform.12
Legal Structure, Financial Structure and Monetary Policy Transmission 177
3 Likely strength of the transmission mechanism In assessing the likely impact of an interest rate change on output and prices in the various countries of the EMU, I follow the work of Kashyap and Stein (1997) and assemble data on the size and concentration of the banking systems, along with measures of banking system health, the importance of bank ®nancing and the size of ®rms. The indicators are chosen to conform as closely as possible to the economic quantities that the lending view suggests should be important. The balance sheets of large, healthy banks are not as sensitive to policy, because reserve contractions can be readily offset with alternative forms of ®nance that do not attract reserve requirements. In addition, I examine measures of the development of equity and debt markets in the EMU countries. Firms with ready capital market access, which are more likely to be found in countries with extensive secondary securities markets, will be better insulated from bank loan-supply contractions. Combining these measures, I construct an index of the probable strength of the monetary transmission mechanism.13 Table 5.2 Size and concentration of the banking industry, by country, 1996 Banks per million people (2)
Concentration ratios: top ®ve banks (3)
1019 140 350 1373 3517 62 937 172 51 313
126 14 68 24 43 18 16 11 5 8
48 57 78 40 17 41 25 79 76 44
Members of the EU not in EMU Denmark 117 Greece 20 Sweden 124 United Kingdom 478
22 2 14 8
17 71 90 28
Other countries Japan United States
4 40
30 17
Country
No. of credit institutions (1)
Monetary union members Austria Belgium Finland France Germany Ireland Italy Netherlands Portugal Spain
556 10 803
Note: Cobcentration ratios are calculated as the percentage of each country's bank assets accounted
for by the ®ve largest banks.
Sources: See the Data Appendix (pp. 189±90).
178 Stephen G. Cecchetti
To assess the importance of small banks in a country's ®nancial system, Table 5.2 reports the number of banks, the number of banks per million population, and measures of concentration for all of the EU countries plus Japan and the United States.14 The data reveal that Austria and Finland have many more banks per capita ± 126 and 68 per million people, respectively ± than any of the other countries. The remaining countries fall into roughly three groups: the United Kingdom, Japan and the southern European countries of Spain, Portugal and Greece have less than 10 banks per million; the United States and Germany have 40 or slightly more; and the remaining countries have between 13 and 25. Turning to the concentration measures in column (4) of the table, it is interesting to note that countries with more banks do not necessarily have less concentrated banking systems. France, for example, with 1373 banks and just under 60 million people, has a fairly high concentration ratio: the top ®ve French banks account for a sizable 40 per cent of total banking system assets and the top ten for nearly two-thirds. Overall, Denmark and Germany have the least concentrated banking systems in Europe. By contrast, large banks clearly dominate Sweden, Finland, Belgium, the Netherlands and Greece. The remaining countries are somewhere in between. What do these ®ndings imply for the strength of the transmission mechanisms in the countries examined? Austria, Germany and the United States have systems composed of a network of small banks, and so one would expect the lending channel to be relatively strong in those countries. At the other end of the spectrum, Belgium, Finland, Ireland, the Netherlands, Portugal, Sweden and the United Kingdom all have banking industries dominated by a small number of relatively large banks, with a modest periphery of small institutions. The remaining countries ± Denmark, France, Greece, Italy and Japan ± fall in a middle group. The weaker a nation's banking system, the stronger the expected impact of policy movements. With this in mind, I have collected a set of standard gauges of banking system health ± return on assets, loan loss provisions, net interest margin and operating costs ± and I have calculated a summary rating of overall system soundness (Table 5.3). Focusing primarily on the return on assets and the average Thomson ratings in Table 5.3 leads to the following rankings: Ireland, the United Kingdom and the United States have the healthiest banks; Austria, Belgium, Germany, the Netherlands, Spain, Denmark and Greece are second; Finland, France, Italy, Portugal and Sweden are third; and Japan is alone at the bottom. Finally, I turn to the availability of non-bank ®nance for ®rms in EU and other countries. The relevant data are reported in Table 5.4. Following Kashyap and Stein (1997) and La Porta et al. (1997), I examine the number of publicly listed ®rms, the extent of secondary equity and debt markets and the ratio of bank loans to all forms of ®nance. Although these are crude measures of access to external ®nance, they are informative. As in the case of Table 5.2, the
Legal Structure, Financial Structure and Monetary Policy Transmission 179 Table 5.3 Measures of banking industry health, by (country, 1996, per cent
Country
Return on assets (1)
Loan loss Net interest Operating Average Thomson provisions margin costs rating (2) (3) (4) (5)
Monetary union members Austria 0.38 Belgium 0.52 Finland 0.50 France 0.36 Germany 0.44 Ireland 1.57 Italy 0.33 Netherlands 0.75 Portugal 0.91 Spain 0.76
0.59 0.17 0.78 0.24 0.18 0.17 0.62 0.26 0.42 0.32
1.67 1.41 2.07 1.43 1.24 3.36 2.32 2.06 2.60 2.20
2.45 1.67 3.05 1.84 2.19 3.32 3.19 2.48 3.80 2.69
2.38 (4) 2.00 (6) 2.83 (3) 2.28 (16) 1.97(19) 1.83 (3) 2.57 (15) 2.10 (5) 2.30 (5) 1.79(11)
Members of the EU not in EMU Denmark 0.91 Greece 1.11 Sweden 1.28 United Kingdom 1.28
0.11 0.18 0.25 0.18
1.28 1.98 1.90 2.15
0.97 2.77 1.77 2.42
2.33 (3) 2.50(6) 2.50 (5) 2.04(23)
Other countries Japan United States
0.75 0.10
1.17 2.68
1.03 3.51
3.32(44) 1.73 (344)
0.01 1.42
Notes: Except for the Thomson ratings, all ®gures in the table are calculated as a percentage of total bank assets. In column (5), the number of banks rated by Thomson in each country and used to compute the average appears in parentheses. Sources: See the Data Appendix (p. 190).
countries can be divided into three groups. Austria, Ireland, Italy, Portugal and Greece appear to have the least well developed external capital markets. They have small equity and bond markets, and bank loans account for a high percentage of ®rm ®nancing. By contrast, Belgium, Denmark, Sweden, the United Kingdom and the United States all have substantial secondary capital markets, and banks are a less important source of ®nance. The remaining six countries are somewhere in between these two groups. Table 5.5 summarises the material in Tables 5.2±5.4 and suggests the overall relative strength of monetary policy in the fourteen EU countries, Japan and the United States. The ®nal column, `Predicted effectiveness of monetary policy,' reports a measure of the effects of monetary policy on output and in¯ation, where higher values suggest a stronger lending channel and therefore a larger impact. Overall, the pattern is very similar to the one reported in Kashyap and Stein (1997, Table 6). Most important, the predicted effects of interest rate movements vary greatly across countries. For example, looking at the EMU countries, one would expect that a given interest rate
180 Stephen G. Cecchetti Table 5.4 Importance of external and bank ®nance, by country, 1996
Country
Publicly Market Number traded capitalisation of publicly ®rms per as a percentage traded ®rms capita of GDP (1) (2) (3)
Corporate Bank loans as debt as a a percentage percentage of all forms of of GDP ®nance (4) (5)
Monetary union members Austria 106 Belgium 139 Finland 71 rance 686 Germany 681 Ireland 76 Italy 217 Netherlands 217 Portugal 158 Spain 357
13.15 13.68 13.87 11.75 8.32 21.59 3.78 13.97 16.11 9.09
15 45 50 38 29 18 21 96 23 42
46 60 34 49 58 13 37 48 19 11
65 49 39 49 55 80 50 53 62 58
Members of the EU not in EMU Denmark 237 Greece 245 Sweden 229 United Kingdom 2 433
45.06 23.44 25.90 41.39
41 20 99 150
105 3 73 45
25 48 32 37
Other countries Japan United States
18.56 31.94
67 111
39 64
59 21
2;34 8,479
Notes: Market capitalisation is the year-end value of ®rms listed on major exchanges. For the United States, three exchanges are used; for Japan, eight; and for each of the remaining countries, one. Sources: See the Data Appendix (p. 190).
change would have the most impact in Austria and Italy, countries in which small banks are relatively important, the banking systems are less healthy and ®rms have little access to non-bank sources of ®nance. The opposite is true of Belgium, Ireland and the Netherlands, where the banking systems are large and healthy and non-bank ®nance is readily available; in these countries, interest rate movements would be expected to have a more muted impact.15 The conclusions of this section could be criticised as applying only to the pre-EMU period. But will the introduction of the Euro be a catalyst for the harmonisation of ®nancial structure across the EMU? I take this question up in more detail later, but at this point I will simply mention that the European Central Bank (1999) report, Possible Effects of EMU on the EU Banking Systems in the Medium to Long Term, provides very little evidence to suggest that an increase in either international banking competition or securitisation and disintermediation will occur quickly.
Legal Structure, Financial Structure and Monetary Policy Transmission 181 Table 5.5 Summary of factors affecting the strength of the monetary transmissions mechanism
Country
Importance of small banks (1)
Monetary union members Austria Belgium Finland France Germany Ireland Italy Netherlands Portugal Spain
Availability Bank of alternative health ®nance (2) (3)
Predicted effectiveness of monetary policy (4)
3 1 1
2
3
1
2
1
1
2
2 2 3 3 2 1 3 2 3 2
3 1 2 2 2 3 3 2 3 2
2.67 1.33 2.00 2.33 2.33 1.67 2.67 1.67 2.33 2.00
Members of the EU not in EMU Denmark 2 Greece 2 Sweden 1 United Kingdom 1
2 2 3 1
1 3 1 1
1.67 2.33 1.67 1.00
Other countries Japan United States
4 1
2 1
2.67 1.67
2 3
Notes: Column (1) is based on Table 5.2; column (2), on Table 5.3; and column (3), on Table 5.4. Column (4) is an average of columns (1), (2) and (3).
4 Measuring the impact of policy on output and prices Testing the proposition that the banking system's concentration, health and importance have a material impact on the monetary transmission mechanism requires an estimate of the effects of an interest rate change on output and prices. Numerous studies report such estimates for some or all of the countries of the EU. These include Gerlach and Smets (1995), who estimate a threevariable structural vector autoregression based on long-run restrictions; de Bondt (1997), who presents estimates of the impact of policy on output and prices for Germany, France, Italy, the United Kingdom, Belgium and the Netherlands that are based on the work of other authors; Dornbusch, Favero and Giavazzi (1998), who report estimates of the impact of policy on output and prices derived from both small vector-autoregressive models and large structural models for Italy, Germany, France, Spain, Sweden and the United Kingdom; Kieler and Saarenheimo (1998), who study France, Germany and
182 Stephen G. Cecchetti
Per cent 1.0 Belgium 0.5
Per cent 1.0 France 0.5
0
0
–0.5
–0.5
–1.0
–1.0
–1.5
–1.5 2
1.0
4
6
8
10
12
14
16
18
20
22
24
26
2 1.0
Germany
0
0
–0.5
–0.5
2
4
6
8
10
12
14
16
18
20
22
24
2
26 1.0
Italy
12
14
16
18
20
22
24
26
4
6
8
10
12
14
16
18
20
22
24
26
6
8
10
12
14
16
18
20
22
24
26
6
8
10
12
14
16
18
20
22
24
26
10
12
14
16
18
20
22
24
26
Portugal
0.5
0.5 0
0
–0.5
–0.5
–1.0
–1.0 –1.5
–1.5 2
4
6
8
10
12
14
16
18
20
22
24
2
26 1.0
Spain
4
Denmark
0.5
0.5 0
0
–0.5
–0.5
–1.0
–1.0 –1.5
–1.5 2
4
6
8
10
12
14
16
18
20
22
24
2
26 1.0
Sweden
4
United Kingdom
0.5
0.5 0
0
–0.5
–0.5
–1.0
–1.0 –1.5
–1.5 2 1.0
10
–1.5
–1.5
1.0
8
–1.0
–1.0
1.0
6
0.5
0.5
1.0
4
Ireland
4
6
8
10
12
14
16
18
20
22
24
26
2
4
6
8
United States
0.5 0
Output response Inflation response
–0.5 –1.0 –1.5 2
4
6
8
10
12
14
16
18
20
22
24
26
Figure 5.1 Reaction of output and in¯ation to an interest rate increase of 100 basis points, quarterly, by country Sources: Cecchetti (1996); Cecchetti, McConnell and Perez Quiros (1999).
Legal Structure, Financial Structure and Monetary Policy Transmission 183
the United Kingdom, concluding that the transmission mechanism is not signi®cantly different across the three countries; Vlaar and Schuberth (1998), who examine money demand functions for fourteen EU countries; Ehrmann (1998), who estimates structural vector autoregressions for thirteen countries and ®nds considerable differences in the intensity of the response of output and prices to monetary shocks across countries; and Cecchetti and Rich (1999), who look at a simple two-variable system for Australia, Canada, France, Italy, Switzerland, the United Kingdom and the United States, and ®nd large differences in the implied impacts. Each of these studies has advantages and disadvantages. Overall, I have chosen to examine the results reported by Ehrmann (1998). The appeal of Ehrmann's approach is that it yields a series of estimates, all based on the same methodology, for nearly the full set of EU countries. Ehrmann uses techniques devised by King et al. (1991). In effect, Ehrmann identi®es monetary shocks using a combination of long-run and short-run restrictions. The methods are described both in his paper and in Cecchetti, McConnell and Perez Quiros (1999). For each country, the model has either four or ®ve variables, including output, in¯ation and an interest rate, and ± with the exception of Germany ± an exchange rate. When a ®fth variable is present, it is either a second interest rate or a commodity price index.16 The chart plots the responses of output and in¯ation to an interest rate movement of 100 basis points for ten EU countries and the United States (see Figure 5.1).17 These ten countries are the ones for which Ehrmann is able to generate consistent and plausible results.18 As is clear from these plots, the point estimates of the impulse response functions vary dramatically across countries. Looking at the impact of interest rate movements on output, note that for France and Germany, the peak impact is nearly twice what it is in the remaining European countries, and ®fteen times the estimated impact in the United States. The impact of policy on in¯ation also varies substantially. Table 5.6 reports the maximum impact of a 100-basis-point monetary contraction on output and in¯ation for all of the countries for which I have estimates. I also include a measure of the timing of the impact ± the quarter at which the maximum effect occurs. Column (5) presents a measure of the ratio of the average output response to the average in¯ation response. This measure is related to the sacri®ce ratio because it is roughly the output loss for an in¯ation decline of one percentage point over a horizon of approximately three years. Unfortunately, these estimates are not terribly precise, a point that is clear from the results in Ehrmann's paper,19 and so we should not take some of the numbers too seriously.
5 Systematic differences in national legal systems If differences in ®nancial systems are creating the cross-sectional variation in the transmission mechanism, it is natural to look for the causes of these
Output
Country
Maximum impact (1)
184
Table 5.6. Impact on output and in¯ation of a 100-basis-point increase in interest rates In¯ation
Quarter of maximum impact (2)
Maximum impact (3)
Quarter of maximum impact (4)
Approximate sacri®ce ratio (5)
Monetary union members Austria Belgium Finland France Germany Ireland Italy Netherlands Portugal Spain
Ð ±0.72 Ð ±1.30 ±1.21 ±0.76 ±0.64 Ð ±0.39 ±0.46
Ð 2 Ð 5 5 4 5 Ð 2 4
Ð ±0.05 Ð ±0.21 ±0.48 ±0.25 ±0.25 Ð ±0.28 ±0.23
Ð 1 Ð 2 2 5 9 Ð 1 4
Ð ±45.29 Ð ±12.07 ±5.83 ±3.45 ±5.01 Ð ±0.58 ±1.34
Members of the EU cnot in EMU Denmark Greece Sweden United Kingdom
±0.48 Ð ±0.56 ±0.53
5 Ð 4 13
±0.34 Ð ±0.11 ±0.37
1 Ð 5 1
±1.69 Ð ±5.61 ±2.57
Other countries Japan United States
Ð ±0.07
Ð 6
Ð ±0.017
Ð 12
Ð ±3.27
Sources: Estimates for the United States are from Cecchetti (1996); those for the remaining countries are from the estimation of Ehrmann's model in Cecchetti, McConnell and Perez Quiros (1999).
Legal Structure, Financial Structure and Monetary Policy Transmission 185
differences. As noted earlier, La Porta et al. (1997, 1998) have examined the relationship between a country's legal system and its ®nancial system. The premise of their work is that investors provide capital to ®rms only if the investors have the ability to get their money back. For equity holders, this means that they must be able to vote out directors and managers who do not pay them. For creditors, this means that they must have the authority to repossess collateral. In addition to having nominal legal rights, these groups must also have con®dence that the laws will be enforced. La Porta et al. (1997, 1998) collect data on the legal systems in forty-nine countries. They argue that all of these legal systems belong to one of four families: English common law, French civil law, Scandinavian civil law and German civil law. With regard to shareholder rights ± speci®cally, the ability of shareholders to vote directors out ± English common law countries have the best protections and French civil law countries have the worst. The pattern is similar for creditor rights, which entail the right to reorganise or liquidate a ®rm. The pattern for enforcement is a bit different: Scandinavian civil law countries have the most rigorous law enforcement, while French civil law countries have the most lax.
Table 5.7. Shareholder rights, creditor rights and enforcement, byCountry Country
Shareholder rights Creditor rights (1) (2)
Enforcement Legal family (3) (4)
Monetary union members Austria 2 Belgium 0 Finland 2 France 2 Germany 1 Ireland 3 Italy 0 Netherlands 2 Portugal 2 Spain 2
3 2 1 0 3 1 2 2 1 2
10.00 10.00 10.00 8.98 9.23 7.80 8.33 10.00 8.68 7.80
German French Scandinavian French German English French French French French
Members of the EU not in EMU Denmark 3 Greece 1 Sweden 2 United Kingdom 4
3 1 2 4
10.00 6.18 10.00 8.57
Scandinavian French Scandinavian English
Other countries Japan United States
2 1
8.98 10.00
German English
3 S
Source: La Porta et al. (1997), Table II.
186 Stephen G. Cecchetti
Table 5.7 reproduces a portion of Table II from La Porta et al. (1997). The column labelled `Shareholder rights' reports an index that is higher when shareholders ®nd it less costly and dif®cult to vote directors out. The column labelled `Creditor rights' reports an analogous index that is lower when creditors experience less dif®culty gaining possession of property that has been used to collateralise a bond or loan. Enforcement is an assessment of countries' rigour in carrying out their laws, with a higher score implying more aggressive enforcement. Finally, the table reports the legal family from which each country's laws are derived. Using these data to examine the relationship between shareholder rights creditor rights, and enforcement on the one hand, and the concentration of ownership and the availability of external ®nance on the other, La Porta et al. (1997, 1998) come to two conclusions. First, corporate ownership is more concentrated in countries where shareholders and creditors are poorly protected by both the substance of the law and its enforcement. Second, and more germane to the current discussion, countries with weaker legal rules and less rigorous law enforcement have smaller and narrower capital markets. Overall, English common law countries have the least concentration of corporate ownership and the largest and deepest capital markets. French civil law countries have the most concentrated ownership and the smallest capital markets. La Porta et al. (1997 p. 1131) conclude that the `differences in the nature and effectiveness of the ®nancial systems around the world can be traced, in part, to differences in investor protection against expropriation by insiders, as re¯ected by legal rules and the quality of their enforcement'. Their ®ndings are con®rmed by the data in Table 5.4, which show clearly that the United States and the United Kingdom have much more extensive capital markets than France and Italy.
6
Relationship of the legal environment to the impact of policy
Following the demonstration in La Porta et al. (1997, 1998) that the systematic variation in systems of corporate governance and ®nance across countries can be tied to the differences in the countries' legal systems, I ask if the variation in the predicted strength of the lending channel and the estimated impact of interest rate movements on output and in¯ation can be traced to these same legal differences.20 To address this question, I combine the data from Table 5.5 on the predicted strength of the lending channel of monetary transmission and from Table 5.6 on the size of the impact of interest rate movements on output and in¯ation with the measures of cross-country differences in legal organization from Table 5.7. In Table 5.8, I report the results of two straightforward exercises. The ®rst separates the countries by the origin of their legal system and constructs group averages for the effectiveness and impact of monetary policy from column (4) of Table 5.5 and columns (1) and (3) of Table 5.6 (Table 5.8, top panel). The results follow the pattern predicted by the index
Legal Structure, Financial Structure and Monetary Policy Transmission 187 Table 5.8 Testing the relationship between cross-country differences in legal structure and monetary policy effectiveness
Legal family
Predicted effectiveness of monetary policy
Impact of policy On output
On in¯ation
Approximate sacri®ce ratio
±0.21 ±0.22 ±0.20 ±0.49
±3.1 ±3.7 4.8a ±5.8
Group mean English Scandinavian French German
1.1 1.8 2.1 2.4
±0.45 ±0.52 ±0.70 ±1.25
Instrumental variables regression Coef®cient Standard error
Ð Ð
±0.46 (0.22)
0.05 (0.08)
±10.4 (10.4)
Notes: `Predicted effectiveness' is drawn from column (4) of Table 5.5; the `Impact of policy', from columns (1) and (3) of Table 5.6. The instrumental variables regression is of columns (1) and (3) of Table 5.6 on column (4) of Table 5.5, with columns (1), (2) and (3) of Table 5.7 as instruments. All of the results in this table use only the eleven countries for which there are estimates in Table 5.6: Ireland, the United Kingdom and the United States (English common law); Denmark and Sweden (Scandinavian common law); Belgium, France, Italy, Portugal and Spain (French civil law); and Germany (German civil law). a average excludes Belgium.
of lending channel effectiveness as the impact of policy on output and the approximate sacri®ce ratio vary systematically ± and as expected ± with the origin of a country's legal system. We can learn a bit more from the data than is recovered from the simple averages reported in the top panel of Table 5.8. The question of greatest interest is whether the cross-country heterogeneity in the real effects of monetary policy can be explained by differences in the countries' ®nancial systems, which have their source in the strength of shareholder and creditor rights and the rigour with which these rights are enforced. We can do this without fully accounting for all of the variation in the transmission mechanism if we assume that the La Porta et al. (1997) measures are valid instruments for the ®nancial variables in a simple regression that has the impact of policy on the left-hand side and the overall measure of the lending channel's effectiveness on the right-hand side. That is, I assume that the shareholder, creditor and enforcement variables are exogenous, while the measure of the effectiveness of the lending channel may not be. The results of these two-stage least squares regressions are reported in the bottom panel of Table 5.8. Again, we see that the countries in which the lending channel is expected to be strongest have the biggest sacri®ce ratios and show the largest impact of interest rate movements on output. The latter of these relationships has a t-ratio of 2.1, and so it may even be signi®cantly
188 Stephen G. Cecchetti
different from zero. The results for in¯ation are much less satisfactory: the measures of ®nancial structure appear to be uncorrelated with the impact of policy on prices. Because of the small size of the sample (eleven countries), the estimates are all fairly imprecise, and so I treat them as being only suggestive.
7
Concluding remarks
Among the many challenges facing the new Eurosystem is the possibility that the regions of the Euro Area will respond differently to interest rate changes. In this chapter, I have suggested that differences in ®nancial structure are a proximate cause for these national asymmetries in the monetary policy transmission mechanism. Moreover, I have proposed that these differences in ®nancial structure are likely a result of the EU countries' diverse legal structures. The evidence, although circumstantial, is consistent with this view. Most economists believe that the monetary transmission mechanism will vary systematically across countries with differences in the size, concentration and health of the banking system, and with differences in the availability of primary capital market ®nancing. The countries of the EU differ quite dramatically in all of the dimensions that would seem to matter, leading to the prediction that the impact of interest rates on output and prices will not be consistent across countries. While the estimates of the impact of interest rate changes on output and in¯ation tend to be quite imprecise, they do differ, and in the way that is predicted by the state of the countries' ®nancial systems. Finally, we can trace differences in ®nancial structure, the size and scope of capital markets, and the availability of alternatives to bank ®nancing to differences in the countries' legal structures. What does this mean for the future of ®nancial markets and monetary policy in the Euro Area? Will the European banking system become more like that of the United States? The arguments presented here suggest that unless legal structures are harmonised across Europe, ®nancial structures will remain diverse, and so will monetary transmission mechanisms. It will not be enough to make regulatory structures more similar, since such a change will not, in and of itself, alter the structure of capital markets. In other words, I do not view regulatory competition as a force to eliminate the asymmetries in the ®nancial intermediation systems of the EU.21 As the ECB (1999) report makes clear, this force has been very weak in the past and is expected to be weak in the future. While we may see cross-border mergers and acquisitions of ®nancial sector ®rms that take advantage of the expertise of those already doing business in a region,22 only a decision to change the existing legal structures so that shareholders and creditors in all EU countries enjoy the same rights will force the movement to a US- style ®nancial structure.
Legal Structure, Financial Structure and Monetary Policy Transmission 189
Data Appendix The data sources for Tables 5.2±5.4 in this chapter are as follows:
Table 5.2 Number of institutions and concentration ratios: For all countries, concentration is calculated as the assets of the top ®ve banks as a percentage of total bank assets. Population: International Monetary Fund, International Financial Statistics (January 1999), country report tables, l. 99z, mid-year estimates for all countries. Austria: Austrian National Bank web pages < http://www.oenb.co.at/stat-monatsheft/ tabellen/2001p.htm >, Ingesamt, Hauptanstalten, for number of institutions; and < http://www.oenb.co.at/stat-monatsheft/tabellen 2000_5p.htm >, Alle Sektoren, Summe Aktiva (Ohne Rediskonte), for total assets; Austrian National Bank, Economic Analysis Division, for assets of top ®ve and top ten banks. Belgium: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on bank balance sheets, p. 36, l. 37 (under supplementary information), for number of institutions; Bank of Belgium, Financial and Economic Statistics Division, for total assets of credit institutions and for share of top ®ve banks. Finland: Bank of Finland, Financial Statistics Desk, for all ®gures. France: Bank of France, Monetary Research and Statistics Division (DESM-SASM) for all ®gures on credit institutions. Germany: Deutsche Bundesbank, Monthly Report (May 1998), p. 16, Table IV.1, column 1, for number of institutions; Deutsche Bundesbank, Department of Controlling, Accounting and Organisation, Division C-2, for share of top ®ve banks. Ireland: Central Bank of Ireland, Monetary Policy and Statistics, for number and total assets of all credit institutions (which include licensed banks, building societies, statesponsored ®nancial institutions and savings banks); IBCA BankScope database, for assets of top ®ve banks. Italy: Bank of Italy, Research Department, for all ®gures. Netherlands: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on bank balance sheets, p. 192, l. 37 (under supplementary information), for number of institutions; De Nederlandsche Bank, Annual Report (1997), Tables 1, 2.1 and 2.2, for assets of top ®ve banks and for total assets of monetary institutions. Portugal: Bank of Portugal web page < http://www.bportugal.pt/publish/frpublish_e.htm >, Chart VIII.1 and Table VIII.2, for number of institutions and share of top ®ve banks. OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on bank balance sheets, p. 231, l. 25, for total assets of commercial banks. Spain: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on bank balance sheets, p. 236, l. 37 (under supplementary information), for number of banks; Bank of Spain, Statistical Bulletin (June 1998), Tables 61.1 (p. 271), 62.1 (p. 281), 63.1 (p. 291), sum of column 1 in all tables, for total assets of banks, savings banks, and credit cooperatives; IBCA Bankscope database, for assets of top ®ve banks. Denmark: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on bank balance sheets, p. 64, l. 37 (under supplementary information), for number of institutions; Denmark National Bank web page < http://www.nationalbanken.dk/nb/nb.nsf/alldocs/ F15D9E8CF275ED1A2412565B4003E8BD5, > for total assets; IBCA BankScope database, for assets of top ®ve banks. Greece: Hellenic Bank Association, The Greek Banking System (April 1998), p. 87, for number of institutions, total assets, and assets of top ®ve banks.
190 Stephen G. Cecchetti Sweden: Sveriges Riksbank, Statistical Yearbook (1996), p. 17, Table 6, for number of banks; Sveriges Riksbank, Financial Statistics Department, for share of top ®ve banks. United Kingdom: British Bankers Association, Annual Abstract of Banking Statistics (1997), Table 1.04, for number of institutions; Bank of England, MFSD, for shares of top ®ve banks (data relate to all banks and building societies operating in the United Kingdom and so include the business of foreign-owned af®liates in the United Kingdom). Japan: Bank of Japan, International Department, for all ®gures for banks and other deposit-taking institutions, end of ®scal year 1996 (March 1997). United States: Federal Financial Institutions Examination Council (FFIEC), Reports of Condition (call reports database), for all ®gures for commercial banks.
Table 5.3 Bank data: McCauley and White (1997), Table 1. Federal Reserve Bank of New York staff calculations for Austria, Belgium, Greece, Ireland, Italy and Portugal, based on ranking of asset size from IBCA BankScope database. In each country, banks were chosen according to 1997 assets. Return on assets, loan loss provisions, net interest margin and operating cost are drawn from IBCA BankScope database. Thomson ratings: Thomson BankWatch database.
Table 5.4 Number of publicly traded ®rms and market capitalisation: International Finance Corporation, Emerging Stock Markets Factbook (1997), pp. 17 and 23 (also available on the Wall Street Journal web site < http://update.wsj.com/public/resources/documents/gitab5.htm >). Population: See sources for population data in Table 5.2. Privately issued debt: Bank for International Settlements, International Banking and Financial Market Developments (February 1998), pp. 46±7, Tables 14 and 15, amount outstanding, December 1996 ®gures; sum of ®gures from Table 14 (international debt securities) and Table 15 (domestic debt securities). GDP: International Monetary Fund, International Financial Statistics (January 1999), country report tables, l. 99b.c for all countries. Year-average exchange rates used for conversion into US dollars (local currency per US dollar, l. rf for all countries). Bank loans: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on bank balance sheets, l. 16 on pp. 27, 35, 63, 67, 91, 115, 143, 159, 163, 167, 191, 231, 235, 251, 259, 263, 303, 307 and 315.
Notes 1. I gratefully acknowledge comments from and discussions with Ignazio Angeloni, Paul Bennett, Stefan Gerlach, Heinz Hermann, Beverly Hirtle, James Kahn, Anil Kashyap, Kenneth Kuttner, Gabriel Perez Quiros, Margaret Mary McConnell, Robert Rich and especially Paolo Pesenti. I thank Michael Ehrmann for providing his programs, and Rama Seth for helping with the data. 2. Throughout the chapter, I refer to the eleven countries of EMU but provide information on only ten. Luxembourg is not included. 3. See Angeloni and Dedola (1998). 4. Similar points are made by White (1998), who suggests that competition in banking may be about to increase in Europe, stimulated by the introduction of the
Legal Structure, Financial Structure and Monetary Policy Transmission 191
5.
6.
7.
8.
9.
Euro. In addition, a ECB (1999) study suggests that EMU may speed up the process of disintermediation and lead to a more geographically diversi®ed and internationalised banking system. For example, within the United States, more than 10 per cent of ®rms with assets exceeding $1 million have chosen to incorporate in Delaware, a state with less than half of one per cent of the country's population. Why is this? The answer can be found by considering how the development of Delaware's legal structure has differed from the development of the legal structure in other states. Originally, large ®rms were incorporated in New Jersey because the state, in exchange for incorporation fees and franchise taxes, had liberalised its corporation law to allow various mergers and cross-holdings that were disallowed elsewhere. State law also gave very strong power to corporations' directors (Grandy, 1989). Delaware copied New Jersey's statutes and then bene®ted from changes made to New Jersey's law by Governor Woodrow Wilson in 1913. As this example suggests, the economic structure has its source in the legal structure. I should note that ®rms in countries that act slowly will be put at a competitive disadvantage, and so they might pressure their governments to speed up the legal changes. The potential strength of such regulatory competition is an open issue. There is an alternative. A company may move to a country where the ®nancial system better suits its needs. The La Porta et al. measures, reported in Table 5.7, suggest that the United Kingdom is the best country in the European Community in which to issue both bonds and stocks, and so ®rms that wish to have ready access to primary capital market ®nancing may tend to concentrate there. But for this strategy to be successful, ®rms would have to reincorporate and move assets into the alternative jurisdiction. The assets must move to provide the proper guarantees to investors. All of this seems unlikely. In addition to the differences in the type of nominal rigidity, there are variations in the way in which the rigidites are modelled. These variations are more than formal; they have very different implications for the dynamic effects of nominal shocks on real variables. Different modelling strategies are based on differences in the timing of price or wage change decisions. There are three basic schemes used, based on Fischer (1977), Taylor (1980) and Calvo (1983), and they create very different dynamic responses of real variables to nominal shocks. Fischer, for example, assumes prices are predetermined, meaning that at some time agents set prices for some number of future periods: the level of prices set on the decision date can differ for the different periods before the next decision date. In this model, the impact of a nominal shock lasts for only as long as it takes for all price setters to have a chance to reset their price schedules. In the Taylor model, prices or wages are assumed to be ®xed, meaning that their nominal value does not vary between decision dates. When prices or wages are ®xed, nominal shocks die out only asymptotically. In Calvo's (1983) model, price-setters change their prices according to a Poisson process, leading to a variety of possible dynamics. Longer-run considerations, such as the potential costs or bene®ts of modest levels of in¯ation, critically depend on understanding the sources of nominal rigidity. For example, Akerlof, Dickens and Perry (1996) and Groshen and Schweitzer (1997) consider whether small positive levels of aggregate in¯ation can facilitate real adjustments in the presence of an aversion to nominal wage declines, suggesting that the long-run goal for in¯ation might be positive. But Feldstein (1996) contends that the tax distortions created by in¯ation reduce the level of output
192 Stephen G. Cecchetti
10. 11.
12. 13. 14. 15.
16. 17. 18.
19. 20
21.
22.
permanently, an argument that suggests that the optimal level of in¯ation may even be negative. Overall, most economists now seem to agree that in¯ation leads to lower levels of real output and may even retard long-run growth. See Feldstein (1999) for a summary. As emphasised by Kashyap and Stein (1994), this assertion applies to both ®nancial and non-®nancial ®rms. This is not to say that the traditional mechanisms, operating through interest rates and exchange rates, are not present as well. Unfortunately, it has proved to be very dif®cult to disentangle the individual importance of the various channels of transmission. It is important to note that there can be signi®cant cyclical and secular changes in the strength of the lending channel as the health and concentration of the banking system change, and as capital markets become deeper and broader. After I collected the data for this section, the ECB issued its report Possible Effects of EMU on the EU Banking Systems in the Medium to Long Term. The Appendix tables in that report contain much of the same information presented here. Throughout the analysis, I omit Luxembourg. A signi®cant failing of this analysis is the assumption that these relative rankings are not changing over time. Surely, if I had chosen different dates to measure the relative health and concentration of countries' banking systems, I would have created a different set of rankings for the ®rst two indicators. It is entirely possible that both the relative importance of small banks and the health of the banking system will become increasingly uniform across countries, leaving only differences in external ®nance. See Appendix A in Ehrmann (1998) for additional details. The results for the United States are derived from Cecchetti (1996). Although he reports estimates for thirteen countries, the estimates for three of these countries appear to be dif®cult to interpret. In the case of Finland, for example, the impact of monetary tightening is to increase output, not decrease it. For Austria and the Netherlands, we have not been able to replicate the results in the current version of Ehrmann's paper. Figures 1±13 in Ehrmann (1998) show that the impulse response functions are rarely signi®cantly different from zero. The same point is made in Cecchetti (1998) and Cecchetti and Rich (1999). White (1998) makes a related point when he notes that the legal, tax, regulatory and supervisory frameworks within which ®nancial institutions operate differ signi®cantly across the various countries of the EU. All of these differences make direct competition more complex and less appealing. He goes on to focus on differences in the EU countries' labour laws and in the regulatory restrictions the countries place on the types of ®nancial products that can be offered. These effects are surely complementary to the ones I address here. It is also extremely unlikely that these dif®culties will be overcome by the issuance of debt and equity in a jurisdiction that offers suf®cient investor protections. But unless ®rms have assets within these jurisdictions, I do not see this as a solution. Such developments would be similar to what has happened with the relaxation of interstate branching regulations in the United States, where banks in one state have purchased a bank in another state in order to obtain the legal and regulatory knowledge to do business in that state. Interstate branching has not meant opening new branches of an existing bank in another region.
Legal Structure, Financial Structure and Monetary Policy Transmission
193
References Akerlof, G. A., W. T. Dickens and G. L. Perry, (1996) `The Macroeconomics of Low In¯ation', Brookings Papers on Economic Activity, 1, 1±59. Angeloni, I., and L. Dedola, (1998) `From the ERM to the Euro: A Soft Transition?', Banca d'Italia, Working paper, June. Bernanke, B. S. (1983) `Nonmonetary Effects of the Financial Crisis in the Propagation of the Great Depression', American Economic Review, 73, 257±76. ÐÐÐÐ (1993) `Credit in the Macroeconomy', Federal Reserve Bank of New York Quarterly Review, Spring, 50±70. Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of Monetary Transmission', American Economic Review, 82, 901±21. Bernanke, B. S. and M. Gertler (1989) `Agency Costs, Net Worth and Business Fluctuations', American Economic Review, 79, 14±31. ÐÐÐÐ (1990) `Financial Fragility and Economic Performance', Quarterly Journal of Economics, 105, 87±114. Calvo, G. A. (1983) `Staggered Price Setting in a Utility-Maximizing Framework', Journal of Monetary Economics, 12, 383±98. Cecchetti, S. G. (1995) `Distinguishing Theories of the Monetary Transmission Mechanism', Federal Reserve Bank of St. Louis Economic Review, 77, 83±97. ÐÐÐÐ (1996) `Practical Issues in Monetary Policy Targeting', Federal Reserve Bank of Cleveland Economic Review, 32, 2±15. ÐÐÐÐ (1998) `Policy Rules and Targets: Framing the Central Banker's Problem', Federal Reserve Bank of New York Economic Policy Review, 4, 1±14. Cecchetti, S. G., M. M. McConnell and G. Perez Quiros (1999) `Policy-makers' Revealed Preferences and the Output±In¯ation Variability Trade-off', Federal Reserve Bank of New York, February, unpublished. Cecchetti, S. G. and R. W. Rich (1999) `International Evidence on the Sacri®ce Ratio: Do We Know How Big It Is?', Federal Reserve Bank of New York, February, unpublished. Christiano, L. J., M. Eichenbaum and C. L. Evans (1997) `Sticky Price and Limited Participation Models of Money: A Comparison', European Economic Review, 41, 1201±49. Danthine, J.-P., F. Giavazzi, X. Vives and E.-L. von Thadden (1999) The Future of European Banking, London, Centre for Economic Policy Research. de Bondt, G. J. (1997) `Monetary Transmission in Six EU-Countries: An Introduction and Overview', De Nederlandsche Bank, Research Memorandum 527. Dornbusch, R., C. Favero and F. Giavazzi (1998) `Immediate Challenges for the European Central Bank', Economic Policy, 28, 17±64. Ehrmann, M. (1998) `Will EMU Generate Asymmetry? Comparing Monetary Policy Transmission Across European Countries', European University Institute Working Paper 98/28, October. European Central Bank (1999) Possible Effects of EMU on the EU Banking Systems in the Medium to Long Term, Frankfurt, European Central Bank, February. Feldstein, M. (1996) `The Costs and Bene®ts of Going from Low In¯ation to Price Stability', In C. Romer and D. Romer (eds)., Reducing In¯ation, Chicago, University of Chicago Press, 123±66. Feldstein, M. (ed.) (1999) The Costs and Bene®ts of Achieving Price Stability, Chicago, University of Chicago Press for NBER. Fischer, S. (1977) `Long-term Contracts, Rational Expectations, and the Optimal Money Supply Rule', Journal of Political Economy, 85, 191±205.
194 Stephen G. Cecchetti Gerlach, S. and F. Smets (1995) `The Monetary Transmission Mechanism: Evidence from the G-7 Countries', Centre for Economic Policy Research Working Paper 1219, July. Gertler, M. and S. Gilchrist (1993) `The Role of Credit Market Imperfections in the Monetary Transmission Mechanism: Arguments and Evidence', Scandinavian Journal of Economics, 95, 43±64. Grandy, C. (1989) `New Jersey Corporate Chartermongering, 1875±1929', Journal of Economic History, 49, 677±92. Groshen, E. and M. Schweitzer (1997) `Identifying In¯ation's Grease and Sand Effects in the Labor Market', NBER Working Paper 6061, June. Hubbard, R. G. (1995) `Is There a `Credit Channel' of Monetary Policy?', Federal Reserve Bank of St Louis Economic Review, 77, May-June, 63±77. Kashyap, A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending', In N. G. Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press for NBER. ÐÐÐÐ (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for the European Monetary Union', Federal Reserve Bank of Chicago Economic Perspectives, September±October, 2±18. Kieler, M. and T. Saarenheimo (1998) `Differences in Monetary Policy Transmission? A Case Not Closed', European Commission Economic Papers, 132, November. King, R. G., C. I. Plosser, J. H. Stock and M. W. Watson (1991) `Stochastic Trends and Economic Fluctuations', American Economic Review, 81, 819±40. La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997) `Legal Determinants of External Finance', Journal of Finance, 52, 1131±50. ÐÐÐÐ (1998) `Law and Finance', Journal of Political Economy, 106, 1113±55. McCauley, R. N. and W. R. White (1997) `The Euro and European Financial Markets', Bank for International Settlements Working Paper 41, May. Rotemberg, J. J. (1984) `A Monetary Equilibrium Model with Transactions Costs', Journal of Political Economy, 92, 40±58. Taylor, J. B. (1980) `Aggregate Dynamics and Staggered Contracts', Journal of Political Economy, 88, 1±23. Vlaar, P. J. G. and H. Schuberth (1998) `Monetary Transmission and Controllability of Money in Europe: A Structural Vector Error Correction Approach', De Nederlandsche Bank, May, unpublished. White, W. R. (1998) `The Coming Transformation of Continental Banking?', Bank for International Settlements Working Paper 54, June.
Discussion
Manfred J. M. Neumann
Cecchetti provides us in Chapter 5 with an empirical study of the monetary transmission mechanism in the European countries. Studying country-speci®c differences of the transmission process is potentially of great relevance in view of the fact that European Monetary Union (EMU) is in operation. Now the eleven countries that have entered EMU are confronted with the single monetary policy run by the independent European Central Bank (ECB). Given the diversity of economic and ®nancial structure it is likely that the ECB's monetary policy actions will affect the member economies at different strengths and speeds. The focus of Cecchetti's study is on assessing the empirical relevance of the bank credit channel of policy transmission which forms an important part of the transmission mechanism of relative prices. The theory of relative prices states that any policy shock destroys portfolio equilibrium and affects all ®nancial and real assets and, consequently, the net worth of ®nancial intermediaries, enterprises and households. Owing to asset-speci®c information and transaction costs as well as rigidities, the shock transmission is not uniform but differs among assets as regards price and quantity response as well as timing. The credit view concentrates on the differential impact of policy shocks on the supply of loans by banks versus the supply of credit by other ®nancial intermediaries or markets. Given that banking regulation makes banks dependent on the supply of reserve money, the prediction is that monetary policy affects banks, hence their clients, quicker and more forcefully than credit markets outside the banking system. Cecchetti's study essentially consists of three parts. Following the work by Kashyap and Stein (1997), he collects data on selected characteristics of the ®nancial structure of fourteen EU countries as well as the United States and Japan. For each country the information is condensed in two steps: ®rst, three partial indicators are computed that are supposed to re¯ect the relative importance of small banks, the health of the banking system and the availability of alternative ®nance; second, a summary indicator of the strength of the bank credit channel is computed, de®ned as the arithmetic mean of the 195
196 Manfred J. M. Neumann
three partial indicators. Cecchetti then addresses the question whether the evidence on country differences in credit channel strength can be explained by differences in the countries` legal systems as regards civil law. Finally ± and this an innovative aspect of the study ± Cecchetti examines whether the computed summary index of credit channel strength or importance is correlated with separate evidence, derived from estimates of structural VARs on the responses of output and in¯ation to monetary policy shocks. He ®nds that the expected correlation exists. While Cecchetti's research agenda is a potentially very fruitful one, the present study is a ®rst, very tentative exploration. My reading of the evidence presented is that it does not lend support to the prediction that the impact of monetary policy is signi®cantly shaped by differences in countries' ®nancial structures.
1
Data evaluation
Cecchetti presents data that describe ®nancial sector characteristics of sixteen countries and aggregates them to derive three indicators that are intended to re¯ect the relative strength of the credit channel. A fundamental problem to be noted is that the indicators are not strictly based on some aggregation formula but are affected by subjective evaluation. Cecchetti's indicator of the importance of small banks can be used to make the point. In the following Table D5.1 I show the two characteristics Cecchetti apparently relies on: (1) the number of banks per million people, and (2) a concentration ratio that aggregates the shares of the top ®ve banks in total bank assets. Compare, for example, Greece with the United Kingdom. Greece is assigned an indicator value of 2 while the United Kingdom receives a 1, meaning that small banks are less important in the United Kingdom than in Greece. But in fact, Greece has less banks per million of people than the United Kingdom (two to be compared to eight) and the Greek concentration ratio is three times the British ratio. Moreover, as the last column of Table D5.1 indicates, the average asset share of a Greek bank (excluding the top ®ve) is twelve times the share a British bank enjoys. On all accounts, it seems that the United Kingdom should be assigned an indicator value of 2 instead of 1 and Greece a value of 1 instead of 2. As a second example, compare Austria with the United Kingdom. Cecchetti's indicator signals that in Austria small banks are more important than in the United Kingdom. Indeed, on account of the number of banks per million of people this judgement is justi®ed. Looking at the concentration ratio, in contrast, the credit channel is less strong in Austria than in the United Kingdom. The top ®ve Austrian banks control 48 per cent of the banking system's assets while in the United Kingdom the top ®ve control only 28 per cent. Thus, we are unable to predict from these data whether a monetary policy shock will have a larger impact in Austria or in the United Kingdom. Given this cognitive problem, Cecchetti resorts to subjective weighting as
Discussion 197 Table D5.1 On the importance of small banks Banks per million people
Concentration ratio: top ®ve banks
Indicator of the importance of small banks
For comparison: average asset share, excluding top ®ve
Germany USA Austria
43 40 126
17 17 48
3 3 3
0.02 0.01 0.05
Denmark Italy Japan France Spain Greece
22 16 4 24 8 2
17 25 30 40 44 71
2 2 2 2 2 2
0.74 0.08 0.13 0.04 0.18 1.93
United Kingdom Belgium Ireland Finland Netherlands Portugal Sweden
8 14 18 68 11 5 14
28 57 41 78 79 76 90
1 1 1 1 1 1 1
0.15 0.32 1.04 0.06 0.13 0.52 0.08
Sources: Cecchetti (this volume), Tables 5.2 and 5.5; own computations.
Kashyap and Stein (1997) have done. But this practice throws considerable doubt on using this as well as similar indicators as inputs in subsequent empirical work. The more types of characteristics are considered, the more relevant becomes the danger of biased aggregation. Table D5.2 shows the two characteristics used by Cecchetti to derive an indicator of the availability of alternative ®nance. Countries with the least (most) developed external capital markets receive an indicator value of 3 (1). Compare, for example, Finland with Belgium. To be sure, on both characteristics the credit channel should be stronger in Belgium than in Finland, implying that the rank order should be reversed. Alternatively, compare Greece with Germany. The ratio of market capitalisation is smaller in Greece but so is the share of bank loans in total ®nance. Thus, it is not clear on what grounds Cecchetti assumes that the credit channel is stronger in Greece than in Germany. Given these inconsistencies in the data transformation, the information content of Cecchetti's indicators is in doubt. I realise that the researchers of the credit channel are aware of the crude nature of this work, and I admit that I do not know how to overcome the methodological problem of aggregating
198 Manfred J. M. Neumann Table D5.2 Availability of alternative ®nance Market capitalization as a percentage of GDP
Bank loans as a percentage of all forms of ®nance
Indicator of availability of alternative ®nance
Austria Ireland Greece Italy Portugal
15 18 20 21 23
65 80 48 50 62
3 3 3 3 3
Germany France Spain Finland Japan Netherlands
29 38 42 50 67 96
55 49 58 39 59 53
2 2 2 2 2 2
41 45 99 150 111
25 49 32 37 21
1 1 1 1 1
Denmark Belgium Sweden United Kingdom United States
Source: Cecchetti (this volume), Tables 5.4 and 5.5.
con¯icting information. Nevertheless it seems that the subjective weighting procedure should be abandoned.
2
Is the evidence in favour of the credit channel?
A novel feature of Cecchetti's study is that he computes a summary indicator of credit channel strength and confronts the predictions embodied in this indicator with estimates of the impact of monetary policy on output and in¯ation. The estimates are taken from a study by Ehrmann (1998) that employs structural VARs. This information, too, is aggregated for different groups of countries. The guiding notion for the grouping is the nature of the countries' legal system as regards civil law. La Porta et al. (1997) pointed out that the legal protection of creditors and shareholders is similar across countries that share the same legal tradition but very different across legal traditions. By comparing group averages Cecchetti shows in his Table 5.1 (p. 173) as well as the upper panel of his Table 5.8 (p. 187) that there apparently is a positive correlation between the estimated (maximum) impact of monetary
Discussion 199
policy on output (though not on in¯ation) and the prediction of relative credit channel strength derived from his summary indicator of ®nancial structures. As an additional underpinning Cecchetti also provides a supporting two-stage least squares regression that employs legal information as instrumental variables. If valid, this is a nice result. Unfortunately, there is reason to conjecture that the result is an artefact. Cecchetti himself notes that the VAR estimates are not very precise, meaning that the con®dence bands for most countries are very large (see Ehrman, 1998). Moreover, the estimates suggest for six out of eleven countries that the maximum impact of monetary policy on in¯ation precedes the impact on output, in most cases by several quarters. To be sure, this timing is dif®cult to rationalise and hence throws doubt on the reliability of the estimates. But let us assume that the VAR estimates for output re¯ect the true impact of monetary policy. On this assumption Cecchetti's regression appears to support the conjecture that monetary policy impacts on output are stronger the stronger is the credit channel. To check for this, Figure D5.1 plots the output responses against the summary indicator of the credit channel. The ®gure reveals that France and Germany, and possibly the United States, are
0.0 USA
Output response
DK –0.5
P
E
UK
S I B
IRE
–1.0 G F –1.5 0.5
1.0
1.5
2.0
2.5
3.0
Summary indicator of credit channel strength Figure D5.1 Maximum impact of an interest rate shock and strength of the credit channel
200 Manfred J. M. Neumann
outliers. For the remaining eight countries, who belong to different legal traditions, we observe about the same output response to monetary policy. In sum, we have to conclude that the evidence collected by Cecchetti does not support the hypothesis that the effectiveness of monetary policy is affected by a country's legal tradition or ®nancial structure. However, I hasten to add that this is not a ®nal verdict on the role of the credit channel as the negative result might be shaped by the shaky nature of the data employed. References Cecchetti, S. G. (2000) `Legal Structure, Financial Structure, and the Monetary Policy Transmission Mechanism', Chapter 5 in this volume. Ehrmann, M. (1998) `Will EMU Generate Asymmetry? Comparing Monetary Policy Transmission Across European Countries', European University Institute Working Paper 98/28, October. Kashyap, A. K. and J. C. Stein (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for the European Monetary Union, Federal Reserve Bank of Chicago Economic Perspectives, September±October, 2±18. La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997) `Legal Determinants of External Finance', Journal of Finance, 52, 1131±50.
Discussion1
Ignazio Angeloni
This is a very interesting and stimulating chapter, one that offers a novel perspective on the transmission process of monetary policy. As in many of his previous contributions, Cecchetti is able to identify a relevant policy issue, bring together stimulating evidence and draw intuitively appealing interpretations from it. If the prime test of the success of a study is to stimulate thinking, this is unquestionably a successful study. No matter how much we may disagree with the interpretations it proposes and with the overall policy message ± and, as I will argue below, there is ample room for disagreement ± I believe that future research on monetary transmision issues, related to the Euro Area in particular, will have to take the ideas that Cecchetti presents here carefully into account. In my comments I will ®rst explain what, in my view, is the main contribution of this chapter and outline the essence of the argument; I will then focus on what I regard as limitations of the analysis and explain why and how these limitations affect the overall policy conclusions. The chapter builds an ambitious bridge between two so far separate strands of literature ± that on the bank credit channel of monetary policy transmission and that on the legal determinants of corporate ®nance and of ®nancial structure in general. Popularised in the 1990s by contributions by Kashyap and Stein (1994, 1997), Hubbard (1995), Gertler and Gilchrist (1993), building on older ideas,2 the `lending view' predicts that the effects of monetary policy on aggregate demand, output and in¯ation should be in¯uenced (normally ampli®ed, as shown by Bernanke and Blinder, 1992) by the autonomous behaviour of commercial banks. The key condition for this to happen is that bank loans be imperfectly substitutable for other assets and liabilities, both in the balance sheet of commercial banks and among the forms of ®nancing available to ®rms and households. The monetary transmission process should thus be related to the structural (legal, regulatory, cultural, etc.) features of each country's banking and ®nancial sector that affect the extent to which direct bank lending can be substituted by other types of credit. As this chapter convincingly argues, a natural connection arises here with the so-called `legal 201
202 Ignazio Angeloni
view' of corporate ®nance and ®nancial structures, put forward more recently by La Porta et al. (1997, 1998) and Rajan and Zingales (1998). According to this literature, cross-country differences in corporate ®nancial structures are related to the way in which national legal systems accord protection to different classes on investors. Speci®cally, the `legal view' predicts that better creditor protection will facilitate external ®nance and help the development of active capital markets, at the disadvantage of bank intermediation, and that shareholder protection should make equity ownership more attractive to investors and thus favour equity relative to debt as a source of capital for non®nancial ®rms. Having laid down this conceptual premise, the chapter then applies it to the present situation of the Euro Area. This application is indeed tempting because the eleven European countries that have adopted a common currency still display considerable differences in ®nancial structures, banking regulatory and supervisory frameworks, let alone the legal systems in a broader sense. Consequently, the chapter's argument goes, the transmission of the single monetary policy to the Euro Area economy cannot be but diverse across countries, and will continue to be so unless and until radical steps are taken to create a common legal framework in the area. Unfortunately ± or perhaps one should say fortunately for the European Central Bank (ECB) and the success of its monetary policy ± after closer inspection such strong policy implication appears less convincing than the basic premise on which it is built. This is not only because, as one concludes from Table 5.7 (p. 185), legal systems in the Euro Area are not so diverse after all ± the only country with an Anglo±Saxon legal system is Ireland, and those belonging to the German or French systems, whose implications for monetary policy effectiveness are broadly similar, cover nine out of eleven countries or 97 per cent of the area's GDP ± but for other and more substantive reasons as well. I will organise my comments on them under three headings: (1) the role of other channels of monetary transmission (speci®cally, the exchange and interest rate channels); (2) the changing nature of the bank lending channel in the Euro Area; (3) the robustness (or lack thereof) of some of the chapter's empirical results.
1
Exchange and interest rate effects
The analysis of this chapter is limited exclusively to the role of commercial banks in the transmission process. However, a more complete and balanced view of the effects of Stage Three of EMU should consider also the changes that will take place in other channels of monetary transmission, particularly the exchange rate and the (structure of) interest rates. From the viewpoint of these two channels, the adoption of a single currency in the Euro zone should unambiguously result in a more homogeneous transmission of monetary policy across countries. Starting with the exchange rate, let us distinguish its contribution to the transmission of monetary policy in two steps. First, the exchange rate reacts to
Discussion 203
changes in the central bank controlled interest rates. This stage of the transmission chain, which crucially depends on monetary policy credibility and on the ®scal±monetary policy mix, is obviously identical across countries in a monetary union, while it could differ sharply when each participating central bank was acting in isolation within its own institutional and national macroeconomic policy framework. The second step, the impact of exchange rate changes on the economy, depends on the economy's degree of openness to international trade. Table D5.3 shows the percentage ratios exports over GDP in the individual countries of the Euro Area and in the area as a whole, where in the latter case only extra-area trade is considered. Not only is the average degree of openness of the area as a whole lower than that of each of the participating countries, but its dispersion across countries is also signi®cantly reduced when attention is focused on extra-area trade only. The impact on individual countries of any given change of the exchange rate of the Euro is less differentiated than those of each country's exchange rate changes before monetary union. This argument can be extended to the role that yield curve changes play in the transmission mechanism. Again, two steps can be distinguished. First, monetary policy shocks affect the level and shape of the yield curve, through a mix of liquidity and expectations effects. It has been shown (Angeloni and Rovelli, 1998) that the impact on the yield curve of monetary policy changes varies across monetary and ®scal regimes, and that it is related to monetary policy credibility. Such differences, again, are eliminated in a monetary union, where a unique yield curve prevails across the whole area with the only exception of small cross-country spreads that depend on bond market liquidity and on different credit ratings of sovereign debtors. The second step is given by the reaction of the private sector's spending behaviour to yield curve changes. Here the judgement of the possible effects of the adoption of the single currency is, as in the case of the exchange rate, more complex. Interest rate effects depend partly on the shape of private agents' intertemporal preferences, for which no change should a priori be expected. In addition, life cycle theory suggests also that these effects should depend also on the size and composition of the household sector's ®nancial wealth. The income effects of a given interest rate change will be larger, ceteris paribus, the shorter the ®nancial duration of the household net ®nancial wealth and, for a given duration, the higher the size of wealth. Euro Area countries have historically been characterised by widely divergent degrees of liquidity of ®nancial assets, due to different public debt management policies.3 As a result of EMU, and in the context of lower and more stable in¯ation, a convergence process is taking place across Euro Area countries in the maturity structure of public debts; moreover a convergence of debt ratios will result from the application of the Stability and Growth Pact. As this process unfolds, income and substitution effects on consumers following interest rate changes are likely to follow a more homogeneous pattern within the Euro Area.
204 Ignazio Angeloni Table D5.3 Openness to international trade in the Euro Area Total exports/ GDP
Share outside EU-11
Exports outside the Euro Area/ GDP
Belgium±Luxembourg Germany Spain France Italy Netherlands Austria Finland Portugal Ireland
62.6 24.3 19.6 20.4 20.8 43.4 28.4 32.9 22.7 44.0
42.8 57.7 41.2 51.6 56.2 42.7 45.2 70.8 35.2 61.3
26.8 14.0 8.1 10.5 11.7 18.5 12.9 23.3 8.0 27.0
Average
Unweighted Weighted by GDP
31.9 25.5
50.5 52.8
16.1 13.3
Standard Unweighted deviation Weighted by GDP
14.1 10.8
11.0 25.2
7.3 5.9
Source: IMF Direction of Trade Statistics (1997 data).
Note: GDP weights are obtained from national GDP converted into 1997 ECU exchange rates.
To summarise, these arguments suggest that the introduction of the Euro is likely to bring about a convergence of the monetary policy transmission among participating countries, if attention is focused on exchange and interest rate effects.
2
The lending channel: how will it change in Stage Three?
According to Kashyap and Stein (1994), three conditions must be present for such channel to operate. First, monetary policy shocks must affect the overall size of the banks' balance sheets ± i.e. money demand must be interest elastic; second, loan supply must respond to money demand changes; third, loans must be `special' ± that is, not easily substitutable by different forms of ®nancing for the private sector. On all these aspects, signi®cant changes are in process in the Euro Area. In recent years, ®nancial innovation has been affecting the nature and properties of money demand in several countries. New ®nancial instruments, close substitutes to money, have enabled banks to lower the burden of reserve requirements and to increase the ¯exibility and diversi®cation of their balance sheet's liability side. Often, new bank debt instruments have been introduced, or existing ones have been modi®ed, in response to tax or regulatory changes.
Discussion 205
The growing importance of near-money instruments such as certi®cates of deposit, repurchase agreements, short-term bank bonds, etc. has induced central banks to broaden the de®nition of their monetary aggregates. The M3 de®nition adopted by the ECB includes all bank liabilities with a maturity up to two years; within this maturity span, banks may be able to compensate a policy induced shortage of sight deposits by raising funds in the near-money segments, thereby reducing the central bank's ability to control the bank balance sheets and the supply of loans.4 This may actually not be true for all individual banks or bank categories, as an active liability management may be possible only for the more sophisticated and larger players. However, the process of bank concentration experienced in recent years in Europe (Table 5.2, p. 177) will tend to spread this phenomenon across the area, though perhaps at different paces in each national market. A parallel process of liability diversi®cation is likely to take place, over time, also for non-®nancial ®rms. Though only very preliminary signs are yet apparent, developed and diversi®ed private securities markets are likely to emerge as a consequence of EMU, as investors will look for new instruments of risk taking and diversi®cation. This should go in the direction of reducing the dependence of corporate borrowers on bank loans as a form of ®nancing. Finally, one should not underestimate the contribution that EMU will provide in enhancing cross-border competition in bank lending. With the elimination of currency risk, borrowers will increasingly be able to search for the best credit conditions available to them across the whole area. In practice, informational asymmetries and other tax and regulatory obstacles will continue to be important factors in cross-border lending, speci®cally in relation to small lenders and borrowers. However, competition will progressively erode these segmentations and tend to give rise to broadly uniform credit conditions for homogeneous classes of borrowers,5 thereby mitigating any differences that exist, at the national level, in the impact of bank lending behaviour in the transmission of monetary policy.
3 Two puzzling empirical results Before concluding, I turn to the empirical evidence presented in the chapter. I do it brie¯y because such evidence is admittedly suggestive, and therefore should be taken more as an indication of how and in what direction further research should proceed. Nevertheless, two speci®c results should, in my view, be singled out as particularly surprising and worthy of deeper investigation. First, looking at the indexes presented in Table 5.5 (p. 181), and sumarised in Table 5.8 (p. 187), countries belonging to the English-type legal family ± notably, the United Kingdom ± emerge as being characterised by a comparatively low effectiveness of monetary policy. This is in sharp contrast with a commonly held view that Anglo±Saxon countries' household sectors
206 Ignazio Angeloni
should be relatively more sensitive to interest rate changes. At least two factors justify this presumption: the higher relevance of consumer credit and housing mortgages, that tend to strengthen liquidity and income effects following short interest rate adjustments, and the higher incidence of shares in private portfolios, that give rise to stronger wealth effects. The empirical literature is in agreement with this presumption: for example, the detailed cross-country study by the BIS (1995) provides convincing empirical backing for it, based on structural econometric models. Again, this puzzle may partly be explained by the fact that this chapter concentrates solely on bank lending effects, disregarding the interest and exchange rate channels. Second, I found somewhat confusing that the signs of the predicted in¯uence of the legal family on the effectiveness of monetary policy differ in sign, in Table 5.8, according to whether the impact on output or in¯ation is considered. Since all indicators analysed in the chapter refer to effectiveness of monetary policy in affecting aggregate demand, one would expect the effects on prices and quantities to have the same sign, and to vary in relative size according to the nature of the in¯ation±output tradeoff.
4
Conclusions
As I stated at the outset, this is an interesting chapter, fun to read, one that stimulates thinking. Its central statement, that the legal system may be an important factor shaping the transmission of monetary policy, is convincing and justi®es further research. However, its empirical results and policy conclusions regarding the Euro Area should be looked at with considerable caution. Relevant areas of the overall monetary transmission process are not given explicit consideration, in order to focus speci®cally on bank lending behaviour and on the nature of the borrower ± lender relationships. Though this simpli®cation may be expositionally convenient, it should be borne in mind that it could change not only the size, but also the sign, of the results. Some puzzling empirical estimates mentioned above may derive precisely from this somewhat unbalanced view of the transmission process. The evolution in process in the European banking markets does, in my view, deserve more attention than it receives here. Restructuring, consolidation and competition are likely to change considerably the European banking landscape and the way bank behaviour affects the transmission process, both in the aggregate and across different countries, sectors and classes of borrowers. The `credit channel' in the Euro Area may well have to be studied anew, building on existing knowledge but taking into account the relevance of the regime change we are living through. Indeed, for the Eurosystem this may be one of the most challenging and policy-relevant research issues in the years to come.
Discussion 207
Notes 1. The views expressed here are the author's, and do not necessarily re¯ect those of the European Central Bank. I am grateful to Benoit Mojon for useful discussions. 2. The modern literature on the `lending view' has elements in common with the old one on the `availability doctrine' (see Roosa, 1966). The main difference is that the new view does not necessarily rely on credit rationing: loan market effects can take place in equilibrium, through changes in the spread between the loan rate and the money market rate. 3. I am assuming here that households do not fully internalise future tax liabilities, ± or, equivalently, that the public debt is at least partly perceived as net wealth by consumers. 4. According to econometric estimates (Coenen and Vega, 1999) the area-wide M3 aggregate is not controllable in the long run via interest rate changes only. Controllability is restored if one also considers the effects on output and prices. 5. Since the mid-1990s, a sharp convergence was indeed observed in lending rates across EMU countries. Such convergence was, however, related not only ± or not much ± to cross-border competition in lending, but to a large extent to the general convergence of in¯ation and interest rates.
References Angeloni, I. and Rovelli, R. (1998) Monetary Policy and Interest Rates, London, Macmillan. Bank for International Settlements (BIS) (1995) Financial Structure and the Monetary Policy Transmission Mechanism, Basle, BIS, March. Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of Monetary Transmission' American Economic Review, 82, 901±21. Coenen, G. and J. L. Vega (1999) `The Demand for M3 in the Euro-Area', European Central Bank, Working Paper Series 6, September. Gertler, M. and S. Gilchrist (1993) `The Role of Credit Market Imperfections in the Monetary Transmission Mechanism: Arguments and Evidence', Scandinavian Journal of Economics, 95, 43±64. Hubbard, R. G. (1995) `Is There a ``Credit Channel'' of Monetary Policy?', Federal Reserve Bank of St Louis Economic Review, 77, May±June, 63±77. Kashyap A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending', In N. G. Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press for NBER. ÐÐÐÐ (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for the European Monetary Union', Federal Reserve Bank of Chicago Economic Perspectives, September±October, 2±18. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (1997) `Legal Determinants of External Finance', Journal of Finance, 52, 1131±50. ÐÐÐÐ (1998) `Law and Finance', Journal of Political Economy, 106, 1113±55. Rajan, R. G. and L. Zingales (1998) `Which Capitalism? Lessons from the East Asian Crisis', Bank of America, Journal of Applied Corporate Finance, 11, 40±8. Roosa, R. V. (1966) Interest Rates and the Central Bank; reprinted in Richard S. Thorn, (ed.), Monetary Theory and Policy, New York, Random House, 559±83.
6
Differences Between Financial Systems in European Countries: Consequences for EMU1 Reinhard H. Schmidt
1
Introduction
There are three main issues which must be discussed if one wishes to answer the question which this chapter addresses: (1) Which features of a ®nancial system are important for monetary policy or, in other words, how is monetary policy conducted, and how does it affect the real economy, and how and to what extent does this depend on the speci®c features of an economy? (2) How different are the ®nancial sectors ± or more generally the ®nancial systems ± in Europe? (3) If signi®cant differences existed between countries, would this have consequences for how monetary policy should, and can, be conducted in a common currency framework? I want to warn readers at the outset that I am not a ®nancial macroeconomist. Because this is not my ®eld of specialisation, my knowledge of how monetary policy is conducted and how it affects the real economy ± that is, of the so-called transmission mechanism ± is only super®cial. I can only hope that this does not invalidate all that I would like to offer in my attempt to ful®l the role assigned to me as a microeconomically oriented scholar in banking and business ®nance. In order to clarify the basis on which I stand, I want to ®rst present two simple ideas about why differences between the national ®nancial systems in Europe might matter for monetary policy in `Euroland'. In essence, I see monetary policy and its effects as follows. There is a central bank which has the ability to provide a monetary policy impulse: the central bank can, for instance, raise the short-term interest rate or reduce the quantity of central bank money in the economy (and will typically do both at the same time). This impulse has immediate consequences for the interest rate structure and the quantity of money (according to some speci®c de®nition of a 208
Differences Between Financial Systems in European Countries 209
monetary aggregate) in the economy and for the banking sector. To the extent that banks re®nance themselves, or hold voluntary or compulsory reserves, at the central bank, they have fewer funds to offer to their potential borrowers and/or lending becomes more costly to them. The resulting change in the situation of the banking sector has, in turn, consequences for the real economy to the extent that the ®rms, households and government bodies that want to borrow from the banks have less credit available, or have to pay more for the funds borrowed from the banking sector. The restrictive monetary policy impulse thus works itself through the system and determines aggregate demand and ultimately output and income, and possibly the price level as well. Some details of this process have to be left for further discussion, but evidently I assume that monetary policy is effective, at least in the short run. The second idea is taken from a paper by Dornbusch, Favero and Giavazzi (1998). Differences between the national ®nancial systems might matter because different national economies might react differently to monetary policy impulses of a given kind and size. If this is so, it might be so because of differences in the ways in which central bank action in¯uences the respective ®nancial sectors and in particular the banking sectors, or because of differences in the ways in which a change in the situation of the ®nancial sectors affects the real economy. If there are two countries, A and B, in a monetary union, with A being more sensitive to monetary policy impulses than B, then a given monetary policy measure might be too strong for country A and too weak for country B even if it is just right for the intended effect on the (weighted) average of the two economies in terms of, say, their in¯ation rates. If I may claim any comparative advantage at all for discussing the consequences of national differences for EMU, it has to do with the second question ± that is, the differences between the ®nancial systems in Europe.2 However, here I must also make two reservations: From the way I have formulated the title of my chapter and the questions listed above, it can be seen that I will not con®ne myself to looking at differences between ®nancial sectors, but wish to ± and will later show that I need to ± discuss differences between national ®nancial systems. This is not simply a terminological difference, but a difference in substance. I use the term `®nancial sector' in a narrow way: it denotes those specialised institutions such as banks, pension funds, securities markets, etc. which provide ®nancial services to the non-®nancial sectors of the economy, and to the ways in which these institutions are shaped and managed and how they operate and are regulated. The term `®nancial system' encompasses not only the ®nancial sector, but also the real sectors to the extent that they demand the ®nancial services of the ®nancial sector and also to the extent to which they forgo using the ®nancial sector, as well as the interaction between the demand for and the supply of the services of the ®nancial sector. Thus, for instance, the extent to which internal ®nancing of investment takes place, the extent to which saving takes the form of real investment, the extent to which banking services are
210 Reinhard H. Schmidt
appropriate to the demand for them, etc. are features of a given ®nancial system.3 The second reservation refers to the countries which I will take into consideration. Anticipating a likely future course of events, I will assume that the United Kingdom is already part of the European Monetary Union (EMU). Moreover, my empirical references will be mainly restricted to the three largest ®nancial systems in the `enlarged EMU' ± namely those of France, Germany and the United Kingdom. Even if the Italian economy were larger than that of Great Britain, I would wish to include that of the United Kingdom because, as I will describe below, its ®nancial system differs markedly from those of the continental European economies, and from that of Germany in particular. An understanding of these differences helps to underscore the heterogeneity that makes for national diversity, and thus for possible problems of a common monetary policy, in the EMU. The chapter is structured as follows: In Section 2, I shall discuss the transmission mechanisms. Given my background, it appears almost selfevident to me that the so-called `credit channel' is important,4 though not so much as an alternative to the `conventional' channels of the transmission of monetary policy, but rather as a complement to them. The discussion of the various channels serves the purpose of providing the criterion for selecting those aspects of national ®nancial systems which I will then go on to characterise in some detail. In Section 3, I will argue that the ®nancial systems ± as well as the ®nancial sectors ± of the three countries are vastly different, and point out the main differences. In Section 4, I will bring the two lines of reasoning together and argue that, in spite of what I perceive as important differences, I do not think that these differences matter very much for the design and conduct of a common monetary policy. The main reason for this is that what I consider to be the characteristic features of the different ®nancial systems ± though not of the different ®nancial sectors ± might have two effects on the functioning of monetary policy which tend to offset each other.
2
Transmission mechanisms or channels of monetary policy
The classi®cation of channels As the name suggests, a transmission mechanism is a conceptual or formal model of the ways through which monetary policy in¯uences the real economy. These ways are complicated and, as far as I can judge, still imperfectly understood. In the relevant literature,5 four different transmission mechanisms or channels are distinguished: (a) (b) (c) (d)
the interest rate channel the channel of relative prices the exchange rate channel, and the credit channel.
Differences Between Financial Systems in European Countries 211
The interest rate channel (a) and to a certain extent also the channel of relative prices (b) are standard elements of what may be regarded as the traditional view of the transmission mechanism. The exchange rate channel (c) is not relevant in the context of the present chapter, as the ± enlarged ± EMU area is a very large economy for which exchange rates are not a terribly important factor because external trade accounts for only a relatively small share of total GDP. The credit channel (d) is the `newcomer' in the market of ideas and by now the main `competitor' to the incumbents (a) and (b). The interest rate channel and the channel of relative prices The interest rate channel is based on the conventional Keynesian IS ± LM model. According to this model, the central bank determines short-term interest rates. With given and unchanged expectations about the in¯ation rate, this also has an effect on real longer-term interest rates, which determine the investment decisions of pro®t maximising ®rms, as the ®rms compare marginal internal rates of return on their investment projects with `the cost of capital' when they decide whether to invest or not. Similar considerations apply to certain consumption decisions,6 so that an unexpected change in that interest rate which the central bank can set in¯uences aggregate demand and ultimately also output. Given that prices are in¯exible only in the short run ± not only by de®nition, but also in reality ± an expansionary or restrictive monetary policy impulse provided by the central bank loses its effect on the real economy over the course of time when prices start to react. In order for the interest rate channel to function, two conditions must be met: for one thing, monetary policy must not only affect short-term interest rates but also (real) medium to long-term rates, and, for another, investment decisions must be interest-elastic. Differences between ®nancial sectors and, more generally, ®nancial systems can translate into differences in terms of the strength of the links between nominal short-term and real longer-term interest rates and in terms of the strength of the links between these rates and investment and consumption decisions. The effects of monetary policy are stronger if ®nancial contracts and, in particular, the terms over which interest rates are contractually ®xed are shorter and thus more easily adjusted. Another factor which may in¯uence the effectiveness of monetary policy, as its working is described in the interest rate channel view, is the extent to which central bank money is being used in the economy. The channel of relative prices (b) ± which is also sometimes called the `asset price channel', which is presented in two different versions in the literature, namely in the `monetarist' version and that developed by Tobin (1969) ± assumes that central banks in¯uence the composition and/or prices of the assets which are held in the portfolios of economic agents. An unexpected monetary policy impulse disturbs the equilibrium composition of the portfolios and induces attempts to adjust their composition. Ultimately, an
212 Reinhard H. Schmidt
expansion of the monetary base by open market operations (OMOs) leads to more demand for securities, to rising security prices and thus falling interest rates and possibly also to an increase in consumption and investment expenditure. The extent to which these effects on asset prices merely lead to price-level changes, or also to real effects, depends on the rigidity of prices and the `disturbances' which may arise from changes in the expectations concerning the future monetary policy. The two versions mentioned above differ with respect to the asset categories which the agents try to bring into balance, and their respective rates of return and also as regards the extent to which the process of the restructuring portfolios directly affects the demand for real versus ®nancial assets and thus investment decisions. According to the proponents of the asset price channel, the main instrument of the central bank is its in¯uence on the quantity of central bank money in the economy. This suggests that different procedures in which central banks operate in practice, and, in particular, differences in terms of the types of ®nancial assets which are eligible as reserves, may lead to differences in the effectiveness of monetary policy. However, even also including the United Kingdom, the differences between the various economies in Europe which existed in these respects in the past, are giving way to a common approach.7 Another aspect, and one in which the various European economies still differ considerably and which is important for this channel, is the extent to which the agents hold ®nancial assets whose prices may vary in reaction to central bank policy. The larger the share of such ®nancial assets in agents' portfolios, the greater will be the likelihood that central bank impulses will work in the ways postulated by the advocates of the asset price channel.8 It appears plausible that monetary policy is able to have an effect on the real economy primarily because of interest effects and also, to a certain extent, because of asset price effects. However, the recent empirical literature argues that in reality the effects of monetary policy, in particular those of a monetary contraction, are stronger and of a different pattern than those which could be expected if the interest rate and asset price channels were the only relevant mechanisms, and also that they exhibit a different pattern than the effects one would expect to encounter if these were the only pertinent mechanisms.9 At a theoretical level, the interest and the asset price channels assume that the agents in the economy behave in an overly mechanistic fashion, and fail to address directly the question of how the ®nancial sector, and, in particular, the banks, react to monetary policy impulses. These two weaknesses suggest that looking only at the traditional channels might prevent one from acquiring a deeper understanding of the implications of intercountry differences for a common monetary policy. The credit channel At the heart of the so-called credit channel are those aspects which are largely left out by the two traditional channels. Because of the information and
Differences Between Financial Systems in European Countries 213
incentive problems which are widely discussed in the current corporate ®nance literature, ®nancial systems do not function in a frictionless manner, and for many non-®nancial ®rms external ®nancing is simply dif®cult to obtain and more costly than internal ®nancing. The cost difference is called the `external ®nance premium'. In the relevant theoretical and empirical literature this external ®nance premium is assumed to be not only an expense incurred in addition to the basic interest costs, but also a positive function of the interest rate. In other words, if the central bank raises or lowers `the interest rate', the external ®nance premium will also go up or down.10 The second aspect which is highlighted in the credit channel view of the transmission mechanism comprises the availability of bank credit and the speci®c quality ± i.e. the limited substitutability ± of bank credit. Working in combination, the external ®nance premium and the availability of bank credit strengthen the effectiveness of monetary policy considerably, and it seems thoroughly plausible to assume that these two factors may differ much more between countries than those at which the traditional views of the transmission mechanism focus. As this brief introduction suggests, the credit channel can be broken down into two separate but complementary channels. The ®rst one, the broader of the two, is called the `balance sheet channel'. It focuses on the ability to borrow. The external ®nance premium for a given borrower is determined by its ®nancial position, in particular by its net worth, and the value of the collateral which it can provide. The borrower's ®nancial position is in¯uenced by the monetary policy and by the business cycle. A restrictive monetary policy raises the interest rates, reduces the cash ¯ow of ®rms and depresses net asset values of borrowers and the value of their collateral, and thus may severely restrict their ®nancing options and raise the premium. The likely consequence is a reduction of investment and in particular of investment in working capital, which is typically ®nanced by short-term bank credit. Owing to economy-wide accelerator effects, the impact of a monetary contraction may not only be stronger, it may also last longer than the traditional view suggests.11 The second `branch' of the credit channel is the more narrowly de®ned bank lending channel. Its proponents proceed from the highly plausible assumption that the central bank is in a position to limit the quantity of credit which the banking sector can provide to borrowers. A restriction or rationing of bank credit in turn restricts the scope of ®rm investment; and this is all the more likely the less bank credit can be substituted by other sources of funding at the bank and ®rm level. In claiming that bank credit is indeed dif®cult and in some cases even impossible to substitute as it provides a certain liquidity insurance, advocates of the credit channel of the transmission mechanism borrow heavily from the recent advances in the theory of ®nancial intermediation which shows why `bank loans are unique'.12 Banks are specialists in lending to ®rms in those cases in which it is important to
214 Reinhard H. Schmidt
monitor the borrower carefully ± or, in other words, in overcoming information and incentive problems. The balance sheet channel and the bank lending channel, moreover, interact in such a way that the effects of monetary policy on the ability of ®rms to borrow from banks and on the ability of banks to lend, reinforce each other. This makes their relevance for the effects of monetary policy all the greater. As these brief explanations suggest, it would be wrong to consider the credit channel as an outright alternative to the interest rate channel. Instead, the effects of borrowing capacity and the availability of bank credit which this channel emphasises reinforce those effects which have traditionally been assumed to exist and to underlie the transmission of monetary policy into the real economy. But both for practical monetary policy and for the problem which is discussed in this chapter ± namely that of possible consequences of differences between national ®nancial systems ± it is important to know more than merely that monetary policy has an effect. One needs to know why it matters and how it affects the real economy in order to be able to determine the direction and the strength of policy measures.13 The credit channel view suggests a list of items which can be used to check which elements of a ®nancial system might be particularly relevant when one tries to analyse the implications of differences in ®nancial systems between countries for monetary policy. This list would be too comprehensive to develop here, and I will thus restrict myself to indicating classes of factors. The central bank and the money market structure are the ®rst category. Some of the key questions here are: How is monetary policy implemented? Who interacts directly with the central bank? How deep and liquid are the markets for short-term securities like CDs, commercial paper and government bills? What scope for administrative interference in the short-term operations of banks do the government and the central bank have? At the level of the banking sector, some of the relevant issues are: What are the most prevalent types of banks? What is the typical relationship between banks and ®rms? What are the legal forms and ownership patterns of banks? What is the level of concentration and competition in the banking sector? What is the nature of the relationship between banks and non-bank ®nancial intermediaries (NBFIs)? Are the NBFIs competitors or subsidiaries of banks? What is the asset and liability structure of the banks' balance sheets? NBFIs and capital markets constitute the main alternatives to banks. They need to be looked at if we wish to determine the extent to which ®rms and banks are able to circumvent the effects of a restrictive monetary policy which would be transmitted through the bank lending channel. As far as NBFIs are concerned, some of the most important factors are (a) the nature of their relationship to the central bank and to the banks, (b) how important they are in the ®nancial sector, and (c) their asset and liability structures. As far as capital markets are concerned it is important to know which markets exist, who has access to them and how deep and how liquid they are.
Differences Between Financial Systems in European Countries 215
Finally it would be important to look in detail at the sector of the non®nancial ®rms and at the household sector to assess how the actors in these sectors are affected by monetary policy measures of the central bank. Key questions here would be: How are they typically ®nanced? What is the prevailing mode of holding liquidity? How many of them have access to security markets or to foreign sources of funding? And, last but not least, how are they governed and what is the time horizon of their strategies? If the goal is to conduct a reasonably comprehensive comparison of ®nancial systems to assess the consequences of national differences for a common monetary policy, then the view one adopts concerning the transmission mechanism of monetary policy evidently has a strong in¯uence on the speci®c choice of aspects which must be examined. The credit channel implies a much longer list of factors than the traditional channels. The characteristics of some of these items in the ®nancial systems of France, Germany and United Kingdom will be described on p. 218 and in the Appendix. However, even more important than the individual items in this list ± which is not intended to be complete ± is the question how they are related to each other. This is a topic to which I will return in the next section. In my opinion, the credit channel view is extremely helpful for an analysis of the consequences which differences in entire ®nancial systems have for the topic under discussion here ± an assertion whose validity I will seek to demonstrate in Section 4.
3 How different are ®nancial systems in Europe? How different can ®nancial systems be? To start this section, I want to recall the de®nition of the ®nancial system provided in the Introduction. The ®nancial system includes the ®nancial sector as the provider of ®nancial services as well as the real sectors of the economy in so far as they demand ± or, as the case may be, fail to demand ± these services, and the complex relationships between the ®nancial and the non-®nancial sectors. Evidently, it does not only depend on the ®nancial sector but also on the real sectors and the relationships between them how monetary policy affects the real economy. Thus, the broad de®nition is appropriate for any analysis of monetary policy. This de®nition is also important for another reason. It helps to get a more comprehensive picture of the extent to which the ®nancial systems of countries differ, and the respects in which they differ, than could be obtained using the narrow concept of the ®nancial sector. Indeed, as I will argue in this section, both ®nancial sectors and ®nancial systems differ between countries, but the latter are more dissimilar than the former, and differences in other parts of the ®nancial systems ± i.e. in parts other than the ®nancial sectors ± could be highly relevant for the design and conduct of a common monetary policy. Thus the broad de®nition of the ®nancial system is particularly
216 Reinhard H. Schmidt
Financial System (1) Financial sector
(2) Financial patterns
(Role of institutions and market)
(Saving and financing)
(3) Corporate governance
(4) Corporate strategy
(Roles of various stakeholders)
(Ability regarding fundamental change)
Figure 6.1 The ®nancial system
appropriate for studying the problems of a common monetary policy in different countries, as it points to the fact that those aspects of the real economy which are important for the working of the transmission mechanism and differ between countries may correspond in a systematic or nonaccidental way to the differences between the ®nancial sectors. This correspondence is crucial for the effects of a common monetary policy. If I make this claim, I am obliged to take a closer ± which is at the same time also a more abstract ± look at the concept of a ®nancial system.14 The broadly de®ned ®nancial system is an open system; there is no clear factual criterion to determine what belongs to it and what does not. Rather, the boundaries of the system at large are drawn by the observer who, in a given case, seeks to address a certain problem in the most fruitful way. For the purpose of this chapter, one can distinguish four closely interrelated sub-systems which are, in turn, composed of several elements. Figure 6.1 shows these sub-systems, which are at the same time elements of the larger system, the ®nancial system. They include, ®rst of all, the ®nancial sector. The second sub-system consists of the surplus and the de®cit units in so far as they provide funds to the ®nancial sector or obtain funds from it, and thus of the savings behaviour of households and of the ®nancing of business or corporations and governmental bodies, and of the ®nancial instruments used by the parties to ®nancial contracts. It is advisable to include corporate governance as the third sub-system and corporate strategies and structures ± for short: the business system ± as the fourth. Note that there is no need to de®ne these sub-systems in such a way that they do not overlap. It is imperative to see how these sub-systems are composed and function individually, and how they are related to each other. As the colloquial use of the term `system' suggests, a system is more than a collection of elements. For a system to be `really a system', its elements must `®t together'. In economic life,
Differences Between Financial Systems in European Countries 217
one can assume that it has positive consequences if this is the case, and negative consequences if the elements do not ®t. One can speak of `important' differences between ®nancial systems, if not only the main elements of which they are composed, but also the way in which they are related, are different. The interesting thing about the sub-systems and their interrelationship, and thus the interesting thing about the entire ®nancial system, is that they are composed of complementary elements. Elements of a system are called `complementary' (to each other) if they mutually increase the `bene®t' they yield in terms of whatever the objective function or the standard for evaluating the system may be, and mutually reduce their disadvantages or `costs'.15 A system is called consistent if its complementary elements indeed take on the values which make the system attain a local optimum. Systems of complementary elements typically have more than one optimum, and the local optima are clearly distinct con®gurations of the values of the elements. Financial systems are systems in this speci®c sense. The complementarity of their elements and the economic bene®ts which a consistent ®nancial system can be assumed to produce are the factors which account for the tendency of countries' ®nancial sectors and non-®nancial sectors to co-vary in a systematic way. To illustrate the concepts of complementarity and consistency in the case of ®nancial systems and also to lay the foundation for the next argument, I would now like to take a look at common two-way classi®cations from the literature for the four sub-systems shown in Figure 6.1. (1) For the sub-system `®nancial sector' it is common to distinguish between a bank-dominated and a capital market-dominated variant based on certain measures of the size and `importance' of banks and capital markets, respectively. (2) The second sub-system refers to how savers hold their ®nancial wealth and the sources from which corporations obtain the bulk of the funding for their investments. This suggests a similar two-way classi®cation distinguishing a bank- and a capital market-oriented system.16 (3) With respect to the third sub-system, corporate governance, there are two-way classi®cations according to different criteria depending on whether the safeguarding of the interests of owners, of creditors, or of the staff is involved. The details need not interest us here.17 Suf®ce it to say that these classi®cations can be brought together under the heading of `insider-' and `outsider-controlled' corporate governance systems.18 (4) Lastly, ®rms can be classi®ed according to their business systems into two groups: those which make ample use of ®rm-speci®c human capital, rely to a large extent on long-term and trust-based relationships ± or implicit contracts ± with their various `stakeholders' and partners, and are more suited to undertake strategic adjustments through longer sequences of small changes, and those with the opposite characteristics.19
218 Reinhard H. Schmidt
Looked at individually, each of these eight speci®c sub-systems can be interpreted as consistent systems of complementary elements. Owing to space limitations, I cannot demonstrate this here. Instead, I will concentrate on the question of how the four sets of sub-systems are related to each other: The subsystems are themselves complementary elements of the overall system; and only two speci®c combinations constitute consistent systems. As far as the ®rst two elements are concerned, this seems plausible. Even though there may be an empirical problem in demonstrating this correspondence, which I shall discuss below, one would tend to think that there is a one-to-one correspondence between the two classi®cations in these two sub-systems: a system in which savers entrust a large fraction of their wealth to banks or closely bank-related non-bank ®nancial intermediaries should also be one in which the banks are the dominant element in the ®nancial sector; and capital market dominance should equally be re¯ected in the patterns of saving and ®nancing and the size and role of the capital market as an institution. Moreover, a bank-dominated ®nancial sector and a bank-oriented system of savings and ®nancing also `®t', or correspond in functional terms, to an insider-controlled governance system and a business system with much ®rmspeci®c human capital, many implicit contracts and a tendency towards gradual change.20 The set of other `values' for the four sub-systems as elements also constitutes a consistent ®nancial system. It is characterised by a greater importance for capital markets than for banks in the ®nancial sector, more saving and ®nancing via the capital market than with and from banks, outsider-controlled corporate governance and highly ¯exible corporations which are able to undertake strategic adjustments with `big leaps'. Evidently, these two systems are clearly distinguishable types of ®nancial systems or, to use Williamson's suggestive term, `discrete alternatives'.21 So at least in theory, ®nancial systems can be fundamentally different. And since the four sub-systems in¯uence each other in their functional values, entire ®nancial systems can be `more different' ± i.e. different in a more fundamental way ± than ®nancial sectors alone might. How different are those in Europe really? The fact that one can distinguish two types of ± consistent ± ®nancial systems at a conceptual level does not imply that one could also ®nd these different types in reality, nor does it imply that real ®nancial systems do in fact correspond to this typology. Therefore, the questions of whether real ®nancial systems have complementary elements, and thus whether they are `really systems' ± and, if they are, whether they are also consistent systems ± are empirical questions. It should also be noted that the theoretical distinction does not tell us whether the ®nancial systems of the major European economies are different. Thus, we shall now turn to the empirical issues. It goes without saying that ± not only owing to the limited space available ± only a very brief discussion of this complicated subject can be provided here;
Differences Between Financial Systems in European Countries 219
however, I hope that by drawing attention to a few key aspects I will be able to make my point. It is equally clear that one cannot expect to ®nd a perfect correspondence between a given type of ®nancial system and a given real ®nancial system ± i.e. that of a speci®c country. Nevertheless, the classi®cation of ®nancial systems suggests differences in reality.22 The example which immediately come to mind are the German and the British ®nancial systems ± and this is the main reason why I anticipate a likely future course of events and assume that the United Kingdom is already part of `Euroland'. The German ®nancial system is of the ®rst type. The German banking sector is big and some banks are important and powerful, whereas capital markets and capital market-oriented institutions such as pension funds are `underdeveloped', as can be measured by indicators such as the total assets of banks and stock market capitalisation as percentages of GDP, respectively.23 Corporate governance in Germany functions mainly through internal mechanisms and involves `insiders' to the corporations ± or, in other words, people and institutions which have long-term interests in safeguarding their speci®c relationships with a corporation and that are typically better informed about its prospects and problems than anonymous market participants could be.24 There does not seem to be an active market for takeovers. The role of banks, of other corporations and of employees having codetermination rights in the governance of corporations epitomise this system, and despite the widely publicised declarations to the contrary, the maximisation of shareholder value is not the dominant objective of most large German ®rms. Even though this may be more speculative, one can also add the observation that, at least on an economy-wide level, German ®rms are more woven into nets of implicit contracts, that labour turnover is lower and that the corporations are less able and willing to undertake abrupt adjustments to changing circumstances than British corporations. Thus it is fair to say that at least as regards the ®rst, the third and, to a certain degree, the fourth element (or sub-system), the German ®nancial system is consistent. Without doing too much violence to a reality which is of course much more complex, one can characterise the British ®nancial system as one in which core elements ± or sub-systems ± (1) and (3), and possibly also (4),25 take on values which are precisely the opposite of those determined for Germany. In the United Kingdom, the relative importance of banks in the ®nancial sector is not as great; capital markets and institutions ± notably pension funds, which are capital market-oriented and in most cases independent of the banks ± play an important role; there is no formal and no de facto codetermination; and ®nancial intermediaries are scarcely involved in the governance of corporations.26 The corporate objective is unrestricted maximisation of pro®ts or shareholder value, and given the nature of the UK ®nancial system there is also no reason why this should be otherwise. The basic mechanism of corporate governance is the takeover market, and the entire corporate
220 Reinhard H. Schmidt
governance system is clearly an outsider control system.27 Overall, the British ®nancial system relies very much on market mechanisms and less on institutions in the conventional meaning of the term ± i.e. it relies less on groups of people who interact directly with each other over a long period and not only have a certain degree of mutual commitment but also ®nd that their interaction gives rise to intense con¯icts over important issues which are, however, resolved internally and ultimately in a peaceful way.28 Despite far-reaching changes in the regulatory and competitive environment in which ®nancial sectors and non-®nancial corporations in Germany and the United Kingdom operate, it is fair to say that the characteristic features of these two ®nancial systems, as far as I have characterised them up to now, are surprisingly stable. Many details of the ®nancial systems have been modi®ed, but their fundamental or structural features have not changed over the last twenty years ± and, indeed, they have probably remained essentially unaltered for much longer than that. In the case of the French ®nancial system, it is more dif®cult to ®nd a unifying `logique'. At least the efforts in this area which I have undertaken with colleagues and students ± for some time now, we have been trying to locate the French ®nancial system along the spectrum of market and institution-based (or Anglo±Saxon and German-type) systems ± have not been successful.29 This is so for at least two reasons. One is the strong role which the French state (or government) has played in the ®nancial system over the years. It is extremely dif®cult to integrate the extent and the forms of state or government intervention into the two-way classi®cation which was described on p. 218, or to invent a different classi®cation incorporating the additional dimension of state activity.30 The second reason is that the French ®nancial system and its four sub-systems have experienced dramatic and fundamental changes during the last two decades. If one looks at individual elements, it appears that these changes have already converted the French ®nancial system from being of the German type ± except for the pervasive state intervention ± to one of the Anglo±Saxon type.31 This may indeed be the case. But in the past the French ®nancial system had a number of elements which were certainly not congruent with the German model or the German ®nancial sector type, and today it includes several elements which do not `®t' the general structure of the Anglo±Saxon model or ®nancial sector type. Referring not only to European ®nancial systems, but also to those of Japan and the United States, Figure 6.2 summarises the characteristic features of consistent ®nancial systems. The tables in the Appendix provide empirical support for my brief summary description of the three ®nancial systems. Nevertheless, it would be good to have more empirical evidence to underscore the profound differences between Germany and the United Kingdom and to support my view that the system in France does not conform to either of the two `polar' models and is at present unstable in so far as it is undergoing a process of rapid change. In particular, additional evidence is
Differences Between Financial Systems in European Countries 221 Stability of business environment Higher
– Control – Relationships – Inside influence – Insider control
BE
BE
Limited coherence
Higher
Lower
FP – Households hold primarily bank deposits – Bank loans as main financial source of firms CG
Higher
System-wide role of market mechanisms
Lower
FS – Banks play dominant role
Lower Germany Japan
Limited coherence
CS – Firms-specific investments – Continuous adjustments
FS – Organised capital markets play dominant role
FP – HHs hold claims on funds and securities – Market-based firm financing
CG – Liquidity – Arm's length – Outside option – Outsider control Lower
USA UK
CS – Marketable investment – Fundamental strategic shifts
Higher
BE
BE
Figure 6.2 Consistent ®nancial systems Note: FS = Financial sector; FP = Financial patterns; CG = Corporate governance CS = Corporate strategy
called for in one respect: in my account of the ®nancial systems, I have intentionally left out the second sub-system of ®nancial patterns. It is logical to ask why, and of course the obvious answer is that it might not accord with the general pictures of the systems. And while the savings part does not offer any problem in this respect, the ®nancing part does. At least this appears to be the case if one accepts at face value the results of an important strand in the academic research about how ®rm investment is ®nanced. The ®ndings presented in this literature are plainly inconsistent with the view underlying my argument which is inspired by the theory-based assumption that there should also be differences in the ®nancing patterns of the ®rms in the different countries and that these should correspond to the differences between, both, the respective ®nancial sectors and corporate governance systems. This inconsistency has to be clari®ed if my claim that there are in fact fundamental differences is to be regarded as valid. In a series of highly in¯uential papers, Colin Mayer and others who followed his line of research32 have arrived at the opposite conclusion. First of all, they take the position that there are hardly any systematic differences between national patterns of corporate ®nancing, and secondly that those differences
222 Reinhard H. Schmidt
which do exist, are inconsistent with what one would expect on the basis of theory and of the empirical features of sub-systems (1) and (3) in the various countries. For instance, Mayer (1988, 1990) found in his early studies that bank ®nancing is more important in the United Kingdom than in Germany and that equity ®nancing is even negative in the United Kingdom. Corbett and Jenkinson (1996, 1997) supported these results in more recent and much more extensive studies. These ®ndings do indeed pose a puzzle. If they were true, they would break the logical chain connecting the four sub-systems and contradict the proposition that the German and the British ®nancial systems are consistent. Moreover, they would invalidate the whole concept of ®nancial systems as consistent sets of complementary elements. From the perspective adopted in the present chapter, it is fortunate that the methodology developed by Mayer and used by him and others has a ¯aw. Mayer and his followers look at net ¯ows of funds between economic sectors whereas they should have looked at gross ¯ows, because in the process of aggregating over real and ®nancial investments and over time, netting eliminates the relevant differences which might exist, and thus the role of external debt ®nancing disappears almost completely. In his dissertation, Andreas Hackethal (1999) has identi®ed this ¯aw, suggested a way to avoid it and recalculated the ®nancing structures of ®rms in Germany, Japan and the United States. As the ®rst columns per country (with the heading `gross') in Table 6.1 show, his correction yields patterns of long-term ®nancing which are consistent with expectations based on the structure of the entire ®nancial systems. The second columns (`net') and the ®rst line concerning internal
Table 6.1 Financing patterns in various countries, 1970±96, per cent of physical investment (gross: 1970±96; net: 1970±94) Germany Sources of ®nance Internal funds Bank ®nance NBFI ®nance Bonds Commercial paper New equity Trade credit, others and statistical adjustment
Gross Net 78.9 72.9 11.9 9.3 n.a. 7.0 ±1.0 0.0 n.a. 3.8 0.1 n.a. 10.1
Japan
USA
Gross
Net
Gross
69.9 152.3 0.0 14.3 5.7 3.8 n.a.
26.7 n.a. 4.0 n.a. 3.5 ±4.1
96.1 65.3 24.6 48.5 9.2 14.9 n.a.
UK Net
Net
93.3 11.1 1 4.6 n.a. n.a. 15.4 4.2 n.a. n.a. ±7.6 ±4.6 ±15.0 ±7.5
Note:
Net ®gures and those for `Internal funds' are taken from Corbett and Jenkinson (1997). They add up
to 100 per cent. These authors include both `NBFI ®nance' and `Commercial paper' in `Bank
®nance'(except for the United States where `Commercial paper' is included in `Others').
Gross ®gures, which by construction do not add up to 100 per cent, are taken from Hackethal (1999).
Only long-term external funds are shown.
Differences Between Financial Systems in European Countries 223
funds, for which the net and gross ®gures are identical, show the corresponding net ®gures derived by Corbett and Jenkinson (1997) for comparison.33 The data needed to recalculate the ®gures for French and British ®rms, and thus to provide a more accurate picture of their ®nancing patterns, are not yet available, but it seems plausible that, given the similarity of the ®nancial systems of the United States and the United Kingdom, those for the United Kingdom will not differ too much from those for the United States. If this is indeed the case, the claim that the British and the German ®nancial systems are consistent and differ in a fundamental way can be upheld. In another empirical study we have found further interesting evidence to support the assumption of fundamental differences between the British and the German ®nancing systems and the proposition that dramatic changes are under way in France.34 In this paper, we have calculated various intersectoral intermediation and securitisation ratios for the three countries over a span of ®fteen years. In contrast to the `conventional wisdom' that there is a general tendency towards disintermediation and securitisation and that, overall, the ®nancial systems in Europe are becoming more similar, our study shows that the levels of disintermediation and securitisation differ substantially between countries and that, except for France, they are surprisingly stable over time. As Figure 6.3 on p. 224 shows, the `liability±intermediation ratios of non®nancial companies', which measures the fraction of external ®nancing which comes from intermediaries, (Figure 6.3a) and the `liability-intermediation ratios of non-®nancial companies to banks', which measure the share of bank ®nancing in the total external ®nancing of ®rms (Figure 6.3b), and the ratios of securitisation of corporate ®nancing (Figure 6.3c) differ greatly between Germany and Great Britain, and are almost completely stable in these two countries. For France, these ratios exhibit not only a great instability, but also a tendency to move away from the German to the British model. This may suf®ce to demonstrate that the general characterisations of the three ®nancial systems are empirically valid, that the overall ®nancial systems of Germany and the United Kingdom are consistent con®gurations of the subsystems as complementary elements and that there are indeed considerable and, at least in some cases, persistent differences between the ®nancial systems of the major European countries. Having established this, I can now analyse what consequences this might have for a common monetary policy within EMU.
4 Consequences for EMU: ®nancial systems' structures and common monetary policy Is there a common reaction of the different national systems to monetary policy? After having argued that the credit channel is probably relevant in Europe, too, and that ®nancial systems in Europe differ in fundamental ways, I will now
224 Reinhard H. Schmidt
of NFCs banks to banks b) Liability–intermediation ratios of
a) Liability–intermediation ratios of non-financial Companies (NFCs) a) Liability–intermediation ratios of non-financial companies (NFCs)
80 70 70 60 60 50 50
40 40 30 30
20 20
10 0
Germany
United Kingdom
France
Germany
France
United Kingdom
c)c)Liability–securitisation ratios of of non-financial companies Liability–Securitisation ratios Non-financial companies
d) d) Liability Liability [Asset]–intermediation [Asset]–intermediation ratios ratios of of banks banks
Securitised itisation liabilities NFCs / Total financial liabilities of NFCs Secur Liabilities of of NFCs / Total financial Liabilities of NFCs
FinancialLiabilities liabilities[assets] [assets]of of banks to [from] NBFIs/ Financial banks to [from] NBFIs Totalfinancial financialliabilities liabilities [assets] of NFCs Total [assets] of banks
20
United Kingdom
Liability–intermediation ratios
France
15
France
10 d
Germany United Kingdom
5 0
A-IRs
1981 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
Germany
1981 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
80 70 60 50 40 30 20 10 0
Financial liabilities of NFCs financialLiabilities liabilitiesofofNFCs NFCs Financial Liabilities NFCstotobanks/ banks/ TTotal otal financial
1981 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
1981 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
85 80 75 70
65 60 55 50 45
40
35
Financial Liabilities NFCstotoFS/ FS/ T Total Financial Financial liabilitiesofNFCs financial Liabilities liabilities of of NFCs NFCs
Figure 6.3 Intermediation and securitisation ratios (in percent), 1981±96
offer a theoretically inspired speculation as to what main consequence for a common monetary policy these differences entail. One could be inclined to think that marked differences make it very dif®cult to design and to implement a common monetary policy. Is this conclusion justi®ed or would adopting this position ± which would, of course, also have signi®cant political implications ± mean simply to jump to conclusions? The transmission of monetary policy goes from the central bank (CB) to the (other parts of the) ®nancial sector (FSec) and from there to the real economy (RE), which reacts to monetary developments by changing investment and possibly consumption decisions, which in turn determines aggregate income, employment and the price level. The top line in Figure 6.4 (p. 228) shows the elements of the transmission mechanism. A denotes the relationship between CB and FSec while B denotes the relationship between FSec and RE. Since I wish to focus on different ®nancial systems in a monetary union, differences between countries at the level of CB are of no concern for my mental experiment, as I assume that there is just one single CB. But ®nancial systems may vary in four respects: (1) The way in which the ®nancial sector (FSec) is in¯uenced by central bank policy measures (A) can, and probably does,
Differences Between Financial Systems in European Countries 225
differ; and (2) the FSecs may be, and indeed are, dissimilar; and (3) the ways in which the ®nancial sector changes credit supply to ®rms and households, and in which the real sectors change their demand for ®nancial services, in reactions to the changing conditions in the ®nancial sector (B) seem to be different; and (4) important aspects of RE ± notably company ®nancing, governance and strategies and structures ± are likely to differ. Or to put it more succinctly: ®nancial and real sectors and the sensitivities of the ®nancial sector to central bank policies, and of the real sectors to changing conditions in the ®nancial sector may differ between countries. Comparing Germany and the United Kingdom, one can ask: in which system does a central bank impulse of a given kind and size have a stronger effect on the ®nancial sector (A)? And in which of the two countries does a given amount of in¯uence of the central bank on the ®nancial sector have a stronger effect on the aggregate demand of the real sectors (B) ± and ultimately on income, employment and prices? The answers depend on how one sees these relationships on the structures of the ®nancial systems and in particular on the roles of the banking sector in the respective ®nancial system. I use the concepts underlying the notion of a credit channel. Here the immediate effect of central bank action is assumed to be mainly an effect on the banks. Thus, the strength of the relation A between the central bank and the ®nancial sector is likely to be stronger in a country in which the relative importance of banks in the ®nancial sector is greater. This should be the case in Germany because of the following main reason. The German ®nancial sector consists mainly of universal banks which supply nearly all kinds of ®nancial services under one roof.35 In contrast to the British ®nancial sector banks do not have any serious NBFI-competitors in Germany.36 Thus the central bank can determine more directly the re®nancing conditions of a much larger part of the ®nancial sector. I therefore hypothesise that the ®nancial sector of Germany reacts more strongly to monetary policy impulses than the British ®nancial sector. As a second step, one has to ask how changing conditions in the entire ®nancial system ± and, in particular, in the banking sector ± affect decisions of ®rms and households. To provide a concrete example, let us examine the case of a tightening of monetary policy.37 Would funding opportunities and funding costs for the real sectors be restricted or raised more in Germany or in the United Kingdom? I presume that the restrictive effect of a given in¯uence of the central bank on the ®nancial sector ± and, in particular, on the situation of the banks ± would be less in Germany than in the United Kingdom, and this for three reasons. The ®rst reason is that the relationship between banks on the one side and ®rms and households on the other is closer in Germany than in the United Kingdom. Owing to their closer multifaceted and more long-term-oriented relationships with their customers, German banks might be more hesitant to tighten credit terms for their clients in order not to burden them too much and
226 Reinhard H. Schmidt
not to disturb their longer-term investment and business strategies in which a `housebank'38 has a lively interest. In a system like that of the United Kingdom, in which arm's length banking prevails, banks and other ®nancial intermediaries simply have less reason to refrain from passing on restrictive monetary policy impulses to their customers; they are less committed to their long-term strategies.39 An additional argument is that, compared with a typical British bank, the typical German bank would have more information on its customers, and would thus need to worry less about their net values declining as a consequence of higher interest rates.40 The second reason is that credit terms differ systematically between countries, and that these differences are in line with the general characteristics of the respective ®nancial systems. In Germany, credit contracts have a longer maturity on average, and interest rates are typically less adjustable than in the United Kingdom. Thus, the ®nancial sector would simply ®nd it more dif®cult to pass on rising interest rates. Furthermore, German accounting principles allow ®rms to build up hidden reserves to a much greater extent. As a consequence, the investment decisions of German ®rms should be less sensitive to short-run changes in the costs and availability of external funds. By contrast, UK accounting principles are such that ¯uctuations in the cash ¯ow have a greater in¯uence on investment, which in turn ampli®es the effects of larger ¯uctuations in the cash ¯ow on aggregate demand and income. The third reason why non-®nancial companies in Germany do not react as sensitively when the banks and the rest of the ®nancial sector ®nd themselves in a more restrictive situation is to be found in the prevailing governance and ownership patterns in both the ®nancial and the non-®nancial sectors. The standard models in the macroeconomic literature about the reactions of banks to a restrictive monetary policy of the central bank and about the reactions of ®rms' investment decisions to a tightening of credit terms are almost always based on the assumption that all economic agents are strictly pro®t-oriented and have a rather short-term perspective. This assumption is not equally valid for German banks and corporations on the one side and UK banks and corporations on the other. The bulk of the institutions comprising the German banking system are not privately owned, pro®t-oriented joint stock corporations or limited liability companies. And even those big banks in Germany which are joint stock corporations have a governance system which shields their management almost perfectly from the performance pressure of their shareholders, and thus their behaviour may not be too different from that of a `Sparkasse' or a `Volksbank'. Thus the assumption that, as a general rule, banks will react to central bank impulses in accordance with standard models, may be wrong, at least for certain ®nancial systems. In Germany, for instance, banks have a long-term interest in maintaining stable and `healthy' relations with `healthy' clients, and it seems plausible to assume that they pursue this interest by dampening monetary shocks, irrespective of whether these are intended by
Differences Between Financial Systems in European Countries 227
the central bank or not. At the level of large companies, a strict and exclusive short-term pro®t orientation41 is also the exception and not the rule in Germany. The more moderate pro®t orientation of many large corporations tends to make them `ignore' changing ®nancial conditions so as to promote their long-term strategies, and they probably are more prone to do this than their counterparts in the United Kingdom. Here I would like to reiterate that this behavioural pattern on the part of German banks and the large ®rms is not necessarily a `problem' or a `¯aw'; rather, it is a feature that is consistent with the general logic of the German ®nancial system.42 In the United Kingdom, both banks and corporations are more forced to strive for pro®t and in so doing react to changing price signals. Both in their relationships with clients and in the way they value outstanding securities, privately owned and unambiguously pro®t-oriented banks and other ®nancial intermediaries react more strongly to central bank activity; and private ®rms which have to be extremely conscious of their stock price performance are more strongly motivated ± indeed, forced ± to react more to price signals, as it is to be expected in a more market-oriented system. Thus one can conclude that the two systems also have different reaction functions with respect to the relationship between the ®nancial sector and the non-®nancial sectors (B). The British system reacts more, and the German system reacts less, to changes in the situation of the respective ®nancial sectors. Figure 6.4 summarises the two parts of the argument presented so far. As can be seen, the German system combines a stronger in¯uence of the central bank on the situation of the ®nancial sector (A, with four stars in the second line of Figure 6.4) and a weaker in¯uence of the ®nancial sector on the real economy (B, with two stars). The situation in the United Kingdom is the reverse. The dominance of the central bank over the ®nancial system is weaker (A, with two stars in the third line) and that of the ®nancial sector over the real sector is stronger (B, with four stars). It would, of course, be naive to quantify the strength of these effects solely on the basis of the arguments which I have provided. Accordingly, all that I would like to do here is to present the idea that the relative strengths of the two relationships A and B are inversely related. That this should be so, is not by coincidence. Instead, it follows from the systemic features of the ± broadly de®ned ± ®nancial systems, which are consistent in each of these two countries. If one carried the exercise of `counting stars' to an extreme, one would ®nd that the net effect of monetary policy may be the same in both countries under consideration here. But I do not mean to imply that the two differences between the partial channels A and B in the two countries cancel out completely. I chose to use stars, and not numbers, because the latter might have suggested a greater degree of precision, perhaps leading to misunderstandings. However, it does seem safe to assume that the differences between the stylised ®nancial systems have offsetting effects with respect to A and B, and in
228 Reinhard H. Schmidt
A CB
Credit channel
Interest rate channel
B FSec
RE
Germany United Kingdom
Germany United Kingdom
Figure 6.4 Monetary policy effectiveness in Germany and the United Kingdom
itself this may be an important factor for assessing the potential for a common monetary policy in an enlarged `Euroland'. The two ®nancial systems may differ in a fundamental way, and yet the net effectiveness of monetary policy may not be all that different. If this is news, it is good news for the proponents of a monetary union in Europe and its future enlargement. The argument presented here in a very non-technical manner implicitly draws on the credit channel literature. It could be rendered more precise and more technical by making this basis explicit and spelling out in some detail how the differences between the ®nancial systems determine the strength of the relationships A and B in the two countries. Not unexpectedly, this exercise would demonstrate that the effects of institutional features on the transmission mechanism are complex, dif®cult to aggregate and in some cases ambiguous. Nevertheless, the basic thrust of the argument would not change; the overall effects of the working of the bank lending and the balance sheet channels are in line with my intuitive argument.43 But even a more technical approach would not solve one important aggregation problem: How can one `add' the different effects of the links A and B given that the relevance of the link B depends on how strong the link A is? It seems that this question requires much more work and in-depth econometric studies. I would be delighted if the brief non-technical discussion I have presented here were to be the inspiration for such a research project. Having admitted that my attribution of stars to the partial channels A and B in the two stylised ®nancial systems owes a lot to the credit channel literature, I should brie¯y add a conjecture concerning whether the result is stable with
Differences Between Financial Systems in European Countries 229
respect to the theoretical basis. The general thrust of the credit channel argument is that the possible effects of monetary shocks, including those engineered by a central bank, are probably greater than those which the literature based on the interest rate channel takes into account. If one retains the notion of the ®nancial system as a system of complementary elements, and the premise that Germany and the United Kingdom have consistent but differing ®nancial systems, and combines them with the assumptions of the interest rate channel, then the effects for both partial channels would be weaker: the effect of the central bank on the banking system (A), which could be approximated by the multiplier, would be less pronounced in both countries and the difference between the effect in Germany and the effect in the United Kingdom would still be about the same. The effect of a change in the situation of the banks and other ®nancial institutions on the real economy (B) would probably also be weaker in general ± and at least some of the arguments which would suggest that in the United Kingdom this effect is stronger would also still apply. Thus, the main result ± namely, that the two parts, A and B, of the transmission mechanisms tend to exhibit offsetting differences between the two countries ± would also be valid if one disregarded the speci®c aspects of the credit channel. This is shown by the stars bracketed on the lower lines in Figure 6.4. Are the different reaction functions likely to be stable? I would like to conclude the chapter with some brief remarks about dynamic issues. Financial systems and their sub-systems are not immutable. European integration, regulatory convergence and, last but not least, the common currency have already brought important changes, and are likely to lead to more change in the future. What does this imply for the argument presented in the preceding section? I do not believe that the introduction of the Euro and even the enlargement to include the United Kingdom will change the basic structures of the ®nancial systems of the participating countries in any fundamental way. Money and central banks are not core elements of the concept of ®nancial systems on which the main proposition in this chapter is based. Different countries can have different ®nancial systems in the sense of consistent sets of complementary elements and sub-systems even though they have a common currency and thus necessarily an almost identical monetary policy. This suggests that the introduction of the Euro per se will also not invalidate my proposition that some ®nancial systems in Europe are fundamentally different and that a common monetary policy may nevertheless be feasible. The common currency and especially the enlargement to include the United Kingdom will, however, magnify all of the forces which are currently working to transform the ®nancial systems, and speed up ongoing processes of change. This raises the question of how ®nancial systems
230 Reinhard H. Schmidt
change in general and what this implies for the common monetary policy in `Euroland'. Owing to space constraints, I will have to restrict myself to a few brief remarks. The `fact' that ®nancial systems have the property that every element of a ®nancial system is functionally related to many others and that a consistent ®nancial system constitutes a local optimum is relevant here. It implies that partial reforms are not likely to be sustainable, and this in turn implies that ®nancial systems are not likely to change and to converge gradually. But if there is no gradual convergence, what else could happen? If, for instance, political forces or dynamic entrepreneurs in the ®nancial service industry succeed in introducing elements of one system ± for instance, the active takeover market of the British system ± into the other ± for example, that of Germany ± the immediate result would be an inconsistent `non-system'. If this happens, there will be pressure to restore consistency. A `restauration' of the old system is one way of achieving consistency. In this case the `innovation' is rejected. This is one possible course of events. The other possibility is that the forces which make changes appear desirable are strong and the forces of `restauration' weak and that the formerly consistent system of a given country undergoes not only a series of partial changes, but will in fact experience a `Gestalt switch' from one type to the other. The resulting new con®guration of the elements of the system may be better or worse; one local optimum is replaced by another one.44 In both cases ± that is, in the case of partial reforms which only lead to a `restauration' and in the case of a complete transformation of the system's architecture ± the task of monetary policy makers will be extremely dif®cult because the transmission mechanism will have become unstable. In such a situation the European Central Bank would scarcely be able to predict the overall effects of its policy on the economy of `Euroland' and would therefore ®nd it dif®cult to determine the precise strength of the monetary impulses which it should provide. There would be less of a chance of such a destabilisation and loss of orientation of monetary policy makers occurring in Europe as a consequence of changes in the national ®nancial systems if different national currencies and monetary policies had been retained and not replaced by a uni®ed currency regime; and under the old system of national monetary policies, such a destabilisation would have less serious consequences. I want to conclude by expressing my belief that this increased risk of instability and disorientation constitutes a bigger problem for a common currency than the need to design and implement a common monetary policy for different, but essentially stable, ®nancial systems, on which the existing literature focuses.
Differences Between Financial Systems in European Countries 231
Appendix
Table 6.A1 Financial system indicators: ®nancial sectors Germany (a) Capital markets Number of domestic listed shares (1997) Market capitalisation of shares related to GDP (1996) (%) New listings between 1995 and 1997 Certi®cates of deposit as a percentage of GDP (1997) (b) Banks Market share as percentage of deposits by the non-bank public (1990) Commercial banksa Savings banks Co-operative and rural banksb Five-®rm concentration ratio of commercial banks (1995) Interest margin (1995) Pro®t before tax (average 1990±4 as percentage of balance sheet total) (c) Non-bank ®nancial intermediaries Private pension ®nancing as percentage of GDP (1996) Pension funds portfolio composition (1996)(%) equities bonds property liquid assets
United Kingdom
France
699
2465
683
22 60
136 765
30 109
0.32
10
18.1
39.7 36.1 19.6
56.7 42.7
53.6 3.4 30.1
17 2.66
57 2.21
47 1.66
0.55
0.66
0.06
53
75.6
8 74 7 12
78 14 5 4
Notes:
a For UK authorised banks.
b UK building societies.
Sources: Davis (1998); Wymeersch (1998); European Central Bank (1999).
4.5 14 38 8 40
232 Reinhard H. Schmidt Table 6.A2 Financial system indicators: corporate governance Germany (a) Takeovers (1988±96) (b) CEO compensation structure (1998) Total remuneration (average) in (US $) Variable bonus as a percentage of annual basic compensation Options/long-term incentive plans as a percentage of annual basic compensation (c) Ownership concentration of listed companies (1994±5) Main shareholder > 50% 25±50% < 25% (d) Share ownership structure (1995) (%) Private households/individuals Public sector Insurance companies Pension/Investment-funds Banks Enterprises/commercial corporations Rest of the world
United Kingdom
France
4
1190
155
398 430
645 540
520 389
27
22
19
2
38
30
68 21 11
7 12 81
37 32 31
14.6 4.3 12.4 7.6 10.3 42.1 8.7
20.3 0.8 21.9 36.6 0.4 1.1 16.3
19.4 3.4 1.9 2.0 4.0 58.0 11.2
Sources: Wymeersch (1998); Towers Perrin's 1998 Worldwide Total Remuneration Report.
Differences Between Financial Systems in European Countries 233 Table 6.A3 Financial system indicators: ®nancial patterns Germany (a) Corporations Dividends/cash ¯ow ratio for listed companies (1994) (%) l2.7 Corporate loans collateralised by real estate as percentage of total corporate loan (1993) 36 Short-term corporate credit as percentage of total corporate loan (1993) 22 Long-term corporate credit as percentage of total corporate loan (1993) 78 (b) Households Mortgage credit interest adjustment (%) Fixed 20 Renegotiable 40 Variable 0 Reviewable 40 Typical loan±value ratio 60±80
United Kingdom
l6.67
France
9.46
59
4l
50
27
50
73
0 30 0 70 90±95
80 0 20 0 70±80
Notes:
Fixed: rate ®xed until ®nal maturity.
Renegotiable: rate not ®xed over entire term, but more than one year
Variable: rate adjustable according to index reference rate.
Reviewable: rate adjustable at discretion of lender.
Sources: Maclennan, Muellbauer and Stephen (1998); Borio (1995); La Porta et al. (1999)
234 Reinhard H. Schmidt
Notes 1. The author wishes to thank Falko Fecht, Andreas Hackethal and Adalbert Winkler, and in particular Marcel Tyrell, for extremely valuable advice and research support. Financial support of the Deutsche Forschungsgemeinschaft is gratefully acknowledged. 2. There is a growing body of econometric literature on differences regarding the monetary transmission mechanisms in European countries (see for instance Bank for International Settlements, 1995; Dornbusch, Favero and Giovazzi, 1998; Britton and Whitley 1997; Ramaswamy and Sloek, 1997; Giovanetti and Marimon 1998). To me this literature, which is partly surveyed in Dornbusch, Favero and Giovazzi (1998), does not provide unambiguous evidence of great differences in the transmission mechanisms. However, to the extent that these papers discuss differences between national ®nancial systems at all, they do not go very far in this respect. But see Kashyap and Stein (1997a), who follow a similar approach to the one in the present chapter. 3. See Schmidt and Tyrell (1997) for this terminological distinction. 4. See Schmidt (1990). 5. See for instance the articles by Bernanke and Gertler (1995); Meltzer (1995); Mishkin (1995) and Taylor (1995) in the Journal of Economic Perspectives and symposium on monetary transmission mechanism, and Goodhart (1989); Cecchetti (1995) and Illing (1997, pp. 145 ff.) for overviews. 6. Especially to households' decisions concerning the acquisition of homes and of consumer durables. 7. See Borio (1997) for an exhaustive analysis of different monetary policy procedures and their recent convergence. 8. See Meltzer (1995). 9. See Bernanke and Gertler (1995) for this argument. 10. For surveys of this transmission channel with empirical results for the United States, see also Bernanke, Gertler and Gilchrist (1996) and Kashyap and Stein (1997b). 11. See Bernanke, Gertler and Gilchrist (1996). 12. For an overview see Freixas and Rochet (1997). More recent contributions include Rajan (1996) and Kashyap, Rajan and Stein (1999). The quotation paraphrases the title of an in¯uential article by James (1987). 13. So far, and according to my knowledge, the credit channel has been the subject of È û (1996) and Guender and only a few empirical investigations in Germany. Sto Moersch (1997) come to a negative conclusion concerning the importance of the bank lending channel in Germany. Worms (1997) ®nds some positive evidence (1998a, 1998b) forcefully with respect to the balance sheet channel. Kuppers È û and ®nds strong support for a criticises the results of Guender, Moersch and Sto credit channel in his own empirical study. For the United Kingdom Dale and Haldane (1995) and Ganley and Salmon (1997) show some importance of the credit channel. More recent research on the credit channel in France includes Goux (1996), Candelon and Cudeville (1996) and Payelle (1996). Their results are somewhat ambiguous, but support the assumption that the credit channel is relevant in France too. 14. The following discussion is based on Hackethal and Schmidt (2000). See also Milgrom and Roberts (1995); Hackethal and Tyrell (1998) and Aoki (1999). 15. For this de®nition see also Milgrom and Roberts (1995). The mathematics behind the concept of complementarity are surveyed by Topkis (1998).
Differences Between Financial Systems in European Countries 235 (1990) for this classi®cation with respect to the 16. See Rybczinski (1984) and Berglof ®rst and second sub-system. 17. These dichotomies and the way in which they are related, are discussed in Hackethal and Schmidt (2000). 18. See Franks and Mayer (1994); Schmidt (1997b) and Tirole (1999). 19. See Boot and Macey (1998), Hackethal and Schmidt (2000); Aoki (1999). 20. The link between bank-oriented ®nancing, insider-controlled governance and ®rmspeci®c human capital is more deeply analysed in Hackethal and Tyrell (1998) and Berkovitch and Israel (1998). The correspondence to the business systems is discussed by Milgrom and Roberts (1995); Aoki (1999) and Hackethal and Schmidt (2000). See also Mayer (1998). 21. See Williamson (1988). 22. See also Goodhart (1993). 23. See the empirical results in European Central Bank (1999) and Davis (1998). Some indicators are presented in the Appendix. 24. See Schmidt and Tyrell (1997) and Prigge (1998). 25. For the United Kingdom we do not know enough about (4), but see Hackethal and Schmidt (2000) for ®rst results and some rather speculative conclusions. 26. See Goergen and Renneboog (1998) and Franks, Mayer and Renneboog (1998). 27. See Charkham (1994); Franks and Mayer (1997) and Wymeersch (1998). 28. For a similar characterisation of the British system, see also Prevezer and Ricketts (1994). 29. See Schmidt (1997a) for some details. The recent book by Plihon (1998, p. 79), is among the numerous supporting French references which one could quote here. 30. For a similar conclusion see OECD (1995). 31. For instance, this can be concluded from the more active market for corporate control in recent years (Wymeersch, 1998). On the other hand, ownership concentration and voting power in French public corporations indicates an insidercontrol system (Bloch and Kremp, 1998). 32. See Mayer (1988, 1990) and Corbett and Jenkinson (1996, 1997) for international comparisons of ®nancing patterns and Edwards and Fischer (1994) for a study of Germany and Bertero (1994) for France. 33. In the work of Mayer and his followers, one can ®nd another distinction. It is the distinction between net and gross ®gures, which concerns a different aspect from the one under discussion here. Their 'gross ®gure' are calculated after the aggregation which is identi®ed in Hackethal (1999) and Hackethal and Schmidt (1999) as the cause of the bias. 34. See Schmidt, Hackethal and Tyrell (1999). 35. See for instance the detailed descriptions of the British and German ®nancial sectors in Saunders and Walter (1996). 36. The predominance of the universal banks in Germany may be due to regulatory conditions partly resulting from the monetary policy strategy of the Deutsche Bundesbank, which heavily depended on a stable money demand. See for instance the Deutsche Bundesbank (1995). 37. This is more than a way of making the discussion more concrete. It might well be that the arguments which follow, only apply to the special case of a restrictive monetary policy. 38. This re¯ects that there is some truth to the belief that the old system of 'housebanks' still prevails. This has been vigorously challenged in a well known book by Edwards and Fischer (1994). But note that the empirical basis of their
236 Reinhard H. Schmidt
39.
40. 41.
42.
43. 44.
attack on this presumed myth is an empirical analysis of ®nancing patterns using the methodology of Colin Mayer (1988), which was discussed as yielding unreliable empirical results on p. 222. Studies by Elsas and Krahnen (1998) and Harhoff and (1998) indicate that banks which perceive themselves as the 'housebank' of Korting a given customer ®rm, and which are also perceived as such by the customer ®rm itself, behave systematically differently, that is in a more cooperative and more long-term oriented manner. Thus relationship banking and housebanks still seems to be real factors. Furthermore, a study of the monetary transmission mechanism in Germany with (1998a) demonstrates the segregated bank balance sheet variables by Kuppers relevance of the 'housebank relationship'. His results show that, in contrast to empirical ®ndings for the United States by Kashyap and Stein (1997b), credit terms and loan volumes of smaller German banks, namely the Savings Banks and the Cooperative Banks, react to a monetary tightening to a much smaller extend than the German 'Grossbanken'. In doing so, they isolate their customers from argues that Savings Banks and Cooperative Banks are monetary shocks. Kuppers those credit institutions in Germany which conform most to the model of a housebank with respect to a large number of their customers. That this more ef®cient way of gathering and processing information by banks can also result in more ef®cient investment decisions by ®rms, is theoretically shown in Dewatripont and Maskin (1995) and von Thadden (1995). A genuine pro®t orientation is typically an orientation towards short-term pro®ts because otherwise it would be dif®cult to make this orientation operational. Longterm pro®t orientation would be indistinguishable from an orientation towards the maximisation of total value or growth potential or 'strategic advantage'; see Schmidt and Massmann (1999). This argument re¯ects the effect of intergenerational risk smoothing through the ®nancial system which is discussed on a theoretical level by Allen and Gale (1997). These authors show that the limited ± or merely long-term ± pro®t orientation of the banking sector can lead to welfare gains. I am extremely grateful to Marcel Tyrell and Falko Fecht for having helped me to arrive at this result, which is not elaborated in the present chapter because of space limitations. In Schmidt and Spindler (1999) it is shown that, and why, there is a possibility that in the 'competition' between different (consistent) national systems of corporate governance the one which is less ef®cient under stable conditions may be universally adopted.
References Allen, F. and D. Gale (1997), `Financial Markets, Intermediaries and Intertemporal Smoothing', Journal of Political Economy, 105, 23±46. Aoki, M. (1999), 'An Information Theoretic Approach to Comparative Corporate Governance', Department of Economics, Stanford University, Working Paper 99±04. Bank for International Settlements (1995), Financial Structure and the Monetary Policy Transmission Mechanism, Basle: Bank for International Settlements. E. (1990), `Capital Structure as a Mechanism of Control: A Comparison of Berglof, Financial Systems', in M. Aoki, B. Gustafsson and O. E. Williamson (eds), The Firm as a Nexus of Treaties, London: Sage, 237±62. Berkovitch, E. and R. Israel (1998), 'Optimal Bankruptcy Laws Across Different Economic Systems', University of Michigan, Working Paper.
Differences Between Financial Systems in European Countries 237 Bernanke B. and M. Gertler (1995), `Inside the Black Box: The Credit Channel of Monetary Policy Transmission', Journal of Economic Perspectives, 9, (4) 27±48. Bernanke, B., M. Gertler and S. Gilchrist (1996), `The Financial Accelerator and the Flight to Quality', Review of Economics and Statistics, 78, 1±15. Bertero, E. (1994), `The Banking System, Financial Markets, and Capital Structure: Some New Evidence from France', Oxford Review of Economic Policy, 10, (4),68±78. Bloch, L. and E. Kremp (1998), Ownership and Voting Power, Louvain, European Corporate Governance Network. Boot, A. W. A. and J. R. Macey (1998), 'Objectivity, Control and Adaptability in Corporate Governance', Tinbergen Institute, Discussion Paper, 98±064. Borio C. E. V. (1995), `The Structure of Credit to the Non-Government Sector and the Transmission Mechanism of Monetary Policy: A Cross-Country Comparison', in Bank for International Settlements (ed.), Financial Structure and the Monetary Policy Transmission Mechanism, Basle: BIS, March, 59±105. ÐÐÐÐ (1997), `The Implementation of Monetary Policy in Industrial Countries: A Survey', BIS Economic Paper, 47. Britton, E. and J. Whitley (1997), `Comparing the Monetary Transmission Mechanism in France, Germany and the UK: Some Issues and Results', Bank of England Quarterly Bulletin, May,152±62. et canal de credit: Candelon, B. and E. Cudeville (1996), `Politique monetaire une estimation empirique sur l'economie politique, 104, francËaise', Revue d'economie 781±807. Cecchetti, S. G. (1995), `Distinguishing Theories of the Monetary Transmission Mechanism', Federal Reserve Bank of St. Louis Review, May±June, 83±97. Charkham, J. P. (1994), Keeping Good Company: A Study of Corporate Governance in Five Countries, Oxford, Oxford University Press. Corbett, J. and T. Jenkinson (1996), `The Financing of Industry, 1970±1989: An International Comparison', Journal of the Japanese and International Economies, 10, 71±96. ÐÐÐÐ (1997), `How is Investment Financed? A Study of Germany, Japan, The United Kingdom and the United States', The Manchester School Supplement, 69±93. Dale, S. and A. G. Haldane (1995), `Interest Rates and the Channels of Monetary Transmission: Some Sectoral Estimates', European Economic Review, 39, 1611±1622. Davis E. P. (1998), `Pension Fund Reforms and European Financial Markets', Financial Markets Group, London School of Economics, Special Paper 107. Deutsche Bundesbank (1995), `Verbriefungstendenzen im deutschen Finanzsystem und ihre geldpolitische Bedeutung', Monatsbericht , April, 19±34. Dewatripont, M. and E. Maskin (1995), `Credit and Ef®ciency in Centralized and Decentralized Economies', Review of Economic Studies, 62, 541±55. Dornbusch, R., C. Favero and F. Giavazzi (1998), `Immediate Challenges for the European Central Bank', Economic Policy, 26, 15±64 (including discussion). Edwards, J. and K. Fischer (1994), Banks, Finance and Investment in Germany, Cambridge, Cambridge University Press. Elsas, R. and J. P. Krahnen (1998), `Is Relationship Lending Special? Evidence from Credit-File Data in Germany', Journal of Banking and Finance, 22, Special Issue on Credit Risk Assessment and Relationship Lending, 1283±1316. European Central Bank (1999), Possible Effects of EMU on the EU Banking Systems in the Medium to Long Term, Frankfurt: European Central Bank. Franks, J. and C. Mayer (1994), 'Corporate Control: A Comparison of Insider and Outsider Systems', London Business School, Working Paper.
238 Reinhard H. Schmidt Franks, J. and C. Mayer (1997): Corporate Ownership and Control in the United Kingdom, Germany and France, Journal of Applied Corporate Finance, 9, 30±45. Franks, J., C. Mayer and L. Renneboog (1998), 'Who Disciplines Bad Management', London Business School, Working Paper. Freixas, X. and J. C. Rochet (1997), Microeconomics of Banking , Cambridge, Mass: The MIT Press. Ganley, J. and C. Salmon (1997), 'The Industrial Impact of Monetary Policy Shocks: Some Stylised Facts', Bank of England Working Paper 68. Giovannetti, G. and R. Marimon (1998), 'An EMU with Different Transmission Mechanisms', CEPR Working Paper 2016. Goodhart, C. A. E. (1989), `The Conduct of Monetary Policy', Economic Journal, 99, 293±346. ÐÐÐÐ (1993), `Banks and the Control of Corporations', The Cyprus Journal of Economics, 6 (2), 91±108. Goergen, M. and L. Renneboog (1998), Strong Managers and Passive Institutional Investors in the UK, Louvain, European Corporate Governance Network. Goux, J.-F. (1996), `Le canal etroit du credit en France', Revue d'economie politique, 104, 655±681. Guender, A. and M. Moersch (1997), `On the Existence of a Credit Channel of Monetary Policy in Germany', Kredit und Kapital, 30, 173±185. Hackethal, A. (1999), Unternehmens®nanzierung in Deutschland, USA und Japan, PhD dissertation, University of Frankfurt (fortcoming), Frankfurt, Lang. Hackethal, A. and R. H. Schmidt (1999), 'Financing Patterns: Measurement Concepts and Empirical Results', University of Frankfurt,Working Paper Series: Finance and Accounting, 33. ÐÐÐÐ (2000), Komplementarita t und Finanzsystem, Kredit und Kapital, Beiheft 15 (inprint). Hackethal, A. and M. Tyrell (1998), 'Complementarity and Financial Systems ± A Theoretical Approach', University of Frankfurt, Working Paper Series: Finance and Accounting, 11. Harhoff, D. and T. Korting (1998), `Lending Relationships in Germany ± Empirical Evidence from Survey Data', Journal of Banking and Finance, 22, Special Issue on Credit Risk Assessment and Relationship Lending, 1317±53. Illing, G. (1997), Theorie der Geldpolitik: Eine spieltheoretische Einfuhrung, Berlin: Springer. James, C. (1987), `Some Evidence on the Uniqueness of Bank Loans', Journal of Financial Economics, 19, (2), 217±35. Kashyap, A. K. and J. C. Stein (1997a), `The Role of Banks in Monetary Policy: A Survey with Implications for the European Monetary Union', Economic Perspectives, Federal Reserve Bank of Chicago, 22, (5), 2±18. ÐÐÐÐ (1997b), 'What Do a Million Banks Have to Say About the Transmission of Monetary Policy?', NBER Working Paper, 6056. Kashyap, A. K. , R. Rajan and J. C. Stein (1999), 'Banks as Liquidity Providers: An Explanation for the Co-Existence of Lending and Deposit-Taking', NBER Working Paper, 6962. Kuppers, M. (1998a), 'Curtailing the Black Box: German Banking Groups in the Transmission of Monetary Policy', University of Cologne, Staatswissenschaftliches Seminar Discussion Paper Series, 17. ÐÐÐÐ (1998b), 'On the Existence of a Credit Channel of Monetary Policy in Germany: A Comment', University of Cologne, Staatswissenschaftliches Seminar Discussion Paper Series 19. La Porta, R., F. Lopez-de-Silvanes, A. Shleifer and R. W. Vishny (1999), 'Agency Problems and Dividend Policies Around the World', Harvard University, Working Paper.
Differences Between Financial Systems in European Countries 239 MacLennan, D., J. Muellbauer and M. Stephens (1998), `Asymmetries in Housing and Financial Market Institutions and EMU', Oxford Review of Economic Policy, 14, (3), 54±80. Mayer, C. (1988), `Financial Systems and Corporate Governance: A Review of the International Evidence', Journal of Institutional and Theoretical Economics, 154, 144±165. ÐÐÐÐ (1990), `Financial Systems, Corporate Finance, and Economic Development', in R. G. Hubbard (ed.), Asymmetric Information, Corporate Finance and Investment, Chicago, Chicago University Press, 307±332. ÐÐÐÐ (1998), `New Issues in Corporate Finance', European Economic Review, 32, 1167±88. Meltzer, A. H. (1995), `Monetary, Credit and (Other) Transmission Processes: A Monetarist Perspective', Journal of Economic Perspectives, 9, (4), 50±72. Milgrom, P. and J. Roberts (1995), `Complementarities and Fit ± Strategy, Structure, and Organizational Change in Manufacturing', Journal of Accounting and Economics, 19, 179±208. Mishkin, F. S. (1995), `Symposium on the Monetary Transmission Mechanism', Journal of Economic Perspectives, 9, (4), 3±10. Organisation for Economic Co-operation and Development (OECD) (1995), National Systems for Financing Innovations, Paris, OECD. Payelle, N. (1996), `Conditions de foncionnement du canal du cre dit et gestion des bilans bancaires: un approche empirique dans le cas francËais', Revue d'e conomie politique, 104, 683±703. Plihon, D. (1998), Les Banques ± Nouveaux Enjeux, Nouvelles Strate gies, Paris, La Documentation FrancËaise. Prevezer, M. and M. Ricketts (1994), 'Corporate Governance: The UK Compared with Germany and Japan', in N. Dimsdale and M. Prevezer (eds), Capital Markets and Corporate Governance, Oxford, Clarendon Press, 237±56. Prigge, S. (1998), `A Survey of German Corporate Governance', in K. Hopt et al. (eds), Comparative Corporate Governance ± The State of the Art and Emerging Research , Oxford, Clarendon Press, 943±1044. Rajan, R. (1996), `Why Banks Have a Future: Toward a New Theory of Commercial Banking', Journal of Applied Corporate Finance, 9, 114±35. Ramaswamy, R. and T. Sloek (1997), 'The Real Effects of Monetary Policy in the European Union: What Are the Differences?', IMF Working Paper 97±160. Rybczinski, T. (1984), `Industrial Financial Systems in Europe, US and Japan', Journal of Economic Behaviour and Organization, 5, 275±286. Saunders, A. and I. Walter (1996), Universal Banking: Financial System Design Reconsidered, Chicago, Irwin. Schmidt, R. H. (1990), `Informationsokonomie und Preisentwicklung an Finanzmarkten: Abschied von neoklassischen Optimierungsvorstellungen?', in W. FilcË and C. Kohler (eds), Kooperation, Autonomie und Devisenmarkt , Berlin: Duncker & Humblot, 13±34. ÐÐÐÐ (1997a), `Comparing the French and German Financial System', in H. Kossbiel Finanzsysteme, Frankfurt: Schulz-Kirchner, 9±33. (ed.), Internationale und Europaische ÐÐÐÐ (1997b), `Corporate Governance: The Role of Other Constituencies', in A. Pezard and J.-M. Thiveaud (eds), Corporate Governance: Les Perspectives Internationales, Paris, Montchre tien, 61±74. Schmidt, R. H., A. Hackethal and M. Tyrell, `Disintermediation and the Role of Banks in Europe: An International Comparison', Journal of Financial Intermediation, 8, 36±67. Schmidt, R. H. and J. Massmann, J. (1999), `Drei Miûversta ndnisse zum Thema ``Shareholder Value''', in B. Kumar et al. (eds), Unternehmensethik und die Transformation des Wettbewerbs, Stuttgart: SchaÈffer-Poeschel, 125±7.
240 Reinhard H. Schmidt Schmidt, R. H. and G. Spindler (1999), 'Path Dependence, Corporate Governance and Complementarity ± A Comment on Bebchuk and Roe', University of Frankfurt, Working Paper Series: Finance and Accounting, 27. Schmidt, R. H. and M. Tyrell (1997), `Financial Systems, Corporate Finance and Corporate Governance', European Financial Management, 3, 159±187. E. (1996), 'Die Finanzierungsstruktur der Unternehmen und deren Reaktionen auf Stoû, moneta re Impulse ± Eine Analyse anhand der Unternehmensbilanzstatistik der Deutschen Bundesbank', Deutsche Bundesbank, Discussion Paper 9/96. Taylor, J. B. (1995), `The Monetary Transmission Mechanism: An Empirical Framework', Journal of Economic Perspectives, 9, (4), 11±26. Tirole, J. (1999), 'Corporate Governance', University of Toulouse, Working Paper. Tobin, J. (1969), `A General Equilibrium Approach to Monetary Policy', Journal of Money, Credit, and Banking, 1, 15±29. Topkis, D. (1998), Supermodularity and Complementarity, Princeton, Princeton University Press. Von Thadden, E.-L. (1995), `Long-term Contracts, Short-term Investment, and Monitoring', Review of Economic Studies, 62, 557±575. Williamson, O. E. (1988), `The Logic of Economic Organization', Journal of Law, Economics, and Organization, 4, 65±93. Worms, A. (1997), Bankkredite an Unternehmen und ihre Rolle in der geldpolitischen Transmission in Deutschland, Frankfurt: Lang. Wymeersch, E. (1998), 'A Status Report on Corporate Governance Rules and Practices in Some Continental European States', in K. Hopt et al. (eds), Comparative Corporate Governance ± The State of the Art and Emerging Research, Oxford: Clarendon Press, 1045±1210.
Discussion
Charles A. E. Goodhart
Schmidt's main result, as you will have read, is that ®nancial systems consist of a number of complementary elements, with both Germany and the United Kingdom having consistent but differing ± even polar opposite ± systems. I accept this, and I am strongly predisposed towards accepting the counterattack launched against the Mayer position by Schmidt and his co-author and student, Hackethal ± which is that the appropriate comparison is on gross, not net, ®nancial ¯ows, and that on this basis the differentiation between the German±Japanese and Anglo±Saxon ®nancial systems does show up reasonably clearly, (whereas it does not do so on a net basis). I was also very interested in Schmidt's Figure 6.3 (p. 224) showing that the French ®nancial system of intermediation was in several respects in the process of moving from the German towards the British camp, as they had presumably surmised in the Banque National de Paris (when launching a hostile takeover bid). Indeed I had very little to criticise in sections on the classi®cation of the various channels of monetary effects or on the general characterisations of the ®nancial systems in these three countries. But my role as a discussant is to comment critically and my particular interest is on the effect of monetary policy on the real economies in Germany and the UK. For this purpose I want to focus on Schmidt's Figure 6.4 (p. 228) (see Figure D6.1). Let me start with the credit channel. Here Schmidt claims that the effect on the ®nancial sector is larger in Germany than in the United Kingdom, largely because the banks play a more central role in Germany, and I am prepared to accept that. He also claims that the credit channel has a greater overall effect than the interest rate channel and I again agree that that certainly can be the case, (consider, for example, current Japanese problems), but I do not feel that this is necessarily so. The credit channel, in my view, does not operate in a strictly linear fashion. In some cases ± for example, where bank capital is already very plentiful ± it may be weak, almost non-existent. In other circumstances it may, indeed, be ferocious. So my classi®cation of the credit channel is as shown Figure D6.1b, distinguishing between the weak and the (bracketed) strong cases. 241
242 Charles A. E. Goodhart
My main disagreement with Schmidt relates, however, to the interest rate channel. With the United Kingdom being more market-oriented, and typically with more short-term, impersonal arms-length contracts, I would expect both links in the chain between monetary policy and the real economy to be signi®cantly stronger in the UK than in Germany as shown in Figure D6.1b. So my general assessment is that unless the monetary authorities press hard enough to instigate a severe credit crunch ± and I rather doubt whether there are any clear examples of that in Germany in recent years ± monetary policy measures would otherwise bite quicker and harder in the United Kingdom than in Germany. Schmidt argues that the asymmetries balance out, so that there would be no problem for overall euro monetary management when, and if, the United Kingdom joins. Estimates from various models, VARs, structural VARs, structural models, multicountry models, etc. of the comparative effects of monetary policy on activity and in¯ation in the main European countries have come up with a somewhat bewildering range of answers. Stephen Cecchetti in Chapter 5 in this volume quotes work by Ehrmann, as well as his own prior work, indicating that the effects on activity (and to a less marked extent on in¯ation), are less marked in the Anglo±Saxon and Scandinavian countries than in France and Germany. I recall other estimates that reverse this ranking, and certainly my own priors are that monetary policy should bite harder in the Anglo±Saxon countries. And there are yet other models that ®nd, or assume, no difference. Be that as it may, my own position on this, that pure interest rate policy bites harder in the United Kingdom, leads towards a further problem of consistency with the evidence. If monetary policy does, indeed, bite harder in the United Kingdom, then one would on the face of it expect interest rates to have needed to be much less variable in the United Kingdom. Instead they have been much more variable (see Figure D6.2). But these are nominal interest rates. One answer is that German policy has been better in controlling in¯ation, so that UK nominal rates have had to ¯uctuate more to offset the larger variations in in¯ation; and there is some truth in that. But even so, the variance of real interest rates has also, been somewhat greater in the UK than in Germany (Figure D6.3). Why might this have been so, if in the United Kingdom monetary policy does bite harder? Possible reasons include: (1) Greater shocks in United Kingdom; (2) More nominal, fewer real, rigidities in Germany than in United Kingdom ± i.e. a ¯atter Phillips curve; (3) Mismeasurement of the relevant index for in¯ation; and importance of asset (housing) prices; (4) More policy errors, and abrupt regime changes, in the United Kingdom; Thus the in the last two decades has suffered far more variability in housing prices than Germany.
As proposed by Schmidt CB
FS FSec
RE
Credit channel Germany United Kingdom Interest rate channel Germany United Kingdom
As revised by Goodhart Credit channel Germany
United Kingdom
Interest rate channel Germany United Kingdom
Figure D6.1 Monetary policy effectiveness in Germany and the United Kingdom 243
244
18
16
14
12
10
United Kingdom Kingdom Germany
8
6
4
2
0 1980:1
1982:1
1984:1
1986:1
1988:1
Figure D6.2 Nominal money market rates, 1980:1±1998:1
1990:1
1992:1
1994:1
1996:1
1998:1
10
8
6
4
United Kingdom Kingdom Germany
2
–0
–2
–4
–6 1980:1
1982:1
1984:1
1986:1
1980:1
1990:1
1994:1
1996:1
1998:1
245
Figure D6.3 Real money market rates, 1980:1±1998:1
1988:1
246 Charles A. E. Goodhart Table D6.1 A descriptive statistical analysis of nominal and real money market rates for the United Kingdom and Germany, l980±l998
UK nominal money market rate German nominal money market rate UK real money market rate German real money market rate
Mean
Standard deviation
Coef®cient of variation
9.92
3.33
1.12
6.21
2.43
0.95
4.09
2.37
1.37
3.36
1.24
0.46
But the truth is that I do not have a good, cast-iron, explanation for my quandary ± which is, to repeat, that, although I believe monetary policy bites harder on output in the United Kingdom than in Germany, such policy has nonetheless been more variable in the United Kingdom. Let me conclude with a brief comment on dynamics. Because Schmidt sees the current static position on the effectiveness of monetary policy in the United Kingdom and Germany as balanced, he is more worried about dynamic shifts in structure causing future problems. I do not share that worry. I worry that doubts whether `one size ®ts all' in monetary policy in the Euro Zone could be initially exacerbated by different ®nancial structures, with adverse political repercussions. So I view a growing homogenisation of ®nancial structures across Europe with more and longer ®xed-term mortgages in the United Kingdom and more reliance on securities markets in Germany as an almost unalloyed bene®t.
Discussion
Alain Vienney
After a few theoretical reminders concerning monetary policy transmission channels, Schmidt in Chapter 6 examines the extent to which the structural differences between the German, British and French systems may affect the conduct of the single monetary policy. Schmidt makes the traditional distinction between the German model of the ®nancing of the economy based on a privileged relationship between banks and ®rms ± a model in which the vast majority of the ®nancing and investments of non-®nancial agents are intermediated by the banking system ± and the British market-oriented model ± where the ®nancial markets account for a bigger share of ®nancing and investment. This contrast is still valid today whereas the French ®nancial system has to a great extent already moved from the German to the British model, and the process is probably accelerating. Favouring an analysis in terms of the `credit channel', Schmidt estimates that while the share of intermediated ®nancing is larger in the German than in the British model the elasticity of bank lending to the net situation of borrowers is signi®cantly lower: German banks favour long-term relationships with companies and accordingly `smooth out' their lending. In all, according to Schmidt the German and British models present two structural differences that cancel each other out, with the result that the impact of monetary policy on the real economy does not differ signi®cantly between the two economies. The widening of the Euro Area to include the United Kingdom is therefore not a problem, and that is good news. So Chapter 6 makes an interesting contribution to the debate on monetary policy transmission mechanisms in Europe. In particular I appreciated the presentation of the ®nancial system types which include the inter-relationship between sub-systems in a coherent and stable way. It helps to better understand the monetary transmission mechanism especially when one considers the credit channel. However, it seems to me ± and that is one of the major points I would like to make ± that the analysis shows a somewhat exaggerated bias towards the `credit channel' and as a consequence, at both the theoretical and empirical 247
248 Alain Vienney
levels, underplays the interest rate channel, which is by far the predominant one and explains a number of differences between countries. To judge by the number of stars allocated in Figure 6.4 (p. 228), Schmidt regards the credit channel as even more important than the interest rate channel, which comes as a surprise.
1 The interest rate channel ± which is generally believed to account for most of the differences between the United Kingdom and continental Europe ± is underestimated It seems to me that Schmidt pays insuf®cient attention to certain essential features of the interest rate channel. He does not, for example, mention the differences in the net position of non-®nancial agents (private individuals in Germany and France are net creditors ± that is, their assets exceed their liabilities, while private individuals in the United Kingdom are net debtors1 ), even though they may explain many of the differences between the United Kingdom and continental Europe. Thus, in the United Kingdom income effects and substitution effects have the same impact on consumption (a rise in interest rates depresses consumption in terms of both the substitution and the income effect), whereas in the countries of continental Europe the substitution effect has an opposite impact to the income effect (through the substitution effect an interest rate rise curbs consumption but increases the net ®nancial income of private individuals and thus their disposable income), and so the effect of monetary policy on private consumption is less determinate. Although the author refers to differences in the nature of ®nancial contracts (®xed rates as opposed to variable or adjustable rates2 ), they are not cited as an important source of the differences between the United Kingdom and the countries of continental Europe. It is obviously related to the relative share of short-term credit where the adjustment of rates can take place more rapidly.3 If you take into account that element of adjustability for the whole balance sheet of banks, you end up with a net position in terms of adjustable or nonadjustable rates for assets or liabilities and even if this situation can be dealt with through appropriate off-balance sheet operations it has a bearing on the way in which the changes in interest rates are passed onto the clients. Likewise, the `wealth effects' are not mentioned, although in view of the different degrees of `marketisation' of the economy, these represent another major difference between the Anglo±Saxon countries and the countries of continental Europe. They are clearly evident in the United Kingdom, for example,4 but are indiscernible in continental Europe according to most empirical studies. More generally, the contrast which Schmidt draws between the effect of monetary policy on the ®nancial system and the effect of changes in the environment of the ®nancial system on the real economy seems to be of relatively limited practical relevance. Besides, it is more usual to distinguish
Discussion 249
between the impact of monetary policy on the ®nancial economy (that is the in¯uence of movements in key rates on bank lending rates and on short- and long-term interest rates) and the impact of the ®nancial economy on the real economy.
2 The importance of the credit channel tends to be overestimated, without looking closely to the formation of bank lending rates Although I am not in a position to judge the signi®cance of the credit channel in Germany, the only evidence of its existence in France is indirect.5 As I have already indicated, an analysis of the method whereby bank lending rates are formed (that is their sensitivity to changes in key rates), which does differ between countries,6 would have been useful. I would like in this regard to emphasise two elements:
. The net interest margin may allow some leeway on the bank lending rates and it still differs widely in the Euro Area
. The presence of regulated rates is of primary importance for the setting up of lending rates, in France, one-third of M3 is made up of this kind of assets with administrated rates and it is part of the public debate to argue that these rates prevent an additional lowering of the borrowing terms.
Nonetheless, the analysis of the reasons why German banks, unlike UK banks, `smooth out' their lending (pp. 225±9) is of obvious value.7 It cannot be denied therefore that, if we are considering only the credit channel, the monetary policy transmission channels in the United Kingdom and in Germany would converge on account of two structural differences that cancel each other out.8
3 The UK economy is more responsive to monetary policy decisions than the German or French economies Simulations conducted, under the aegis of the BIS, by the central banks using their national macroeconomic models and collated by Smets9 demonstrate a greater responsiveness to a `standardised' monetary shock in the case of the UK economy than in the case of France or Germany. Thus, a one-point increase in the key rate over two years leads in the very ®rst year to a 0.85 per cent contraction in GDP in the United Kingdom, compared with more limited declines of 0.4 per cent and 0.37 per cent, respectively, in France and Germany.10 While the VAR models11 produce more contrasting results, they also highlight differences in the scale of the adjustment that economies make to a monetary shock, although the speed of adjustment is more or less the same
250 Alain Vienney
from one country to another (in particular, no price effect is evident within a period of under eighteen months±two years). It will be noted, however, that the results for the United Kingdom are very close to those obtained for Italy and that, in this respect also, there is no essential difference between the United Kingdom on the one hand and continental Europe on the other.12
4 The introduction of the Euro will change progressively the basic structures of the ®nancial systems of the participating countries . First, as it is noted in Chapter 6, the overall transmission mechanism is not that different between the United Kingdom and other European countries.
. Second, monetary policy transmission channels in the United Kingdom
and in continental Europe should gradually converge, in line with the prospects of the United Kingdom joining the Euro Area and the development of a large uni®ed capital market in Europe,13 moreover we do have the example of France moving from a German type to an Anglo± Saxon type of ®nancial system: ± In many respects the particular features that characterise the channels of transmission in the United Kingdom re¯ect the high degree of `marketisation' of the economy (widespread holding of ®nancial assets, which explains the wealth effect, and lower rate of intermediation than in the countries of continental Europe). This trend could be followed in continental Europe as a result of a uni®ed ®nancial market. ± In addition, the United Kingdom integration into the Euro Area could encourage convergence in interest rate references, which up to now have been in¯uenced by past in¯ation performance. UK banks might, for example, increasingly lend at long-term ®xed rates, like their counterparts in continental Europe. The latest available statistics would seem to con®rm that this is, in fact, what is happening. ± Competition will be the driving force of the evolution and it will progressively generate more similar ®nancing-supply response, means of payment and asset investments. Finally I dare say that these structural evolutions will probably be conducive to a certain degree of harmonisation of the legal and ®scal frameworks in the member countries.
Notes 1. In 1995, the net position in interest-bearing assets were 48 per cent, for German households, 26 per cent for France and ±9 per cent for the United Kingdom. 2. As a general rule, the United Kingdom prefers variable or adjustable rates (for example, in the case of mortgage lending), whereas most loans and assets in Germany and France are still at ®xed rates.
Discussion 251 3. It should be noted that the proportion is the most important in Italy. 4. An increase of GBP 100 in a private individual's net stock market assets leads to a Great Britain Pound 5 increase in consumption within two years. 5. See for example the article by Barran, Couderc and Mojon (1994). 6. See, for example, Borio and Fritz (1995) and Browne and Fell (1996). The differences are mainly attributable to the existing level of competition among the banks and, in France, to the presence of regulated rates. On this subject, see P®ster Grunspan (1998). 7. Schmidt argues that German banks prefer to take a long-term view of their relationship with their clients and therefore do not shorten their sails at the ®rst sign of trouble. This is particularly true since they are often shareholders in the companies and that, given the long-term nature of the relationship, they are in a better position than UK banks to assess precisely the effect of a rise in interest rates on the ®nancial situation and therefore solvency of their clients. 8. The ®rst structural difference is that, in the case of United Kingdom, credit is less likely to be constrained since ®rms are able to have recourse indifferently either to markets or to banks. 9. Smets (1995). 10. Simulation conducted with an endogenous exchange rate, except for the bilateral FRF/DEM rate, which is assumed to be constant. 11. See, for example Gerlach and Smets (1995) or Dornbush, Favero, Giavazzi (1998). 12. The fact that the Italian economy responds so sharply to monetary shocks is largely due to the importance of short-term ®nance and investment, since the elasticity of short-term interest rates to key rates is more marked than that of long-term rates. 13. Although Schmidt does not accept this argument (`I do not believe that the introduction of the Euro and even the enlargement to include the United Kingdom will change the basic structures of the ®nancial systems of the participating countries in any fundamental way' Df. 334, p. 000), he does not explain why.
References Barrah F., V. Coudert and B. Mojon (1994) `Transmission de la politique mone taire et credit bancaire, une application a cinq pays del', OCDE CEPII Morning Paper, 94±03, June. Borio, C. E. V. and L. S. Fitz (1995) `The Response of Short-term Lending Rates to Policy Rates: A Cross-country Perspective', Basle, BIS, May. Browne, F. and L. Fell (1996) `The Changing Institutional Context of Monetary Policy and its Implication for In¯ation', Frankfurt, EMI, May. Dornbusch, R., C. Favero and F. Giavazzi (1998) `A Red Letter day', CEPR Discussion Paper, 1804. Gerlach, S. and F. Smets (1995) `The Monetary Transmission Mechanism: Evidence from the G7 Countries', Basle, BIS. P®ster, C. and T. Grunspan (1998) `Bank Restructuring, Monetary Policy Ef®ciency and Financial Stability in France', Central Banks' Economists' Metings, Basle, 28±29 October. Smets, F. (1995) `Central Bank Macroeconomic Models and the Monetary Policy Transmission Mechanism', Basle, BIS.
7
European Labour Markets and the Euro: How Much Flexibility Do We Really Need? Michael C. Burda
1
Introduction
In addition to evidence on the nature and source of regional ¯uctuations, European Monetary Union (EMU) will also provide economists with valuable new evidence on the monetary transmission mechanism. Given the scepticism with which macroeconomics currently regards monetary policy, current concern over real effects of EMU comes as a surprise; in a world of ¯exible prices, space-spanning contingent claims markets and complete information, it is dif®cult to see why monetary union matters at all for real integration processes already underway.1 For example, if the real business cycle paradigm (RBC) ± which emphasises disturbances and propagation mechanisms in the non-monetary economy and ignores nominal rigidities ± is approximately correct, the EMU exercise is nothing but a sophisticated veil. To the extent that EMU leaves ®scal policies and real behavioural incentives unchanged, the effects of a common currency are of second order at best. In short, this chapter has no real reason to be written. Yet, the liveliness of the contemporary debate ± among reasonable and coolheaded economists for the most part ± is suggestive of an expectation that, for whatever reasons, real effects of EMU are in the cards. If this is indeed the case, the underlying presumption must be that nominal disturbances to aggregate demand and the money supply, in particular, can in¯uence the short-run path of output and employment, and will continue to do so after EMU is up and running. Not wanting to make my life too easy, I have decided to write this chapter from the perspective of an eclectic who is willing to entertain newKeynesian arguments. These arguments are important, as the survival of monetary union will rest on factors outlined long ago by Mundell (1961) and McKinnon (1963). In Europe, these are perceived to originate chie¯y in labour markets. From a point of departure that money and monetary policy can in¯uence real variables, I will discuss the macroeconomic impact of labour market rigidities on real and nominal adjustment to disturbances in `Euroland'. However, the most interesting aspects involve taking the 252
European Labour Markets and the Euro 253
discussion one step further: for a number of reasons, the arrival of EMU will itself have signi®cant effects on the functioning of labour and product markets and the relative importance of real and nominal rigidities. These feedbacks will ultimately affect the way Europe reacts as a macroeconomic entity to demand disturbances and how its central bank views the effectiveness of monetary policy. This chapter surveys a number of issues too involved to be treated in modeltheoretic detail here. I will furthermore abstain from econometric analyses for reasons which should be clear to all. There is a sense that the macroeconomic regime has changed in a way it has not in several hundred years in Europe: if the Lucas Critique has any relevance at all, it had better be here and now. I will adduce some empirical evidence however, which is suggestive of what one might expect in the future. The chapter is highly speculative, but meant to be so. My discussion is organised as follows. In Section 2, I discuss the macroeconomic impact ± at both regional and pan-European level ± of the current structure of labour markets. Section 2 surveys the multifarious means by which a monetary union could affect the functioning of labour markets. This feedback takes some surprising turns, and may lead to a wholly different perception of the transmission channels of monetary policy in Europe. Section 3 adduces simple but striking evidence in support of my hypotheses and Section 4 concludes.
2 How will labour market in¯exibility affect the macroeconomics of `Euroland'? Real rigidities and regional ¯uctuations Robert Mundell taught us long ago that the key to a monetary union's success can be found in the synchronisation of underlying economic fortunes and, barring this, the mobility of factors of production, especially that of labour.2 Naturally, labour mobility is costly for both natural and man-made reasons, and immobility may be regarded differently across cultures and traditions. Abstracting from social valuations of immobility, losses of output and welfare are involved when labour does not move to job opportunities, in a geographic, industrial or occupational sense. To the extent that regional shocks ± such as an oil discovery in the North Sea or German uni®cation ± continue to occur, they will wreak macroeconomic havoc on the real evolution of output, employment and other important variables in ways which are now well understood. The lack of a ¯exible nominal exchange rate in a world of nominal rigidities may imply protracted adjustment to regional shocks, unless labour and other resources move to follow better economic fortunes. Indeed, the available evidence on labour mobility in the European context is remarkably discouraging and suggests that a major component of rigidity
254 Michael C. Burda
derives from labour's unwillingness to move.3 In addition, Europe is characterised by less in-migration, lower fertility and older demographic structure; all these factors further tend to increase immobility. It would almost seem unfair to compare Europe with the United States, given that the gene pool of the latter constitutes a selection of those of the former who had the strongest incentives to migrate! At the same time, it is worth noting that even within national boundaries, European labour mobility is low and not capable of erasing regional disparities, so it is unrealistic to expect much here.4 Yet factor mobility in a monetary union is not restricted to labour, and under conditions of constant returns one should be indifferent whether the capital migrates to labour or labour migrates to capital.5 In theory, EMU will liberate capital mobility as exchange rate risk vanishes, and in fact intraEuropean capital mobility has surged in recent years. This is documented in Table 7.1, which shows the evolution of intra-EU foreign direct investment (FDI) ¯ows since the 1980. The persistent boom in European equities can be seen in part as a reaction to the increased mobility now afforded to capital by a common currency and increasingly integrated asset markets, combined with ef®ciencies offered by a uni®ed market for goods and services. Whether mobile capital can smooth out ¯uctuations is not well understood; it stands to reason, however, that capital should move to places where labour is in excess supply and could in principle perform this function.
Table 7.1 Intra±EU FDI Flows, 1985±94 (per cent of GDP) Country
Direct investment in¯ows from EU countries
Ireland Portugal Spain Sweden Denmark Netherlands Belgium/Luxembourg United Kingdom Austria Italy Greece Finland Germany France
Balance of direct investment to other EU countries
1985±9
1990±4
1985±9
(0.32)a 1.01 1.02 0.26 0.39 0.91 1.64 0.84 0.24 0.24 0.21 0.23 0.17 0.42
(0.13) 1.72 1.54 1.11 1.05 1.29 3.05 0.69 0.35 0.19 0.53 0.47 0.11 0.67
n.a. 0.96 0.81 ±1.25 ±0.27 ±0.26 0.36 ±0.01 0.07 ±0.03 n.a. ±0.73 ±0.28 ±0.19
Note:
a Numbers in parentheses are described as highly unreliable.
Source: Dohse and Krieger±Boden (1998).
1990±4 n.a. 1.38 1.24 ±0.69 ±0.05 ±1.34 0.73 ±0.17 ±0.08 ±0.17 n.a. ±0.75 ±0.62 ±0.26
European Labour Markets and the Euro 255
Product market integration is potentially more important than either form of factor mobility. Heckscher±Ohlin trade theory under incomplete specialisation implies that harmonised product prices in traded output produced with the same technology leads to wage convergence (the factor price equalisation (FPE) theorem). Consequently the need for factor mobility is eliminated and the market spreads shocks automatically across the currency area. Here evidence by von Hagen and Neumann (1994); Fatas (1997); Frankel and Rose (1996); Bayoumi and Eichengreen (1993, 1996) and others seems to point to increasing product market integration over time, although this literature has tended to emphasise quantities more than prices. Nominal frictions, real rigidities and pan-European macroeconomic ¯uctuations The next point of discussion is the role of nominal frictions in the European context. What could the sources of non-neutralities of money in a future EMU be? Arguing from the status quo, the common perception is that nominal rigidities play a subordinate role in European business cycles. The standard assumption is that the large role of centralised collective bargaining, the use of indexation and a high degree of openness all made Europe more likely to translate demand disturbances rapidly into price level changes than the United States, Canada, or Japan. A thorough if somewhat dated discussion of these issues can be found in the work of Michael Bruno and Jeffrey Sachs,6 who distinguished between US and continental European labour markets by their reaction to nominal demand and supply shocks. For them, the structure of labour markets ± meaning to a large extent institutions of wage determination ± was a key determinant of adjustment to macroeconomic and especially supply-side disturbances. As this chapter's title suggests, the functioning of the labour market will be central to understanding the effects of EMU.7 Mainstream macroeconomics predicts real effects of money and nominal demand ¯uctuations when impediments prevent the clearing of product and especially labour markets. While the origin of these impediments is still poorly understood, it is also clear that the role of rigidities in nominal and real spheres are highly complementary for any neoclassical or `new-Keynesian' account of macroeconomic ¯uctuations (e.g. Ball and Romer, 1990; Blanchard, 1990; Romer, 1996; 1998; Kollmann, 1999). This means that it is not Jeanne, 1998; Roger, suf®cient for nominal rigidities (such as menu costs) to exist, but they must also exist alongside real rigidities. In one widely cited mechanism, coordination failures prevent agents from moving the economy to a better equilibrium. This complementarity lends intuition to Milton Friedman's (1953) argument for ¯oating exchange rates. In a famous analogy, Friedman compared the gains from ¯exible rates to those from setting all clocks back one hour in the fall and forward in the spring: it is more ef®cient to change the nominal time standard (the nominal exchange rate) than it is to require
256 Michael C. Burda
Figure 7.1 Complementarity of real and nominal rigidities for a given price change
millions of individuals to adjust their daily time schedules (nominal domestic prices) to the annual solar cycle (changing demand and supply conditions).8 Blanchard (1990. pp. 810ff.) and especially Romer (1996. p. 283) make the reasoning more explicit: individuals do not change their nominal schedules in the absence of daylight savings time because of the real costs they incur, given that all others do not change their behaviour. We are dealing with ®rms which set prices. The extent of real rigidities for a given price change can be thought of as the resource cost to ®rms of not moving to optimal pricing in the absence of nominal frictions. In the two panels of Figure 7.1, this is given by the shaded areas, which are approximately triangles with base Q � Q* (equal to the output difference between passive quantity adjustment at rigid price p given by Q and the pro®t maximizing output level given by Q*), and height equal to the gap between marginal cost (MC) and marginal revenue (MR) at output level Q. The latter depends on various factors such as the behaviour of the marginal product of labour, marginal capacity costs, and the elasticity of labour supply. In the panel (a), the costs of not changing price from p to p* are relatively small, since the desired quantity change is modest and marginal costs are ¯at. In contrast, the ®rm depicted in panel (b) of Figure 7.1 is under considerable cost pressure to change prices, as can be seen by the vertical difference between marginal revenue and price for the last units produced. Passive quantity adjustment implies a large departure from unconstrained optimal production Q*, while sharply rising marginal costs means that these additional units are being produced at a large loss.9 For a given costly nominal price adjustment, the ®rm in panel (a) is likely to maintain rigid nominal pricing, while the ®rm in panel (b) will adjust its prices. Comparing the two panels, one sees the necessity of real rigidities: individual ®rms have little incentive to change
European Labour Markets and the Euro 257
prices, given that others are not doing so. Strategic complementarity implies that second-order issues for the ®rm can have ®rst-order effects for the macroeconomy. Money wage rigidity could also induce business cycle ¯uctuations. While an important element in the early intellectual development of Keynesian macroeconomics, nominal wage rigidity is not borne out at the micro level (Bils, 1985, Smith, 1999) nor is it particularly supported by aggregate evidence on wage and price dynamics (see references in Blanchard, 1990); Jeanne (1998) (1998) have both shown that nominal price rigidity, combined with and Roger some degree of real wage rigidity, is suf®cient to generate persistent ¯uctuations that resemble US business cycles.10 Summary The previous discussion can lead to rather sombre conclusions about the future of EMU. First, the conventional wisdom of extreme rigidity in labour markets, which now has the OECD seal of approval (OECD, 1994) and is accepted nowadays by everyone except the labour unions and perhaps a few surviving extremists in the German ®nance ministry, should render the EMU a Mundellian nightmare. It won't be necessary, according to this logic, for another German reuni®cation to occur to generate real problems. All we need is some overheating in Ireland, Portugal, or Finland and the whole EMU project will collapse as the other regions slump without any equilibrium mechanism. An equally pessimistic message emerges on the monetary transmission mechanism when considered under these circumstances, in which a rapid pass-through into in¯ation is taken for granted by market participants. Reviews by Buti and Sapir (1998) and Dohse and Krieger-Boden (1998) give rather sombre pictures of the prospects, and Dornbusch, Favero and Giavazzi (1998) raise questions about the asymmetric impact of monetary policy on the participating EMU countries. Moreover, ®scal policy is hamstrung by the Maastricht Treaty and the Pact for Stability and Employment and potential exists for `beggar-thy-neighbour' effects as countries jockey to better their macroeconomic circumstances. This `Flassbeck±Lafontaine Hypothesis' sees purposeful competitive de¯ation just around the corner, as countries unable to devalue are forced to regain competitiveness by more painful means. In this view, governments, robbed of their power to generate instant nominal devaluations will do what Britain did in the ®rst half of the 1920s. Feldstein's (1992) criticism is now widely accepted that politics have outweighed economics; Eichengreen (1998) has already speculated about the `dissolution' of the European Monetary Union before it even begins. Given this doomsday scenario, critical economists are compelled to ask the question: Are rigidities in Europe set in stone? Is it reasonable to assume that the Euro will leave labour markets and their institutions intact ± and, if not, which ones are implicated? What will be the consequences of these changes?
258 Michael C. Burda
What follows is a highly speculative discussion of three areas: (1) nominal rigidities; (2) real rigidities, holding institutions constant; and (3) changing institutions.
3
Will the Euro affect labour market ¯exibility?
Nominal price rigidity should increase First, I speculate that a number of factors will cause nominal rigidities to increase in Euroland, especially that of nominal prices. First, the introduction of a common currency will effectively convert a Europe of many small open economies into a behemoth with an import ± export exposure of 10 per cent of GDP, roughly as closed as the United States and Japan. This is a regime change of striking character. As a consequence, a large share of industry will be moved into the `home goods' sector, and will no longer be exposed to vagaries of nominal exchange rate and international demand ¯uctuations. For small, open economies with output more likely to be concentrated in the valueadded chain, exchange rate disturbances are re¯ected rapidly in both input and output prices; a monetary union in Euroland removes this aspect, as inputs become increasingly non-traded goods invoiced in Euros. Devaluationinduced expenditure-switching is no longer possible on a grand scale.11 Factors favouring nominal rigidities ± i.e. customer relationships, search costs, etc. ± should become relatively more important than costs associated with cross-border transactions.12 Cost pressures will increasingly be restricted to domestic (Euroland) labour markets, marginalising the importance of exchange rate changes for pricing decisions.13 Figure 7.2 illustrates how the reaction of local currency costs to a devaluation are decisive in determining incentives to adjust prices. In panel (a), which corresponds to a small open economy, marginal costs rise in response to a devaluation and the incentive to change prices rise commensurately. In panel (b) ± which corresponds to Euroland ± the incentive is less strong, leading to a larger output effect. A second effect is more subtle (and possibly less relevant). A common currency area is generally assumed to increase competition, as improved price transparency opens up national markets to intra-EMU, cross-border rivals. At the same time, however, monetary union in Europe necessarily implies a signi®cant decrease in the overall relevance of the external market for the representative producer. Assuming that foreign trade is perfectly competitive and priced off the exchange rate according to the law of one price, the representative exporting ®rm pre-EMU, ironically, may face an enlarged domestic market with more pricing power on balance, to the extent that the market using Euros increases relative to that using foreign currencies. This is especially true if the pace of mergers and acquisitions within Euroland continues. To the extent that `inwardisation' increases monopolistic power in price-setting, it will increase incentives not to adjust prices in the domestic
European Labour Markets and the Euro 259
currency, for reasons stressed by Akerlof and Yellen (1985), Mankiw (1985) and Romer (1996). Increased exposure to the sheltered domestic market will mean greater incentives to price to market and to set nominal prices in advance for longer periods, as customer relations become more important and the net bene®ts of charging stable nominal prices increase (Okun, 1981). The third and potentially most important effect ¯ows from the credibility that comes from having a central bank which can `stand above' (i.e. ignore) economic conditions in individual countries and be free of political pressure. To this extent if the European Central Bank (ECB) is really the most independent central bank in the world, agents will be more prone to expect low in¯ation and will not attribute deviations to policy changes. This important source of inertia should be distinguished from the usual wage± price mechanism (e.g. Blanchard, 1990); rather it has to do with the anchoring of in¯ationary expectations and the effect this will have on the willingness to negotiate contracts in nominal terms. To give some sense on the evolution of rigidities, I present some simple statistics for data on comparable price and wage time series from EU member countries.14 Table 7.2 displays average unweighted correlations of bilateral in¯ation rates (®rst difference in the logarithms) for a number of groupings of countries in addition to the Euro-11 since 1961. For comparison, I present data for eight regions of the United States for a similar time period. Clearly, an increase in price convergence has taken place across the board, not only in the smaller `core' groupings. The eigenvalues of the moment matrix indicates the extent to which in¯ation in one country can be expressed as a linear combination in others. Table 7.3 documents that, to a large extent, my
Figure 7.2 The cost of passive quantity adjustment in response to an exchange rate depreciation
260 Michael C. Burda Table 7.2 Synchronisation of price in¯ationa in Europe and the United States, 1961±96 Average correlation coef®cient in group (std. dev.) Total 1961±79 1980±96 sample
Smallest and largest moment matrix eigenvalues (1961±79) and (1980±96) 1961±79
1980±96
Percentage change
Core Europe (B, NL, D, A)
0.76 (0.08)
0.80 (0.06)
0.82 (0.09)
9.82 10�4 0.207
5.89 10�4 0.0928
±39.9 ±55.2
Core Europe F, DK, IT
0.74 (0.11)
0.71 (0.13)
0.81 (0.12)
8.74 10�4 0.560
1.64 10�4 0.363
±81.2 ±35.1
Euro-11 lightb
0.73 (0.14)
0.73 (0.14)
0.80 (0.15)
6.53 10�4 0.983
3.88 10�5 0.602
±94.1 ±38.8
Euro-11 lightb 0.71 DK, S, UK (0.13)
0.70 (0.15)
0.78 (0.14)
3.17 10�4 1.336
3.69 10�5 0.811
±88.4 ±39.3
1.48 10�5 0.376
_
_
Memo: USA 8 regions, 0.95 1978±92, (0.03) GSP de¯ator
_
_
Notes: a In¯ation is measured as ®rst difference in the logarithm of the relevant price index.
b Less Luxembourg, Portugal.
B Belgium; NL Netherlands; D Germany; A Austria; F France; DN Denmark; IT Italy;
S Spain; UK United Kingdom; USA United States.
Sources: US Bureau of Economic Analysis (REIS), IMF, International Monetary Statistics.
conclusions hold when looking at a much smaller time interval and when correcting for exchange rate changes. It has been argued, by Calmfors (1998a) and others, that monetary union could result in increasing nominal money wage rigidity. Presumably this would arise as a result of the low level of in¯ation and resistance to nominal wage reductions. In addition, the alignment of traded goods prices should impose factor price convergence, as long as complete specialisation does not occur ®rst, although this can only be a statement about labour of a given quality. At the same time, Calmfors (1998a) claims that increasingly variable macroeconomic conditions might lead to shorter nominal contract periods and greater nominal wage ¯exibility. Nominal wage behaviour in Europe over the past thirty years lends support to my contention that nominal wages are less likely to be rigid than prices. Table 7.4 and 7.5 clearly show a determination in the strong positive correlation of real wage growth present in the 1960s and 1970s. To the extent
European Labour Markets and the Euro 261 Table 7.3 In¯ation correlations, in national currency and DM terms, 1976±96 Average correlation coef®cient in group Annual OECD in¯ation rate Total 1976±86 Sample
1987±96
Annual OECD in¯ation rate in DMa terms using BLS exchange rates Total Sample
1976±86
1987±96
Core Europe (B, NL, L, D, A)
0.82 (0.10)
0.81 (0.08)
0.77 (0.12)
0.56 (0.27)
0.52 (0.31)
0.70 (0.17)
Core Europe F, DK, IT
0.80 (0.11)
0.79 (0.12)
0.33 (0.50)
0.45 (0.23)
0.45 (0.25)
0.38 (0.41)
Euro-11
0.79 (0.13)
0.67 (0.26)
0.48 (0.37)
0.44 (0.21)
0.48 (0.24)
0.37 (0.40)
Euro-11 DK, S, UK
0.78 (0.12)
0.67 (0.24)
0.45 (0.41)
0.49 (0.21)
0.54 (0.22)
0.38 (0.42)
Notes: a OECD in¯ation corrected using BLS exchange rates.
For country codes, see Table 7.1. Standard deviations in parentheses.
Source: OECD.
that increasing `entropy' in the behaviour of nominal wage movements is re¯ected by decreasing cross-country correlation, this supports the assertion that nominal wage ¯exibility is increasing, not decreasing over time. The largest eigenvalue of the moment matrix for ®rst differences in nominal wages, compared with that of nominal prices, is larger and the decline in the eigenvalues are smaller, suggesting that nominal wages in this context do not seem to merit the description `rigid'. Not only are nominal wages less correlated across European countries than US regions but their levels have exhibited divergence in the past decade. Table 7.6 displays US BLS data on hourly compensation in the EU and computes coef®cients of variation for the groupings CORE (Belgium, Luxembourg, France, Germany and Austria); CORE Denmark France Italy; The EURO-11 (CORE plus Ireland, Finland, Spain, Portugal, France and Italy). For each grouping Germany was retained and dropped to examine the in¯uence of that country, especially in light of German uni®cation. In all cases except the CORE less Germany (the Benelux countries plus Austria), the cross-country variability of nominal wages increased over the ten-year period.
262 Michael C. Burda Table 7.4 Synchronisation of nominal wage growtha in Europe and United States, 1961±96 Average correlation coef®cient in group (std. dev.) Total 1961±79 1980±96 sample
Smallest and largest moment matrix eigenvalues (1961±79) and (1980±96) 1961±79
1980±96
Percentage change
Core Europe (B, NL, D, A)
0.85 (0.06)
0.76 (0.17)
0.46 (0.11)
2.15 10�3 0.688
1.24 10�3 0.130
±42.3 -81.0
Core Europe F, DK, IT
0.72 (0.15)
0.52 (0.32)
0.48 (0.18)
1.49 10�3 1.49
6.59 10�4 0.451
±55.8 ±69.8
Euro-11 lightb
0.71 (0.15)
0.46 (0.35)
0.55 (0.22)
5.58 10�4 2.39
2.57 10�4 0.760
±54.0 ±68.2
Euro-11 lightb 0.66 DK, S, UK (0.18)
0.48 (0.31)
0.50 (0.26)
1.80 10�4 2.96
6.07 10�5 0.981
±66.2 ±66.9 _
Memo: USA 8 regions, 0.92 1978±92, (0.06) annual comp. Memo: USA 8 regions, 1978±92, wages/salaries
0.90 (0.08)
_
_
2.01 10�5 0.449
_
_
_
1.65 10�5 _ 0.425
_
Notes: a Nominal wage growth is measured as ®rst difference in the logarithm of the wage index.
b less Luxembourg, Portugal.
For country codes, see Table 7.1.
Sources: US: Bureau of Economic Analysis (REIS), IMF, International Monetary Statistics.
Real rigidities should decrease given current institutions It is interesting that there are so many who believe that real rigidities in Europe threaten the success of monetary union, and I am sure that my invitation to contribute to this volume was related to my perceived views on real rigidities in European labour markets. Indeed, a number of arguments can be found to buttress the claim that in¯exibility in the labour market will spell the death of EMU. Yet how robust are these arguments to the Lucas Critique ± i.e. the introduction of the Euro? In my view, the more important and subtle effect of EMU has largely escaped scrutiny: How will a common currency affect the functioning of labour markets? Could the vaunted lack of labour market ¯exibility in continental Europe be affected by the introduction of a common currency? If so, how?
European Labour Markets and the Euro 263 Table 7.5 Nominal manufacturing wage growth correlations in national currency and DM termsa Average correlation coef®cient in group Annual nominal wage growth in Annual nominal wage growth in DM manufacturing in local currency basis Total 1976±86 sample
1987±96
Total sample
1976±86
1987±96
Core Europe (B, NL, L, D, A)
0.68 (0.11)
0.64 (0.12)
0.42 (0.34)
0.44 (0.27)
0.44 (0.26)
0.39 (0.33)
Core Europe F, DK, IT
0.66 (0.12)
0.59 (0.22)
0.22 (0.33)
0.29 (0.25)
0.25 (0.31)
0.24 (0.25)
Euro-11
0.68 (0.13)
0.56 (0.19)
0.34 (0.33)
0.30 (0.27)
0.28 (0.32)
0.28 (0.39)
Euro-11 DK, S, UK
0.65 (0.13)
0.55 (0.19)
0.30 (0.35)
0.32 (0.27)
0.33 (0.30)
0.26 (0.40)
Notes: a First differences in log hourly nominal compensation costs for production workers in
manufacturing, in local currency or in DM converted using annual average exchange rates.
For country codes, see Table 7.1. Standard deviations in parentheses.
Sources: US Bureau of Labor Statistics; author's calculations.
Table 7.6 Nominal wages in manufacturing in the EU, 1986 and 1996 Money wages in Europe in dollars (nominal hourly compensation) Land Luxembourg Belgium Germany Netherlands Austria France Denmark Italy Finland Ireland Portugal Spain Sweden United Kingdom memo: USA
Unweighted coef®cients of variation of nominal wages
1986
1996
Grouping
1986
1996
10.86 12.43 13.43 12.22 10.73 10.28 11.07 10.47 10.71 8.02 2.08 6.25 12.43 7.66 13.26
22.55 25.89 31.87 23.14 24.95 21.19 24.24 17.48 24.95 13.85 5.58 13.40 24.56 14.13 17.70
CORE (A, B, D, L, NL)
0.095
0.143
CORE less D
0.077
0.064
CORE DK, I, F
0.098
0.173
CORE less D
0.076
0.123
EURO-11
0.331
0.358
EURO-11 less D
0.336
0.342
Note: For country codes, see Table 7.1; L Luxembourg, I Ireland.
Sources: US Bureau of Labor Statistics, Of®ce of Technology and Productivity.
264 Michael C. Burda
Because the quanti®cation of real rigidities is dif®cult and undoubtedly subject to regime changes (Calmfors 1998a) it seemed unwise to estimate measures of nominal and real wage rigidity, on the other hand it is reasonable to conclude that for the most part the two pressure points on which all real rigidities rest are (1) collective bargaining and unions and (2) the social safety net and especially unemployment bene®ts. My discussion below will concentrate primarily on these. The elasticity of labour demand will increase The ®rst Euro assault on real rigidities is the weakening of union power in wage determination. While unions are already in retreat in much of the OECD (OECD, 1994), in Europe this decline is largely restricted to Britain, membership losses in France and Italy belie an ever-strong in¯uence on central wagesetting institutions, in Germany, membership has declined primarily in the East, where it was arti®cially high to begin with. Yet the brave new world of Euroland portends ill for continental collective bargaining, which has always been a national institution with national idiosyncracies. A simple textbook argument ± namely, the Marshall±Hicks rule of labour demand ± predicts that the melding of European nations into a currency union will attenuate unions' ability to monopolise the supply of labour by increasing the demand elasticity they face.15 Three of the four elements of the Marshall±Hicks rule will be operative. First, labour unions derive their attractiveness from their ability to tap into quasirents that their employers can earn in the market. In a globalising Europe, product market competition among companies operating with quasi-rents will increase dramatically, which translates into an increase in the elasticity of product demand and the elasticity of the derived demand for labour.16 Second, the acceleration of intra-European corporate mergers and takeovers opens up the possibility of easy substitution of capital and cheaper labour for more expensive labour within the Euroland area. This attenuates the bargaining strength of national unions. Third, for any given national labour market, the rest of Euroland is large (and possibly getting larger), meaning that the supply elasticities of competing factors is likely to be high. How will European labour unions cope with these powerful winds of change? Already hamstrung by fragmentation along industrial, regional, or religious lines, they will face language and national cultures as further barriers to their effectiveness. Despite considerable rhetoric, searches of labour union literature (including the Internet) have yielded little concrete evidence of an effective Pan-European labour movement. While a similar argument applies to employer associations, the growing transnationality of capital puts labour at a clear bargaining disadvantage ± a forced decentralisation in the Calmfors± Drif®ll (1988) sense. The potential for coordinated bargaining strategies is presently incompatible with union structures across countries, which is essential to purposeful pattern bargaining. These structures represent decades
European Labour Markets and the Euro 265
of gradual evolution and cannot be changed overnight. To me, at least, it seems highly implausible that Europeans will accept wage leadership of German engineering and public sector workers after having ®nally shaken themselves from the yoke of Teutonic monetary policy!
Strategic interaction of unions with the central bank will change The argument that labour market rigidities might be endogenous has been made by a number of analysts (Danthine and Hunt, 1994, Berthold and Fehn, 1997, Dohse and Krieger-Boden, 1998 among others). While I take the position that competition will impose decentralisation and deregulation of EMU labour markets, a number of analyses emphasise changing strategic interactions between central banks, unions and governments and the effect these can have on aggregate outcomes. In particular the incentives for unions to recognise the effects of their wage demands on the macroeconomy stands at the centre of this discussion.17 An important strand of the literature which has emerged in the run-up to EMU takes Calmfors and Drif®ll's (1986) contribution as a starting point, which relates the centralisation of collective bargaining to the degree to which unions internalise the effect of collective bargaining on the macroeconomy. Early on, the risks of simply extending this analysis to the EMU context were made clear by Danthine and Hunt (1994). They showed that product market integration will play an important role in ¯attening out the `hump' therefore rendering centralisation of collective bargaining less relevant. Another strand has been explored by Cuikierman and Lippi (1998, 1999) who look at strategic interactions of the centralisation of nominal wage-setting and central bank independence. While these analyses are intellectually stimulating, I am convinced that the most pressing effects of monetary union derive from the fact that existing market imperfections and distortions will be subject to forces of competition; these effects are likely to swamp Barro±Gordon (1993) and Calmfors±Drif®ll (1988) effects and issues of time-consistency, reputation and coordination. I would therefore go even farther than Danthine and Hunt (1994) and argue that structural change implied for labour and product markets needs to be studied carefully before venturing guesses on the future strategy spaces of policy makers. It is, of course, the Lucas Critique again: the elasticity of labour demand will change, the objectives of labour unions will change, their constraint sets will change, the analyses cited above generally assume complete product market integration and ignore capital as a competing factor of production. Local national unions which insist on aggressive wage settlements will be faced with higher local joblessness. Only if the social safety net accommodates higher unemployment will unions be able to ignore these factors, and given the hard budget constraint of the monetary union, they will ®nd it increasingly dif®cult to do so.
266 Michael C. Burda
The Euro and labour market institutions An equally interesting hypothesis is that European jurisdictions will adapt and possibly reform labour market regulation in light of the increasing pressures brought about by EMU as well as globalisation and technological innovation. In this view, increased competition among member EU states as well as among regions within EU states will lead to a Nash equilibrium in which each member state disregards the effects its behaviour has on the others. This type of competition might emerge directly, in which some initiate direct labour market reforms in the hope of `beating the competition' and reap short to medium-term gains, the success stories of the Netherlands and Denmark might be viewed in this light. Another channel is increased tax competition ± especially, but not only, corporate taxation ± to enhance the attractiveness of investment in local economies (Standortwettbewerb in the local jargon), as Ireland has done aggressively in recent years. This tax competition puts strain on national member country ®nances and may force spending cuts and structural reforms. The experience of US states in this regard indicate that this mechanism can be powerful indeed. Bean (1998) has discussed this aspect.18 At the same time, it seems unlikely that the EU Commission and Parliament will sit idly and watch this `race to the bottom'. Already minimum capital taxation has been all but agreed to, while the probability of increased international (intra-European) competition along the social dimension is severely hampered by the Social Charter, which was rati®ed at the Strasbourg Summit in 1989 by all EU governments except the United Kingdom.19 The about-face on fast-track membership of the new market economies of Central and Eastern Europe may re¯ect a fear that unbridled competition in both regulatory and tax dimensions might be triggered by early admission of these countries. Yet the lack of consensus for a federal European ®scal policy means that little substantive support for harmonisation will come from the top. Summary What are the macroeconomic implications of increasing nominal rigidity and real ¯exibility, ceteris paribus? The empirical evidence, which is meant to be suggestive, supports the contention that nominal price rigidity has increased as a consequence of product market integration and exchange rate stability. Nominal wages, in contrast, are highly correlated only in the core, and this applies a fortiori to real wages and real exchange rates as well. These ®ndings suggest that the Euro will affect labour market ¯exibility in the direction of more ef®ciency. Without more detailed information on preferences, it is impossible to say whether this increase in ef®ciency will lead to overall welfare gains; some analyses, such as Agell (1998), claim that labour market rigidities may re¯ect welfare-improving policies in the light of other market imperfections. Burda (1995) has presented a related rationale for union wage compression.
European Labour Markets and the Euro 267
Wage-setting will become more fragmented unless pan-European efforts arise to coordinate. On the collective bargaining front, managing this change will require Herculean efforts on the part of national labour movements. In this vein one could expect a restructuring of unions in France, Spain and Italy (and possibly the United Kingdom) towards centralised industrial unions in order to facilitate cross-border cooperation; Dohse and Krieger-Boden (1998) describe the emergence of `European Works Councils' in large enterprises. Yet the reality of labour relations in these countries as well as the divergence of the interests of labour at the national level portend less dramatic changes (Streeck, 1998). While the possibility of pattern bargaining by large industrial unions ± as in Austria, Germany, or Sweden ± is frequently discussed, it is dif®cult to see how it could lead to truly coordinated outcomes without a strong central organisation as is the case in these countries. Because I see pan-European coordination coming in a decade's time at the earliest, a more modest goal for organised labour would simply be to get control over the process. The example of Eastern Germany can be seen as a lesson on how not to do it. Increasing nominal wage ¯exibility combined with nominal price rigidity is likely to lead to increased real wage ¯exibility. Casual evidence I have assembled in Tables 7.7 and 7.8 show that real wage behaviour in EU members has become increasingly uncorrelated over time, and that this tendency increases with the size of the group considered. This can be contrasted with US evidence, which shows a remarkably high correlation given the size of the regions considered. The empirical evidence suggests that while there is enough `insurance potential' in many respects to reduce Europe-wide risks, it is not showing up in wage growth rates. The dramatic deterioration of real wage correlations is evidence, to my mind at least, that there is potential for ¯exibility, at least between the core and the rest of the Euro-11. This ¯exibility supports my contention of a `forced decentralisation', which would not have been less likely had a two-track solution to the monetary union question been implemented. The breakdown of the synchronous behaviour of real wage growth in Germany and Holland in the early 1990s depicted in Figure 7.3 is one example of how this has occurred. The macroeconomic implications of increasing nominal rigidity and declining real rigidity, ceteris paribus, are somewhat surprising. The old conventional wisdom (Sachs, 1979, 1983, Bruno and Sachs, 1985) was that the United States was characterised by nominal rigidity but real wage ¯exibility, the nations of Europe in contrast had real rigidities but not nominal ones, which led to accentuated responsiveness of nominal wages to aggregate demand movements and to an attenuation of policy makers' ability to use monetary policy even to the limited extent now allowed in mainstream macroeconomics. The implications of my analysis is that Europe is likely to develop a more pronounced, common cycle as its own response to monetary policy evolves. This is the conclusion reached by more recent analyses such as Jeanne's (1998).
268 Michael C. Burda Table 7.7 Synchronisation of real wage growtha in Europe and the United States, 1961±96 Average correlation coef®cient in group (std. dev.) Total 1961±79 1980±96 sample
Smallest and largest moment matrix eigenvalues (1961±79) and (1980±96) 1961±79
1980±96
Percentage change
Core Europe (B, NL, D, A)
0.60 (0.08)
0.69 (0.16)
0.24 (0.38)
2.69 10�3 0.170
9.04 10�4 0.014
±66.4 ±91.5
Core Europe F, DK, IT
0.59 (0.13)
0.45 (0.24)
0.08 (0.41)
1.76 10�3 0.291
2.76 10�4 0.018
±84.3 ±93.6
Euro-11 lightb
0.55 (0.13)
0.36 (0.25)
0.06 (0.42)
9.96 10�4 0.405
1.937 10�4 0.026
±80.6 ±93.5
Euro-11 lightb 0.46 DK, S, UK (0.20)
0.35 (0.24)
0.14 (0.39)
5.62 10�4 0.455
1.35 10�5 0.036
±97.6 ±92.1
Memo: USA 8 regions, 1978±92, real comp.)
0.59 (0.18)
US (8 regions,
1978±92, 0.55 real wages and (0.20) salaries)
_
_
6.68 10�5 0.016
_
_
_
_
6.10 10�5 0.016
_
_
Notes: a Real wage growth is measured as ®rst difference in the logarithm of the nominal wage index
reported by the IMF, International Finance Statistics, divided by the IMF/IFS consumer price index.
b Less Luxembourg, Portugal.
For country codes, see Table 7.1.
Concluding remarks In addition to its historic dimensions, EMU will shed new light on a number of old, bothersome questions. Naturally, it will help us understand better how monetary unions function. In the ®rst instance, however, it will teach economists and policy makers the relevance of the new-Keynesian approach to understanding aggregate ¯uctuations, for which there is precious little evidence. It will also help us decide whether nominal price or wage rigidities are more relevant for explaining the real effects of aggregate demand ¯uctuations and thus the transmission mechanism itself. The convergence of exchange rate and especially price dynamics suggests that the preconditions for nominal price rigidities have become more
European Labour Markets and the Euro 269 Table 7.8 Manufacturing real wage growth correlations using different price indexes Average correlation coef®cient in group Wages de¯ated by OECD price index Wages de¯ated by IMF price indexa Total 1976±86 sample
1987±96
Total sample
1976±86
1987±96
Core Europe (B, NL, L, D, A)
0.39 (0.25)
0.49 (0.23)
0.06 (0.49)
0.44 (0.26)
0.50 (0.23)
0.17 (0.59)
Core Europe F, DK, IT
0.22 (0.26)
0.27 (0.30)
0.13 (0.38)
0.23 (0.43)
0.27 (0.27)
0.13 (0.43)
Euro-11
0.13 (0.25)
0.14 (0.30)
0.13 (0.36)
0.12 (0.25)
0.12 (0.29)
0.13 (0.38)
Euro-11 DK, S, UK
0.14 (0.23)
0.17 (0.29)
0.13 (0.35)
0.14 (0.38)
0.16 (0.29)
0.14 (0.36)
Notes: a Luxembourg excluded. Standard deviations in parentheses.
Per cent 14 United States 12
Germany
Changes to previous year
10
Netherlands
8 6 4 2 0 –2
1961
1966
1971
1976
1981
1986
1991
1996
–4 –6
Figure 7.3 Growth rates of real hourly compensation costs in manufacturing, 1960±96
favourable in Euroland. At the same time, trends in money±and, especially, real±wages point to declining importance of real rigidities. Real economic conditions and institutions are increasingly unfavourable for `business as usual' in the European union; the breakaway behaviour of the Netherlands, Denmark and possibly Ireland and Portugal supports the hypothesis that EMU
270 Michael C. Burda
is a Trojan horse of decentralisation ± not only de facto, but more importantly for structural reasons related to product and capital market integration. As many have recognised, the functioning of labour markets is central to the macroeconomic future of Euroland, but the mechanisms are remarkably subtle. The most important of my messages can be summarised as follows. First, the introduction of common currency, price transparency and internal trade integration ± will lead to a `inwardisation' of the European continent with the implication that internal and external nominal shocks will have less impact on nominal wage- and price-setting, and show up more strongly in output variation. Second, the standard analysis suggests that this will be related to the extent the underlying real economy is responsible for output ¯uctuations. In the past continental European countries were known for their `real rigidities' and in¯ation appeared to respond rapidly to changes in nominal demand.20 Yet my prediction that the EMU amounts to a `forced decentralisation programme' which will subject these rigidities to increasing pressure is accompanied by an optimism that a reduction of these rigidities will follow. Most important of the forces are increasing capital mobility, trade integration and competition, which will force wages for labour of given quality to converge (factor price equalisation) as well as to react more ¯exibly to changing local real conditions. Labour mobility, while a central point of discussion, is a side show which is not as relevant in the short run for the United States as it is made out to be.21 As more continental European countries scale back safety nets, it will become increasingly dif®cult for real wage determination to stay out in front of nominal developments. This ¯attening of the Phillips and aggregate supply curves will facilitate a more potent monetary policy. My prediction is that, unless an improbable miracle occurs in pan-European collective bargaining, labour markets will become more and not less ¯exible in the future. Calls for additional ¯exibility may be the economic equivalent of whipping a dead horse, and could provoke counterproductive reactions. As if it were not controversial enough to sell the Euro as the Trojan horse which liberalised labour markets, I ®nd it highly likely that it will change the macroeconomics of Europe fundamentally over the next decade and thus foster in a new regime for ®scal and monetary policy. Monetary policy should gain a new potency, as Europe begins to look more like the United States and Japan and less like Germany and France. A new role for monetary policy should emerge, although the usual caveat remains that the effectiveness of monetary policy is largely an artifact of its not being used in a predictable way to in¯ate the economy (Taylor, 1980). Therefore my chapter should not be construed as endorsing Oskar Lafontaine's `domestic demand strategy', but rather a warning that the temptation to employ such a strategy will increase in future years. Of course, my analysis is predicated on the view that nominal rigidities, especially price rigidities, are important in the evolution of a macroeconomy
European Labour Markets and the Euro 271
in the short run. If I turn out to be wrong and have to eat my hat, this fact will nevertheless have been useful information for our profession as well as policy makers. If I am right, EMU will have delivered the ultimate bonus in real ef®ciency gains for the unemployment-riddled labour markets of the continent.
Appendix In Tables 7.2±7.8, I present some suggestive evidence in support my twin hypotheses of increasing nominal rigidities on the one hand and decreasing real rigidities on the other. The variables considered are (1) consumer prices, (2) nominal wages for the total economy, (3) real wages, all from IMF, International Financial Statistics and using a longer sample (1961±96), data gathered by the US Bureau of Labor Statistics on manufacturing wages and exchange rates, and the OECD price index (1976±96). Correlations of ®rst differences in logarithms of these variables were examined in different grouping: a core group (Germany, Luxembourg, Belgium, Holland, and Austria); the core plus France, Italy and Denmark; the Euro-11; and ®nally the Euro-11 adding back Denmark, plus Sweden and the United Kingdom. The average correlation coef®cient provides a rough indicator of co-movement, while eigenvalues of the moment matrix indicate the extent to which linear combinations of countries' experiences can replicate others, the number of eigenvalues close to zero later indicates the extent to which `insurance' is possible. Notes 1. The view that short-run adjustment costs associated with EMU are small relative to long-run gains has been echoed by Buiter (1995). 2. See Mundell (1961), as well as McKinnon (1963); Kenen (1969). 3. Indirect estimates of labour mobility for the United States by Blanchard and Katz (1992) and for Europe by Decressin and Fatas (1995) show that European regions tend to adjust to adverse employment shocks via changes in labour force participation as opposed to residence. For more detailed summaries of the evidence, see Eichengreen (1993) and Gros and Hefeker (1998) as well as Obstfeld and Peri (1998). 4. See Gros and Hefeker (1998) for an overview. 5. It is remarkable that the optimal currency literature has largely ignored the role of capital mobility ± meaning long run mobility of the means of production ± despite Mundell's own explicit reference to it in his seminal article. For examples, see discussions in Bo®nger (1994); Bayoumi and Eichengreen (1996); Wyplosz (1997); Gros and Hefeker (1998). 6. See Bruno and Sachs (1985); Sachs (1979, 1983); but also Branson and Rotemberg (1980). 7. Among others, Romer (1996) has stressed the labour market as a primary source of real rigidities in the macroeconomy, as complementary to nominal rigidities.
272 Michael C. Burda 8. `The argument for ¯exible exchange rates is, strange to say, very nearly identical with the argument for daylight savings time. Isn't it absurd to change the clock in summer when exactly the same result could be achieved by having each individual change his habits? All that is required is that everyone decide to come to his of®ce an hour earlier, have lunch an hour earlier, etc. But obviously it is much simpler to change the clock that guides all than to have each individual separately change his pattern of reaction to the clock, even though all want to do so.' (Friedman (1953), p. 173, my emphasis). 9. In fact, the ®rm in panel (b) is more likely to ration output, producing only to the point at which price equals marginal cost, and thereby violating the assumption of completely passive (i.e. demand-determined) adjustment of production to demand. In any case the point is made that incentives to change prices in this case are large. 10. For evidence on the rigidity of prices in the United States see Carleton (1986), summaries of empirical evidence are available in Blanchard (1990) and Romer (1996). 11. This argument can also be found in McKinnon (1963), who stresses the role of nontraded goods in the reaction to devaluations. 12. The failure of ®rms selling into the United States fully to pass through exchange rate ¯uctuations is well documented (see Knetter, 1989; Dornbusch, 1987, 1996) and could be seen as an indication of what Euroland can expect. 13. One exception could be energy prices, which continue to be denominated in dollars. As Europe is the largest customer of the oil-exporting Middle East and Russia it may come to pass that oil prices are denominated in Euros. The relevant issue, of course, is whether oil prices in Euros will tend to become more stable over time. 14. The empirical evidence I present in this chapter is rather modest, as it seems foolish to place much weight on estimates of structures in place before monetary union. On the other hand many investigators have looked at the temporal evolution of crosscorrelation of price and quantity variables. (For example DeGrauwe, 1991; von Hagen and Neumann, 1994; Bayoumi and Eichengreen, 1996; Frankel and Rose, 1996). For details on the data used, the reader is referred to the Appendix (p. 271). 15. The Marshall±Hicks rule states that the elasticity of demand for labour is higher, the higher the elasticity of demand for output produced with that labour, the higher the elasticity of substitution between labour and other inputs the lower the elasticity of supply for those competing inputs and the greater the cost share of labour in production. See Hamermesh (1993). 16. For evidence on how product market competition has affected labour unions and labour markets in the United States in general, see Duca (1998). 17. Calmfors (1998a, 1998b); Gru È ner and Hefeker (1998); Cukierman and Lippi (1998, 1999); Soskice and Iversen (1999). 18. Arguing from a Barro±Gordon perspective, Calmfors (1998c) has conjectured that incentives to reform inside the EMU are greater than outside, since countries with control over monetary policy are likely to view labour market reform and monetary policy as substitutes for reducing unemployment, while inside EMU the latter vanishes. Reforming labour markets provides one means of ensuring against idiosyncratic shocks. This effect will be strengthened by ®scal pressures stemming from increasing unemployment, as well as the reorientation of national objective functions when in¯ation can no longer be in¯uenced by national policies. Similarly, Hefeker (1998) assumes unions which choose both the nominal wage and the degree of ¯exibility.
European Labour Markets and the Euro 273 19. For a discussion of these issues see Belke (1996). 20. See Bruno and Sachs (1985), Chapter 11, esp. pp. 232±40. 21. Willem Buiter (1995) has made this point, as have others. The results of Blanchard and Katz (1992) imply adjustment to adverse shocks in the United States which are long and drawn out over several years.
References Agell, J. (1998) `On the Bene®ts from Rigid Labour markets: Norms, Market Failures, and Social Insurance', Uppsala Dept of Economics, Working Paper, 1998:17. Akerlof, G. and J. L. Yellen (1985) `A Near-rational Model of the Business Cycle, with Wage and Price Inertia', Quarterly Journal of Economics, 100, 823±38. Ball, L. and D. Romer (1990) `Real Rigidities and the non-Neutrality of Money', Review of Economics and Statistics, 57, 183±203. Barro, R. and D. Gordon (1983) `A Positive Theory of Monetary Policy in a Natural Rate Model', Journal of Political Economy, 91, 589±610. Bayoumi, T. and B. Eichengreen (1993) `Shocking Aspects of European Monetary Integration', in F. Torres F. and F. Giavazzi, Adjustment and Growth in the European Monetary Union, Cambridge, Cambridge University Press. ÐÐÐÐ (1996) `Operationalizing the Theory of Optimum Currency Areas', CEPR Discussion Paper, 1484. Bean, C. (1998) `The Interaction of Aggregate Demand Policies and Labor Market Reform', Swedish Economic Policy Review 5, 353±82. Belke, A. (1996) `Implikationen einer zentralisierten Geld- und WaÈhrungspolitik fu È r die BeschaÈftigung in Europa', 8 UniversitaÈt Bochum, Diskussionspaper. Berthold, N. and R. Fehn (1998) `Does EMU Promote Labor-Market Reforms?', Kyklos, 51, 509±36. Bils, M. (1985) `Real Wages and the Business Cycle: Evidence from Panel Data', Journal of Political Economy, 93, 666±89. Blanchard, O. (1990) `Why does Money Affect Output: A Survey', Chapter 15 in B. Friedman B. and F. Hahn (eds), Handbook of Monetary Economics, Amsterdam, Elsevier. Blanchard, O. and L. Katz (1992) `Regional Evolutions', Brookings Papers on Economic Activity, 1, 1±75. Bo®nger, P. (1994) `Is Europe an Optimal Currency Area?', CEPR Discussion Paper, 915. Branson, W. and J. Rotemberg (1980) `International Adjustment with Wage Rigidity', European Economic Review, 13 (3), 750±77. Bruno, M. and J. Sachs (1985) Economics of Worldwide Stag¯ation, Cambridge, MA, Harvard University Press. Buiter, W. (1995) `Macroeconomic Policy during a Transition to Monetary Union', CEPR Discussion Paper, 1222. Burda, M. (1995) `Unions and Wage Insurance', CEPR Discussion Paper, 1232. Buti, M. and A. Sapir (1998) Economic Policy in the EMU, Oxford: Oxford University Press. Calmfors, L. (1998a) `Macroeconomic Policy, Wage Setting and Employment ± What Difference does EMU Make?', Oxford Review of Economic Policy, 14, 125±51. ÐÐÐÐ (1998b) `Unemployment, Labour-market Reform and Monetary Union', Institute for International Economics Studies Seminar Paper, 639. ÐÐÐÐ (1998c) `Monetary Union and Precautionary Labour-market Reform', Institute for International Economics Studies Seminar Paper, 659. Calmfors, L. and, Drif®ll (1988) `Bargaining Structure, Corporatism and Macroeconomic Performance', Economic Policy, 6, 3±62. Carleton, D. (1986) `The Rigidity of Prices', American Economic Review, 76, 637±58.
274 Michael C. Burda Cukierman, A. and F. Lippi (1998) `Central Bank Independence, Centralisation of Wage Bargaining, In¯ation and Unemployment ± Theory and Evidence', CEPR Discussion Paper, 1847. Cukierman, Alex and Francesco Lippi (1999) `Labor Markets and Monetary Union: A Preliminary Strategic Analysis', Tilburg University March mimeo. Danthine, J.-P. and J. Hunt (1994) `Wage Bargaining Structure, Employment and Economic Integration', Economic Journal, 104, 528±41. DeÂcressin, J. and A. Fatas (1995) `Regional Labor Market Dynamics in Europe', European Economic Review, 39, 1627±55. DeGrauwe, P. (1991) `Is Europe an Optimal Currency Area? Evidence from Regional Data', CEPR, Discussion Paper DP, Number, 555. Dohse, D. and C. Krieger-Boden (1998) WaÈhrungsunion und Arbeitsmarkt. Auftakt zu È bingen: Mohr. unabdingbaren Reformen, Kieler Studien 290, Tu Dornbusch, R. (1987) `Exchange Rates and Prices', American Economic Review, 77, 93± 106. ÐÐÐÐ (1996) `The Effectiveness of Exchange Rate Changes', Oxford Review of Economic Policy, 12, 26±38. Dornbusch, R., C. Favero and F. Giavazzi (1998) `Immediate Challenges for the European Central Bank', Economic Policy, 26, 15±64. Duca, J. (1998) `How Increased Product Market Competition May Be Reshaping America's Labor Markets', Federal Reserve Bank of Dallas Economic Review, Fourth Quarter, 2±16. Eichengreen, B. (1993) `European Monetary Uni®cation', Journal of Economic Literature, 31, 1321±57. ÐÐÐÐ (1998) `European Monetary Union: A Tour d'Horizon', Oxford Review of Economic Policy, 4, 24±40. Fatas, A. (1997) `EMU: Countries or Regions: Lessons from the EMS Experience', European Economic Review, 41, 743±51. Feldstein, M. (1992) `The Case against EMU', The Economist, 13 June, 19±22. Frankel, J. and A. Rose (1996) `The Endogeneity of the Optimum Currency Area Criteria', CEPR Discussion Paper, 1473. ÐÐÐÐ (1997) `Is EMU More Justi®able ex post than ex ante?', European Economic Review, 41, 753±60. Friedman, M. (1953) `The Case for Flexible Exchange Rates', in M. Fredman, Essays in Positive Economics, Chicago, University of Chicago Press, 153±203. Gros, D. and C. Hefeker (1998) `Factor Mobility, European Integration and Unemployment', Brussels, CEPS, mimeo. Gru È ner, H. and C. Hefeker (1999) `How will EMU Affect In¯ation and Unemployment in Europe', Scandinavian Journal of Economics, 101, 33±47. Hagen, J. von and M. J. M. Neumann (1994) `Real Exchange Rates Within and Between Currency Areas: How Far Away is EMU?', Review of Economics and Statistics, 2, 236±44. Hamermesh, D. (1993) Labour Demand, Princeton, Princeton University Press. Hefeker, C. (1998) `Labour Market Rigidities, Unions and Monetary Union', University of Bale, mimeo. Jeanne, O. (1998) `Generating Real Persistent Effects of Monetary Shocks: How Much Nominal Rigidity Do We Really Need?', European Economic Review, 42, 1009±32. Kenan, P. (1969) `The Theory of Optimum Currency Areas: An Electic View', in R. Mundell and A. Swoboda (eds), Problems of the International Economy, Chicago, University of Chicago Press.
European Labour Markets and the Euro 275 Knetter, M. (1989) `Price Discrimination by US and German Exporters', American Economic Review, 79, 198±210. Kollmann, R. (1999) `Explaining International Comovements of Output and Asset Returns: The Role of Money and Nominal Rigidities', University Paris XII, June, mimeo. Mankiw, N. G. (1985) `Small Menu Cost and Large Business Cycles: A Macroeconomic Model of Monopoly', Quarterly Journal of Economics, 100, 529±39. McKinnon, R. (1963) `Optimal Currency Areas', American Economic Review, 53, 715±25. Mundell, R. (1961) `The Theory of Optimal Currency Areas', American Economic Review, 51, 657±65. Obstfeld, M. and Peri, G. (1998) `Regional Non-Adjusment and Fiscal Policy', Economic Policy, 26: 205±59. OECD (1994) The Job Study, Parts I and II, Paris, OECD. Okun, A. (1981) Prices and Quantities A Macroeconomic Analysis, Washington: Brookings Institution. Ro È ger, W. (1998) `Output, Prices and Interest Rates Over the Business Cycle: How Well Does a Keynesian±Neoclassical Synthesis Model Perform?', European Commission, mimeo. Romer, D. (1996) Advanced Macroeconomics, New York, McGraw Hill. Sachs, J. (1979) `Wages, Pro®ts and Macroeconomic Adjustment: A Comparative Study', Brookings Papers on Economic Activity, 2, 269±319. ÐÐÐÐ (1983) `Real Wages and Unemployment in the OECD Countries', Brookings Papers on Economic Activity, 1, 255±304. Smith, J. (1999) `Nominal Wage Rigidity in the UK', Department of Economics, University of Warwick, March, mimeo. Soskice, D. and T. Iversen (1999) `Multiple Wage-Bargaining Systems in the single European Currency Area', Oxford Review of Economic Policy 14, 110±123. Streeck, W. (1998) `The Internationalisation of Industrial Relations in Europe: Prospects and Problems', Working Paper Series in European Studies, 1. Taylor, J. (1980) `Aggregate Dynamics and Staggered Contracts', Journal of Political Economy, 88, 1±24. Wyplosz, C. (1997) `EMU: Why and How it Might Happen', Journal of Economic Perspectives, 4 (11), 3±21.
Discussion
Hendrik J. Brouwer
Almost everyone agrees that a certain degree of ¯exibilisation of Europe's labour markets will be needed in order to help EMU absorb asymmetric shocks. We don't know much about the probability that these shocks will occur in Europe. What we do know is that the combination of EMU and the current wave of product market liberalisation in Europe will make reallocations of production factors increasingly likely. Production activities may become stronger concentrated along regional lines, as comparative advantages become more important now that exchange rate uncertainty within Europe has been eliminated. We also know that labour mobility is and will remain slower than in the United States. It is therefore important that labour markets are able to accommodate this process. Today's high unemployment in Europe points to the need of more ¯exible labour markets in order to reduce its structural component. The EU Council has acknowledged this and it has put this issue high on the political agenda. One could ask, however, whether it is also put high on the national political agendas. It should be, since the ultimate responsibility rests with the member states themselves. They are in the best position to determine which speci®c measures should be taken for their own economies. What types of labour market rigidities can we observe?
. A slow reaction of nominal wages to changes in real and/or nominal
conditions in the economy. This is the main rigidity addressed in Burda's Chapter 7. . But there is also evidence for lack of suf®cient wage differentiation ± for instance, if wages are based on average productivity movements in the economy. This causes a crowding-out of employment in low productivity sectors, like parts of the service sector. In some countries too high a level of legal or negotiated minimum wages reinforces this effect. . Another phenomenon is the rather persistent high non-wage labour cost. And, last but not least: . Regulation and contracts limit a suf®cient degree of ¯exibility.
276
Discussion 277
Burda tells us that the call for political action to achieve greater ¯exibility does not really matter much, since EMU itself will force labour markets in Europe to become suf®ciently more competitive. His policy message is surprising and indeed provocative. The question is whether that is the whole story. Let me ®rst note the issues on which I agree with Michael Burda:
. First, the chapter's central notion that EMU implies a major regime shift
which will not let the existing labour institutions in Europe remain intact.
. I also agree with the expectation that EMU may increase nominal price
rigidities. Not only the reasons mentioned in the chapter ± a reduced role for exchange rate changes and for outside competition (pp. 258±61) ± could play a role, but also the independent European Central Bank (ECB) with a mandate of maintaining price stability will make a stable price climate in Europe more likely. The general view that ®scal policy should contribute to stability ± as expressed in the Stability and Growth Pact ± gives further support. . I share Burda's scepticism about the likelihood of centralised wage negotiations in Europe (pp. 264±5). But there are some tendencies in that direction. Therefore, I would like to go one step further: Is it really useful to propagate and stimulate a European Social Dialogue? I doubt whether that could enforce labour market ¯exibility. So far, the results of this dialogue have produced the opposite. As I said, however, I am not convinced about all elements of Burda's chapter. Let me mention where I have some dif®culties:
. The chapter states that real rigidities will diminish through more
competition in the labour market which will lead to weakening of union power in wage determination (p. 264). This contrasts with the recent high wage demands in Germany, which do not give the impression of moderation on the side of the unions brought forth by EMU. . Of course, one could argue that such high wage demands are transitory but that still leaves the question of what time frame Burda has in mind with regard to this change in rigidities. In the meantime, when moving from one state of equilibrium to another, things could still go wrong. Hysteresis effects explain why Europe has such a high level of structural unemployment. . What could go wrong is not only that wage demands are excessive, but also that governments resist the transfer of a part of their coordination role to the market. On the contrary, painful adjustment in the labour market may well lead to a call for increased coordination by governments in the direction of short-term `quick ®xes', like the 35-hour working week, while neglecting the need for sometimes unpopular structural adjustment.
278 Hendrik J. Brouwer
. I think the chapter somewhat underestimates the role governments have to play, as this element is hardly discussed at all. It is well documented that current rigidities in the European labour markets relate not just to wage determination but also to laws and regulations, such as rules on working hours and on hiring and ®ring. There is much room for improvement and thus for action by politics in these areas. . I think we cannot exclude that a quite different and less rosy picture than the one sketched by Burda would result if nominal wages tended to exhibit the same degree of stickiness as prices. I see some reasons for such a development, such as (1) a possible lengthening of the duration of wage agreements and (2) a ¯oor in nominal wage contracts in a situation of low in¯ation, which can already be observed in some countries. . Finally, the empirical outcomes which the chapter presents are somewhat puzzling. It seems that some trends of the last decades already display the direction to which Burda's hypotheses point. For instance, it is noted (p. 267) that Tables 7.7 and 7.8 show an increasingly uncorrelated real wage behaviour in the EU over time. This prompts me to ask why these trends have already been operating for so long. Should this not be taken as indication that other variables than the coming of EMU have played a role in this respect? Also, the robustness of some of the outcomes is questionable. For instance, the correlations presented in Tables 7.2 and 7.3 (pp. 260±1), which measure the synchronisation of in¯ation, seem to contradict each other. According to Table 7.2 the degree of synchronisation of in¯ation has remained high, especially in the EU core group, whereas Table 7.3 suggests that this synchronisation has fallen all over the EU.
The conclusion of my comments is that, although enticing, it is better not to believe too much that the `invisible hand' of monetary integration will clear labour markets, but rather focus on how policy makers can help reduce existing rigidities. So I would like to change the sentence on p. 270: `Calls for additional ¯exibility may be the economic equivalent of whipping an apparently dead horse', and that could be a very useful thing to do!
Discussion
LuõÂs Campos e Cunha
1 Introduction Burda's chapter 4 is not an easy chapter to comment on. Its aim is `to entertain new-Keynesian arguments' without a well speci®ed model and, therefore, no clear policy recommendation can be derived from it, despite being related to policy issues. Furthermore, it is also not an empirical paper, because `the empirical evidence is rather modest' and it is questionable that such evidence would be of any relevance owing to the change of regime, as correctly emphasised by Burda. The chapter deals with institutional and sociological matters for discussing the impact of the change of economic regime due to monetary union, with particular emphasis on labour markets. As Burda says, `is highly speculative, but meant to be so'. I would say it is both speculative and provocative, and this is an excellent idea. Given these dif®culties, let me summarise my understanding of the main issues raised by the chapter and make some comments as I go along.
2 EMU and OCA: Mundellian problems Some of the issues the chapter promises to address are those found in the literature on optimal currency areas (OCA) following Mundell (1961). What is the optimal response of monetary policy to idiosyncratic shocks in an economy? Should there be one or multiple currencies, in order to respond optimally to these shocks? As the literature stands now, we are not in a condition to answer all these questions satisfactorily. However, one thing is certain: the Mundellian nightmare regarding OCA in the context of EMU has been oversold. The understanding of exchange rate policy is very different today than it was in 1961. More speci®cally, it is today widely accepted that no exchange rate policy can insulate an economy from real shocks. To assert that exchange rates can do more than insulate a country from nominal shocks is an incorrect reading of Mundell (1961). This clari®cation has been recently made in Mundell (1997, 1998). 279
280 LuõÂs Campos e Cunha
The lack of factor mobility ± namely, labour ± is another related problem raised in Burda's chapter. I do not think that the lack of actual signi®cant labour ¯ows within the EU-11 is a sound proof of rigidities in this area, as claimed by Burda. First, as already noted, due to both factor price equalisation (FPE) and to capital ¯ows. These have been large and increasing during the last decade and, in a certain class of models, they are perfect substitutes for labour mobility. Secondly, the `unwillingness' of people to move, explained by older demographics as referred to by Burda, is not a distortion. If people do not want to move, why should they? Unwillingness to migrate across borders is not a component of rigidity or a distortion of any sort. As a ®nal note, in the 1960s, there was large labour migration within Europe despite its being much more dif®cult than today from any point of view. When people were willing to move, they did.
3
EMU: a change of regime
A second cornerstone of Burda's chapter is the change of regime, in the sense used in Lucas (1976). This is a very important and relevant issue and its impact both in the short run and long run is dif®cult to assess. One should note that an argument of change of regime is an elegant one, but it remains to be seen whether it is relevant as concerns factor market rigidities in `Euroland'. In any case, in the chapter it is argued that nominal rigidities will increase but real wage ¯exibility will prevail. Real wage ¯exibility As regards wages, it is claimed that EMU will lead to a change in institutions of wage determination that will make the EU look more like the United States today, in the sense that the change of regime will imply an increase of nominal rigidities and a reduction of real rigidities. The argument is based on the idea that trade unions will lose much of their bargaining power and, therefore, nominal and real wage will become more ¯exible. On the one hand, it is not clear why national trade unions should be weakened. Alternatively, one could argue that competition among them will increase, leading to a reversed conclusion. On the other hand, most of the rigidities of the labour market mentioned are due to regulation: I have in mind minimum wage laws, unemployment compensation schemes, restrictions to ®rms to adjust their labour force, and so on. These will not vanish either because of monetary union or the lack of international coordination of national labour unions, for the same reason as, in the past, those rigidities were not a consequence of the existence of several currencies or of labour union coordination. Wage bargaining and the role of labour unions under the new regime could be better discussed in a model of insiders/outsiders. We should then look for changes that could induce an increase or decrease in the reaction of wages to local employment conditions and to unemployment persistence.
Discussion 281
Price rigidities The chapter also argues that as the regime changes, price stickiness in the goods market will increase. The author claims that EMU will lead to an increase in nominal rigidities because the share of `home goods', goods traded inside Europe, is larger than the total of home goods across all countries in the Union. The reasoning is that these goods will not be exposed to ¯uctuations and this will favour nominal rigidities. Both parts of the argument need clari®cation: is Burda arguing that shocks to goods traded inside the EU were mainly driven by national monetary policies? Otherwise, ¯uctuations in these sectors will certainly persist. Furthermore, why should a more stable environment lead to higher nominal rigidities? This `inwardisation' effect owing to the single currency is probably too much emphasised, since most of the effects were anticipated at least for the socalled `core' countries. Furthermore, the `closing' of the Euroland economies does not have the same nature of a country imposing tariff barriers and capital controls. With the single currency the domestic economies become bigger, and there is an increase in competition owing to reduced transaction costs and increased price transparency. Market segmentation is less likely to take place and, therefore, ®rms are more ± not less ± exposed to competition as economic frontiers are pushed further away, so it is not obvious that monopolistic power will necessarily increase.
4 Concluding remarks on policy I would like to believe, using the wording of the chapter, that, `EMU is the Trojan horse of decentralisation' and that the `Euro is the Trojan horse which will liberalise labour markets', and therefore, `calls for additional ¯exibility may be the economic equivalent of whipping a dead horse'. Unfortunately, most of us feel that we have to continue to call for some structural reforms leading to institutional and legal changes so that structural unemployment will fall. The Euro will not be the solution for high unemployment rates. Unlike Burda, I am more concerned with the symmetric consequences in the new regime. We could imagine consequences concerning labour markets owing to compulsory centralised harmonisation. In fact, regions of Euroland with lower labour productivity should be vigilant and oppose attempts to harmonise labour market regulation throughout the Union. Just imagine the impact of a forceful and a speedy harmonisation of minimum wage towards the core country levels. It would be the equivalent of Brussels offering a rope to thousands of workers to hang themselves and it would be very dif®cult for them to resist. This would imply a jump in unemployment rate for lowerproductivity regions as real rigidities increased. This is foreseeable and it would be a big step in the wrong direction. As regards monetary policy, Burda makes a point that stronger price persistence in Euroland will facilitate a more powerful monetary policy, which
282 LuõÂs Campos e Cunha
is almost an invitation to a more active monetary policy. Without a more detailed mechanism of how persistence comes about, this is a claim dif®cult to make and at the same time to dispute. For instance, we cannot talk about `monetary policy effectiveness' without saying how price expectations are formed, which is not explained in the chapter. On the other hand, if one believes ± as I do ± that money is basically neutral in the long run, an increase in price persistency, if true, would call for the European Central Bank (ECB) to have an even longer-run orientation while conducting its monetary policy. A mistake made today will be paid further away in time owing to the increased persistence of prices in Euroland: the long lags of monetary policy will become even longer. To sum up, persistency does not necessarily facilitate the job of the ECB. I would like to ®nish by stressing that the chapter makes a valuable contribution to the debate and it was a pleasure to participate in it. References Lucas, R. (1976), `Econometric Policy Evaluation: A critique', in K. Brunner and A. H. Meltzer (eds), The Phillips Curve and Labor Markets, Amsterdam, North-Holland, 19±46. Mundell, R. (1961), `The Theory of Optimal Currency Areas', American Economic Review, 51, 657±65. ÐÐÐÐ (1997) `Currency Areas, Common Currencies and EMU', American Economic Review, (2) 214±16. ÐÐÐÐ (1998) `Great Expectation for the Euro', The Wall Street Journal, March, 24±5.
8
The Monetary Transmission Process: Concluding Remarks1 Otmar Issing
1 Introduction In order to be successful in conducting monetary policy, central banks need to have a good understanding of the working of the economy, including an accurate assessment of the timing and the effects of changes in the policy instrument on in¯ation and economic activity ± that is, the monetary transmission process (MTP). Such an assessment is necessary in order to tailor the policy response to unexpected developments in the economy and successfully maintain price stability. The introduction of the Euro constitutes a unique experience in monetary economics and central banking. This historical regime change is likely to affect the size and the timing of the various transmission channels. Analysing the potential changes in the structure of the economy caused by the start of Stage III of EMU is therefore of great importance and will surely be an object of research for decades to come. Economists are indeed likely to regard the introduction of the Euro as a `laboratory experiment' of the Lucas Critique. However, the Eurosystem does not have the luxury of waiting for the results of this research. The Governing Council started conducting an independent monetary policy on 4 January 1999. Since then it has met twice a month to assess the outlook for in¯ation and decide on the appropriate policy actions. Central bankers are used to making decisions on the basis of incomplete, evolving information. As you all know, it was this lack of precise knowledge about the variable and long transmission lags of monetary policy that led Milton Friedman to warn against an activist monetary policy. The stability-oriented strategy announced by the Council in October 1998 is designed to guide monetary policy through the uncertainties that must inevitably accompany the transition to the single currency. Of course, it is in no way a substitute for a thorough analysis of the state of the Euro Area economy, its implications for the in¯ation outlook and the policy actions necessary to achieve price stability. On the contrary, the implementation of the strategy presupposes a continuing investigation of the working of the Euro 283
284 Otmar Issing
Area economy, including the transmission process. In this respect, conferences like this one are a very important source of information, particularly if they bring together such a distinguished group of outstanding researchers on monetary economics. Let me then use this opportunity to brie¯y give you my own reading of the key messages of the discussion. I would like to review two key aspects. First, what do we know about the MTP in the Euro Area and, second, how it is likely to change?
2
What do we know about the MTP in the Euro Area?
There is by now quite a large literature on the MTP in the Euro Area countries. While the focus of this literature has often been on the cross-country differences and the implications for the common monetary policy, a lot can also be learned from this literature concerning Euro Area-wide transmission. Two approaches have been used which I will discuss in turn. One approach is to describe and analyse the institutional features of the economies which have a bearing on the MTP. Another is to use econometric models to estimate the effects of a change in policy-controlled interest rates on economic activity and in¯ation. Let me brie¯y review the results of both approaches. Institutional comparison The way in which changes in monetary policy feed through the various sectors of the economy depends to a large extent on the structural and institutional features of its ®nancial, labour and goods markets. These institutions are rooted in the history of each country and often resulted from the interplay between particular macroeconomic shocks, the preferences and bargaining powers of various agents in the economy and more or less market-friendly regulatory and government forces. Researchers have devoted a lot of energy to investigating whether such institutional characteristics, which from a theoretical perspective may affect the impact of monetary policy, differ across countries or regions. Often the focus on one particular aspect of the MTP leads to the conclusion that the transmission of monetary policy impulses will be very heterogeneous across Euro Area countries. However, if one brings several of these features together (as is done in Table C1.1), then it is much less obvious that the effects of policy will be very different. Generally, countries that are particularly sensitive to policy changes because of one criterion, will be less sensitive on the basis of other characteristics. Let me brie¯y go through some examples. Goods markets One important characteristic of the goods markets which affects the transmission channels of monetary policy is the degree of openness. While the Euro Area as a whole will be a relatively closed economy ± with a ratio of exports to
Concluding Remarks 285
GDP of about 15 per cent, it is only marginally more open than the United States or Japan ± there is a fair degree of dispersion within the Euro Area (see Table C1.1). For example, Belgium is about three times as open to the nonEuro Area as Spain. As a consequence, while the exchange rate channel may be relatively less important for the Euro Area as a whole, the strength of its impact on output and in¯ation may be quite different across countries. Similarly, the overall interest elasticity of aggregate demand will depend on the importance of interest rate sensitive sectors in the economy. For example, Carlino and DeFina (1998) ®nd that, within the United States, regions with a higher weight of construction and manufacturing in the economy are more strongly affected by the Federal Reserve System's monetary policy. If similar considerations are true within the Euro Area, then one could expect that, for example, Germany and Spain should experience a stronger interest rate channel than the Netherlands or Belgium. Another sector which is highly responsive to the level of the interest rate is the housing sector. Maclennan, Muellbauer and Stephens (1998) have expressed concern about the implications of cross-country differences in the tax and legal framework of housing markets for the transmission mechanism. In particular, transaction costs owing to taxes on new houses and stamp duties are quite heterogeneous across countries. A particularly low tax rate in the United Kingdom is consistent with the high amplitude of its housing market cycle. Within the Euro Area, Germany, Italy and Spain have much smaller transaction costs than the Netherlands, France and Belgium. Financial structure Table C1.1 also summarises a number of features of ®nancial structure across countries. From a comparison of these features across countries, a number of tentative conclusions can be drawn. Countries in the Euro Area appear to have several features in common. Stock markets and corporate debt securities markets are less well developed than in the United States, the United Kingdom or even in Japan. Of the seven largest Euro countries, the Netherlands is the only one where the number of publicly traded ®rms and the stock market capitalisation show levels that are comparable to those of the United States or the United Kingdom. Obviously, this implies that equity prices in the Euro Area should play a less important role in the MTP through both the effect on the cost of capital and wealth effects on consumption. Similarly, France is the only country where the corporate sector issues a signi®cant share of its debt directly on the market. In 1997, up to one-third of the debt of French ®rms consisted of securities, while it was below 10 per cent in the other countries. To the extent that bank lending rates are more sticky than market rates in response to a change in policy-controlled rates, the interest rate channel should be weaker in the Euro Area.2 However, there remain some important differences in ®nancial structure within the Euro Area. For example, the degree of ®nancial intermediation is
Table C1.1 Institutional features with a bearing on the transmission mechanisma Goods markets Non-Euro Manufactur. exports and / GDP construction (per cent) / GDP (per cent) Belgium Germany Spain France Italy Netherlands Austria Average*b
26.8 14.0 8.1 10.5 11.7 18.5 12.9 14.6
USA Japan UK Year
8.5 10 21.4 1997
Source
IMF, Direction of Trade Statistics
27 37 32 27 30 24 29 29
1980±90 Carlino and DeFina (1998)
Financial markets
Stamp duty VAT on new homes (per cent)
Number of publicly traded ®rms / per capita
33 2 9 25.6 8.2 23.5 16.9
13.7 8.3 9.1 11.7 3.8 14.0 13.1 11.3
39 24 35 34 19 90 14 36
7
6
9
15
5
3
2
6.7
38 49 34 / 50 40 47 29 65 44.7
3 3 3 3 3 3 3 3
17 15 19 14 20 9 16 15.7
1 1997
31.9 18.6 41.4 1996
95 72 127 1995
20 10 19 1993
39 71 85 1 993
1 2 2 1989±94
1 8 7 1989±94
25 30 152 1989±94
Prati and Shinasi (1997)
Borio (1995)
Borio (1995)
Nickell (1997)
Nickell (1997)
Nickell (1997)
Maclennan, Muellbauer and Stephens (1998)
Cecchetti (1999)
Stock Securities market / Total ®rm capitalisation debt / GDP (per cent) (per cent)
Labour markets Credit at adjustable rate / GDP (per cent)
Union Employment Unemploy. coverage protection bene®t index index index
240 252 245 171 10 140 100 165.4
Notes: a These indicators should be taken as suggestive. b Simple average of the seven countries above. Goods markets indicators: Share (in GDP) of exports vis-a-vis countries that are not in the Euro Area. For the United States, Japan and the United Kingdom: total exports; Share of manufacturing and construction sectors in GDP; Sum of stamp duty on housing transactions and VAT rate on new houses. Financial markets indicators: Number of publicly traded ®rms per million of population; Stock market capitalisation scaled by GDP; Share of securities debt in non-®nancial corporate sector total debt; Credit at adjustable rate scaled by GDP ± i.e. total credit over GDP times the proportion of total credit at adjustable rate as in Borio (1995).
286
Concluding Remarks 287
quite different among the large Euro Area economies with loans of credit institutions to non-banks ranging from as low as 67 per cent of GDP in Italy, to as high as about 150 per cent of GDP in Germany and in the Netherlands. Similarly, the maturity structure of the debt of the private sector differs considerably across countries (Borio, 1995). Labour market structure The wage-setting process is a key element in the transmission from monetary policy impulses to prices. Continental European countries are generally characterised as having lower nominal wage rigidity but higher real wage rigidity than the US.3 The implications for the transmission process are not unambiguous. Higher real wage rigidity in the Euro Area implies that the output cost of achieving a similar effect on in¯ation may be greater. This would tend to increase the sacri®ce ratio. On the other hand, a high response of nominal wages to expected in¯ation has the potential of making the direct transmission process of monetary policy on wages and prices much more effective to the extent that the central bank's in¯ation objectives are highly credible. A comparison of labour market structures across countries has usually put more emphasis on understanding the level of European structural unemployment than the transmission process.4 Except for the United Kingdom, where the labour market is generally much less regulated than on the continent, it is dif®cult to rank countries because there are many dimensions of labour market rigidities.5 Euro Area labour market institutions are generally characterised by unemployment bene®ts, worker protection and union participation that are much larger than in the United States. Yet, the hierarchy among Euro countries in terms of institutional factors of rigidity is far from obvious. Although most Euro countries have a high degree of centralisation in wage negotiations, there are marked differences along some other dimensions. For example, when it comes to employment protection, Italy and Spain appear to be the most rigid, while the Netherlands is much less so. In contrast, in terms of the generosity and duration of unemployment bene®ts, Italy is far below other Euro countries, while Germany, Spain and Belgium stand at the top. In sum, this brief overview of institutional features shows that in many respects the MTP in Euro Area countries share several common features. Moreover, although important institutional differences remain, the ranking of countries in terms of their implications for the strength of the MTP is not clear. Countries that score low on one criterion may score high on another. Indeed, the average ranking of the seven biggest countries in the Euro Area according to the qualitative impact associated to each of the ten indicators of Table C1.1 suggests there are no clear differences in the strength of the MTP among these countries.
288 Otmar Issing
Macroeconometric analyses of the MTP This general picture is con®rmed by the many studies that have examined the transmission process in Euro Area countries using a variety of macroeconometric approaches (vector autoregressions (VARs), large-scale macroeconometric models, smaller-scale calibrated models, etc). As emphasised by Kieler and Saarenheimo (1998), there are no cross-country differences in the strength and timing of the MTP that are robust across these studies. While this result may in part be due to differences in methodologies and the statistical imprecision of some of the estimates, it suggests that it is not straightforward to ®nd clear cross-country differences. In light of our ®ndings concerning the institutional differences across countries, this is not too surprising. Overall, the response of real GDP and in¯ation to a monetary policy tightening follows the pattern predicted by textbook theory. An increase in the policy-controlled interest rate leads to an immediate decline in GDP, which reaches its trough after ®ve±eight quarters before it comes back to the base-line. The combined impact of a decrease in import prices following the appreciation of the exchange rate and downward pressure on wages coming from rising unemployment leads to a fall in prices. Typically, however, the fall in in¯ation takes somewhat longer to materialise than that of output. This is also the pattern of responses that can be obtained from simulating the effects of a monetary policy tightening in various models estimated on aggregate Euro Area data. Table C1.2 shows the effects of a policy tightening in the Euro Area using three different macroeconometric approaches: a VAR model by Monticelli and Tristani (1998), a two-equation model estimated by Peersman and Smets (1999) and the ECB's area-wide model. There are many reasons why models estimated on synthetic historical data may be biased and give the wrong answers. Nevertheless, on the basis of existing evidence two conclusions can be drawn. First, the qualitative effects are very similar to what I have previously described and what has been found for the United States. Second, there is considerable uncertainty about the precise timing and size of the effects on economic activity and in¯ation. Let me now turn to the question of how the MTP and differences across countries may be affected by the creation of the single currency.
3 The monetary transmission process in the Euro Area: is it likely to change? For the Eurosystem it will be even more important than for most other central banks to continuously monitor any developments that may change the transmission process of monetary policy. I would like to focus on the likely developments in the ®nancial sector and its impact on the MTP. I will not elaborate on the potential effects of the single currency on goods and labour markets. Let me just underline that, particularly in the case of labour markets, further structural reform is needed not only to reduce the
Table C1.2 Simulation of a monetary policy tightening Monticelli and Tristani (1998) Horizon Interest rate GDP Consumer prices
Year 1 0.102 ±0.18 0.07
Peersman and Smets (1999)
ECB area-wide model
Year 2
Year 3
Year 1
Year 2
Year 3
Year 1
Year 2
Year 3
0.044 ±0.4 ±0.16
±0.025 ±0.38 ±0.45
0.5 ±0.18 ±0.11
0.5 ±0.32 ±0.7
0 ±0.27 ±1.8
0.5 ±0.07 ±0.11
0.5 ±0.16 ±0.22
0.5 ±0.23 ±0.25
289
290 Otmar Issing
equilibrium unemployment rate but also to increase labour market ¯exibility and thereby improve the ability of the economy to cope with macroeconomic shocks. Our hope is that the establishment of a stability-oriented single monetary policy in the Euro Area will provide incentives to continue on the path of reform. At this point, I would also like to stress that, given the historically unprecedented nature of EMU, it is natural to emphasise the potential importance of structural breaks owing to the start of Stage III. One should, however, realise that, ®rst, a long and gradual process of monetary convergence has preceded the introduction of the new currency, so that many of the adjustments may already have taken place. Second, the effect of the single currency on competition in goods, labour and ®nancial markets follows previous structural changes such as the Single Market initiative, so that whatever structural change will take place is part of an on-going process. The single monetary policy As of 4 January 1999 the Eurosystem started to implement a single monetary policy for the Euro Area. This has by itself potentially large effects on the way changes in policy instruments affect the economy. In fact, it is generally meaningless to discuss the MTP without a discussion of the monetary policy regime. The monetary policy strategy and its credibility will affect the way changes in policy instruments are interpreted by the private sector and feed through into their expectations and actions. Let me brie¯y give two examples. The effects of the single monetary policy on the MTP will be most visible in the way asset prices respond to policy. For example, while in the past responses of long-term interest rates to policy innovations were quite diverse across Euro Area countries, more recently there have been abundant signs of the development of a single private sector yield curve, which responds homogeneously irrespective of the nationality of the borrower. Second, the credibility of the Eurosystem's price stability objective is likely to promote longer-term nominal contracting in both labour and ®nancial markets. Such developments are again likely to homogenise the transmission process. For example, evidence suggests that in Italy, where historically the average maturity of debt contracts was relatively short owing to high and variable in¯ation, an increasing proportion of the new housing debt is issued at a ®xed interest rate. The emergence of an integrated ®nancial market in the Euro Area Monetary integration is a major step in the completion of the integration of European ®nancial markets. Such an integrated ®nancial market denominated in Euros may provide investors and issuers with levels of deepness and liquidity which they could ®nd before only in the United States. These developments will be accompanied by the spreading of ®nancial market innovation across Europe. For example, direct ®nance through
Concluding Remarks 291
commercial paper and corporate bonds for non-®nancial ®rms may increase in countries where it has not yet developed and reach levels that are currently observed only in France (a third of the corporate sector debt). The potential consequences of the development of direct ®nance and a changing role of the banking sector in Europe for the transmission may be signi®cant. Competition between direct ®nance and bank ®nance will put pressure on banks and may increase the speed of transmission from the money market rate to the rate at which ®rms borrow. For example, in France, evidence shows that the money market rate has taken over the prime rate as the reference to compute the bank credit rate to big ®rms as well as to smaller ones. In the short run, it may, however, also increase the fragility of the banking sector. On the other hand, cultural preferences for ®xed rate contracts and against indebtedness, national consumer protection laws and inertia in consumer habits will tend to slow down cross-country competition in retail banking. Retail banking involves long-term investment in brand names, legal expertise in speci®c national laws and monitoring of borrowers' solvency. These provide incumbent banks with substantial advantages over competitors, especially foreign ones, so that initially the impact of EMU on retail banking could be limited to the harmonisation of banks' funding costs. The development of the Euro as an international currency The development of the Euro as a strong and credible currency traded in deep and liquid ®nancial markets may also affect the importance of the exchange rate channel to the extent that it enhances the role of the Euro as an invoice and anchor currency. While currently the exchange rate channel may be somewhat more important than in the United States, its importance may be further reduced if several trade partners ± for example, in Eastern Europe ± peg their currency to the Euro and to the extent that the price of some commodities could be traded in Euros.
4 Conclusion This review of the existing literature on the MTP in the Euro Area con®rms the validity of Friedman's famous dictum that monetary policy lags are long and variable ± and, I would add, very uncertain. In the case of the Euro Area, uncertainties are magni®ed by the scarcity of reliable aggregate data and the potential effects of a regime change. The monetary policy strategy of the Eurosystem was designed to face, in a pragmatic, yet conceptually sound way, the many uncertainties inherent in this regime change. The strategy is at the same time transparent and ¯exible. It clearly communicates the commitment to price stability by de®ning price stability as an increase of the area-wide harmonised index of consumer prices of below 2 per cent. This helps to anchor in¯ation expectations and preserve the antiin¯ationary reputation inherited from its precursors, the national central
292 Otmar Issing
banks (NCBs). In addition, two pillars ± a reference value for the growth of a broad money aggregate and a broad-based assessment of the outlook for in¯ation ± are used to explain monetary policy decisions necessary to achieve the price stability objective. The prominent role of money ± which we consider as a reference value, not a target in the traditional sense intended in the monetary literature ± underlines the principle of continuity and is rooted in robust theoretical and empirical arguments accumulated over many decades of research. Indeed, probably the only statement with which everybody would agree is that in¯ation is, in the long run, a monetary phenomenon. However, uncertainties in the stability of money demand would make it inappropriate to automatically respond to deviations of the growth in money aggregates from target. This is why the Council decided not to adopt a strict monetary targeting framework. Instead, the prominent role of money as a reference value is complemented by a broad-based assessment of the in¯ation outlook, including both model-based forecasts and a variety of other relevant indicators. Together these two pillars allow for a timely and ¯exible response to a changing environment, while keeping the objective of price stability in clear focus. The strategy will help communicating with the public and organising policy action in a coherent way. We do not see it as an invariable dogma. The ECB closely follows the developments in economic research: this is a key factor in our continuous assessment of the strategy itself. We look forward to a continuous, productive exchange between the ECB and the academic world on these issues. Notes 1. I thank Frank Smets and BenoõÃt Mojon for their valuable assistance. 2. There remains uncertainty about the effect of bank dependence on the MTP. For example Cecchetti (Chapter 5 in this volume) has emphasised that countries in which ®rms are more bank-dependent will be more sensitive to the Eurosystem's decisions to change interest rates through a kind of credit channel. However, Schmidt has argued in chapter 6 in this volume that long-term relationship bank lending, which occurs more naturally in a bank-dominated ®nancial system, has led banks to insure customers against interest rate shocks. 3. See, for example, Calmfors (1998). 4. See, for example, Blanchard (1998) or Bean (1994). 5. In addition, Burda's Chapter 7 in this volume shows that there is not yet a consensus as to how the institutional features of continental European labour markets in¯uence macroeconomic outcomes.
References Bean, C. (1994) `European Unemployment: A Survey', Journal of Economic Literature, 32(2), 573±619. Blanchard, O. (1998) `European Unemployment. Shocks and Institutions', Baf® Lecture, mimeo.
Concluding Remarks 293 Borio, C. E. V. (1995) `The Structure of Credit to the Non-government Sector and the Transmission Mechanism of Monetary Policy: A Cross-country Comparison', in Bank for International Settlements, Financial Structure and the Monetary Policy Transmission Mechanism, CB 394, Basle. Burda, M. (1999) `European Labour Markets and the Euro: How Much Flexibility Do We Really Need?', Chapter 7 in this volume. Calmfors, L. (1998) `Macroeconomic Policy, Wage Setting and Employment ± What Difference Does EMU Make?', Oxford Review of Economic Policy, 14, 125±51. Carlino G. and R. DeFina (1998) `Monetary Policy and the U.S. States and Regions: Some Implications for European Monetary Union', Federal Reserve Bank of Philadelphia, Working Paper, 98±17. Cecchetti, S. (1999) `Legal Structure, Financial Structure and the Monetary Policy Transmission Mechanism', Chapter 5 in this volume. Kieler M. and T. Saarenheimo (1998) `Differences in Monetary Policy Transmission?: A Case not Closed', Economic Papers, 132, European Commission, Directorate-General for Economic and Financial Affairs. Maclennan, D., J. Muellbauer and M. Stephens, (1998) `Asymmetries in Housing and Financial Markets Institutions and EMU', Oxford Review of Economic Policy, 14, 54±81. Monticelli, C. and O. Tristani (1998) `What Does the Single Monetary Policy Do? A SVAR Benchmark for the European Central Bank', DG Research of the European Central Bank, mimeo. Nickell, S. (1997) `Unemployment and Labour Market Rigidities: Europe versus North America', Journal of Economic Perspectives, 11(3), 55±74. Peersman G. and F. Smets (1999) `The Taylor Rule: A Useful Monetary Policy Benchmark for the Euro Area?', International Finance, 2(1), 85±116. Prati, A. and G. Shinasi (1997) `EMU and International Capital Markets', in P. Masson, T. Krueger and B. Turtelboom (eds), EMU and the International Monetary System, Washington, DC, International Monetary Fund. Schmidt, R. `Differences between Financial Systems in European Countries: Consequences for EMU', Chapter 6 in this volume.
Francesco Giavazzi
The wave of mergers and acquisitions that is sweeping the European banking industry should not be a matter of indifference for those concerned with the possibility that the monetary transmission mechanism may work asymmetrically across EMU: for two reasons. First, banks, as is well known, are at the centre of ®nancial markets in continental Europe, and thus of the transmission mechanism. Second, the consolidation of the European banking industry, which is currently happening, may sharpen, rather than reduce, the existing sources of crosscountry asymmetries. Evidence of the importance of banks in the transmission mechanism inside EMU is related to the extent to which monetary policy operates through a `credit' channel, rather than simply through a `money' channel. As is well known, when loans and bonds are imperfect substitutes in the balance sheets of banks, following a squeeze in liquidity, banks reduce the amount of loans they supply. Firms could turn to the bond market, but if bonds and loans are imperfect substitutes the external ®nance premium will go up. This effect [known as the `credit' channel (see, for example, Bernanke and Gertler, 1995)] works on the supply side and ampli®es the more traditional demand effect of a monetary tightening ± the change in interest rates which affects new marginal spending by modifying borrowing conditions and by affecting asset prices, and thus the market value of wealth. Because of the large share of bank loans in the total debt liabilities of European non-®nancial ®rms (85 per cent in Germany, 80 per cent in France and Spain, 95 per cent in Italy, as opposed to 32 per cent in the United States) the behaviour of banks is thus central to understanding the transmission mechanism in Europe. Reliable identi®cation of a `credit' channel requires the use of micro data: macroeconomic time series are ill-suited to identify a `credit' channel from a `money' channel in the transmission of monetary policy from the central bank to banks. This is because the money channel works through banks' liabilities, while the credit channel works through their assets, but assets and liabilities are tightly related by accounting identities. For this reason, the evidence proposed by macroeconomic studies which look at output and price ¯uctuations in response to shifts in the quantities of loans and deposits is rarely decisive. On the contrary, microeconomic data allow one to ask whether the responses of banks and ®rms to a shift in monetary policy differ depending on their characteristics ± their size, for instance: small ®rms are more likely to be liquidity-constrained and to depend on banks for ®nancing; similarly, small banks ®nd it more dif®cult to insulate their loans' portfolio from a squeeze in central bank liquidity, because a small bank typically cannot use bond 294
Concluding Remarks 295
holdings as a buffer (Kashyap and Stein, 1997; Kashyap, Stein and Wilcox, 1993). So far, the empirical evidence on the importance of bank characteristics in determining the response of loans to a shift in monetary policy has been limited to US data. Favero, Flabbi and Giavazzi (1999) have extended this evidence to Europe. Their ®ndings, albeit limited to a speci®c episode ± the EMU-wide monetary contraction which occurred during 1992 ± point in two main directions. In Germany there is clear evidence of a lending channel for all banks, except the smallest ones ± that is, for over 80 per cent of the German banking sector. In France, on the contrary, there is no evidence of a lending channel, independently of the size of the banks considered. The large Italian banks are similar to the corresponding German banks, in the sense that there is evidence of a lending channel: the impact on the supply of loans of a change in reserves is however twice as large as for the corresponding German banks. The difference between Germany and Italy lies in medium-sized banks (whose share of the domestic banking industry is larger in Italy than it is in Germany): while there is evidence of a lending channel for medium-sized German banks, this is not the case for the corresponding Italian banks. In Spain, there is evidence of a lending channel for medium-sized banks: the impact on lending of a cut in reserves is similar to that found in Germany. The empirical evidence thus points to signi®cant cross-sectional and crosscountry differences in the response of individual banks to monetary policy ± at least in the speci®c episode studied in Favero, Flabbi and Giavazzi (1999). In Germany, Italy and Spain monetary policy operates via a lending channel in an important segment of the market ± a segment that accounts for more than 80 per cent of the total German banking industry and about half of the Italian and Spanish industry. The situation is quite different in France, where there is no evidence of a lending channel. Cross-country differences in the process of ®nancial intermediation could be the result of varying national preferences and traditions. Consequently, a consolidated, cross-border, ®nancial institution may wish to continue offering different products in different markets. Similarly, the respective roles of markets and intermediaries may be history-dependent in a way that will not allow for rapid changes. Nevertheless, the creation of cross-border suppliers of ®nancial services, at a time when European consumers and ®rms are likely to become more similar, would plausibly result in a homogenisation of ®nancial practices across EMU. The current consolidation of the European banking industry appears, however, to be moving in the opposite direction. It is instructive to observe what is happening in the light of the parallel experience in the United States. Despite the massive consolidation which has occurred, concentration at the local level has, if anything, decreased. Table C2.1 shows the Her®ndahl index of the concentration of local markets for bank deposits in the United States: consolidation has been accompanied by no signi®cant change in concentration.
296 Francesco Giavazzi Table C2.1 Measures of concentration in US banking markets, 1988 and 1997 Per cent of total assets of Her®ndahl concentration indexa domestic banks held by Metropolitan Statistical Areas Non-MSA the top eight banks (MSA) counties counties 1988 1997
22.3 35.5
2020 1949
4316 4414
Notes: a The deposit Her®ndahl index is 10 000 times the sum of squared market shares based on deposits of banks operating in MSA and non-MSA counties Source: Berger, Demsetz and Strahan (1998), reproduced in Danthine et al. (1999).
Table C2.2 documents the characteristics of the consolidation which has so far occurred in Europe. Unlike in the United States, most European banking deals (half of all mergers or acquisitions in 1997) have involved institutions based in the same country. Cross border activity has been limited to deals involving a bank and a non-bank ®nancial institution (NBFI), mostly an investment bank, an insurance company or an asset manager. While cross-border deals are motivated by the search for experience in corporate ®nance and asset management ± skills that are in scarce supply in continental European banks ± domestic deals are mostly driven by the search for size. There are two reasons to be concerned. Competition is the ®rst. European banks have a natural tendency to consolidate within national boundaries, leading to industry concentration ratios much above those observed in the United States. This is because of potential cost-cutting, culture and trust ± and, indeed, the quest for market power at a time of insecurity and change: sheep get closer together when in danger. In commercial banking, diversi®cation gains explain the success of interstate consolidation in the United States. The anaemia of the equivalent cross-border merger and acquisitions (M&A) business in Europe is worrisome: while it can be explained by the fact that a good deal of the gains from diversi®cation can be obtained within the borders of individual European states, it also generates the concern that European commercial banks will want to reach the higher minimum size in their business simply by acquiring or merging with their national competitors. The Table C2.2 Bank acquisitions in Europe, 1993±7 (value of all deals, US$ billion)
Total Of which: ± Domestic bank/bank ± Cross-border bank/bank ± Bank/non-bank
1993
1995
1997
19
40
122
9 1 9
24 8 8
60 7 55
Source: Goldman Sachs, reproduced in Danthine et al. (1999).
Concluding Remarks 297
observed tendency to in-country consolidation is a challenge for competition authorities as it is likely to reinforce local monopoly power. This is particularly important for small-®rm lending, as large ®rms will access the Euro capital markets directly, while consumers will have the option of turning to specialised asset managers and to direct banking. Second are the consequences for the monetary transmission mechanism. The creation of cross-border suppliers of ®nancial services, at a time when European consumers and ®rms are likely to become more similar, would plausibly result in a homogenisation of ®nancial practices across EMU. One would expect, for instance, that the ability of medium-sized Italian ®rms to make use of the opportunities offered by the emerging Euro-wide ®nancial markets will be improved if Deutsche Bank were to take over Banca Commerciale Italiana and the clients of Banca Commerciale could bene®t from the universal banking experience of Deutsche Bank±Bankers Trust, as compared with the alternative of a local merger between Banca Commerciale and Banca Intesa. In the former case, there is reason to believe that traditional banking relationships will be reconsidered and aligned on the best German and European practices in a way that would not be promoted by a pure domestic consolidation. References Berger, A. N., R. S. Demsetz and P. E. Strahan (1998) `The Consolidation of the Financial Services Industry: Causes, Consequences, and Implications for the Future', Board of Governors of the Federal Reserve System, November, mimeo. Bernanke, B. S. and M. Gertler (1995) `Inside the Black Box: The Credit Channel of Monetary Transmission Mechanism', Journal of Economic Perspectives, 9, 27±48. Danthine, J. P., F. Giavazzi, E. L. von Thadden and X. Vives (1999) `A Brave New World: European Banking After EMU', London, CEPR. Favero, C. A., L. Flabbi and F. Giavazzi (1999) `The Credit Channel and (A)Symmetries in the Monetary Policy Transmission Mechanism in Europe: Evidence from Banks' Balance Sheets', Milan, Boconi University, mimeo. Kashyap A. K. and J. C. Stein (1997) `What do a Million Banks Have to Say about the Transmission of Monetary Policy?, NBER Working Paper, 6056. Kashyap A. K., J. C. Stein and D. W. Wilcox (1993) `Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance', American Economic Review, 83, 78±98.
Claes Berg1
The connection between a country's legal structure, ®nancial structure and monetary transmission mechanism made in Chapter 5 in this volume by Stephen Cecchetti leads to the intriguing conclusion that without legal harmonisation the ®nancial structure will remain heterogeneous across EMU member states, and because of this the monetary transmission mechanism will remain asymmetric within the EMU. Cecchetti has made an attempt at quantitatively substantiating this conclusion. The evidence can, however, be viewed only as tentative, as is also clearly stated in Chapter 5; the comments by Neumann and Angeloni showed that some additional research on re®ning some of the measures used would be useful. The crucial issue, of course, is to what extent the asymmetry in the EMU's transmission mechanism is a problem. I basically agree with what Angeloni has noted ± namely, that other parts of the transmission mechanism, such as the interest rate channel and the exchange rate channel, will become more harmonised in the Euro Area. It is also important to emphasise what Philippe Moutot said, in commenting on Weber's Chapter 4 in this volume, that the security settlement system has been harmonised in Europe and also the system for accepting collateral for the whole Eurosystem. This may have repercussions on the way the legal system will affect the monetary policy process of the ECB. However, as Reinhard Schmidt (Chapter 6 in this volume) and Otmar Issing (p. 283±92) have both pointed out, these countries are no longer small open economies and their response functions will probably have changed as a result. Being a central banker it is encouraging to hear from Bennet McCallum, in Chapter 1 in this volume, that it is systematic behaviour that matters rather than unforeseen monetary shocks. McCallum made this point in a simple three-equation model without optimising behaviour. However, this point can easily be made in a simple optimising dynamic general equilibrium model, where the only real effect of monetary policy is through the in¯ation tax. The effects on welfare of the in¯ation tax can be substantial when the other distortionary taxes are also modelled. Then adding stochastic shocks to the monetary policy rule have minor effects on welfare (see Jonsson and Klein, 1997).2 In Michael Burda's Chapter 7, it is claimed that nominal prices will be more rigid because the sheltered sector of Europe as a whole is relatively larger than in individual countries where external in¯uences on price-setting have been important. On the other hand the new European market is much larger in size than any regional sheltered market, which should improve competition and ¯exibility in pricing in some goods and service markets. In sum, it is dif®cult to 298
Concluding Remarks 299
give a ®rm conclusion whether ¯exibility of nominal prices will improve or not in EMU. The view that institutions are endogenous, and will be affected by the introduction of EMU, seems reasonable in the sense that more ¯exibility in real wages could be expected over time. However, as the comments by Brouwer and Campos e Cunha showed, it seems that the analysis to some extent underestimated the length of the adjustment process. It is reasonable to believe that reforms of labour market regulations and other safety nets will meet political resistance, at least in some countries. The tendency towards less union power could cause reactions. Furthermore, in the area of ®scal policy coordination it will take some time before a new equilibrium will be reached. I will therefore say some words on a somewhat neglected issue ± the possibility of using 'internal devaluations' as a substitute for monetary policy when countries are subject to asymmetric shocks in EMU. Variations in payroll taxes paid by employees could be used to reduce (or increase) production costs without affecting the overall ®scal balance if other taxes or government expenditures were changed at the same time. In a recession, relative unit labour costs can be depreciated through a cut in the payroll tax rate. This could be neutralised (in the budget) by a rise in employee contributions to the social security system, in income taxes or in VAT. In Finland, central labour market organisations have concluded an agreement on so-called 'buffer funds' to help ®nance unemployment insurance and pensions. The idea is to build up such funds in good times and draw on them in recessions and thus avoiding raising employer contributions. The motivation is thus to allow some ¯exibility in wage costs without having to reduce actual nominal wages (see Calmfors, 1998). Let me now give some general remarks on monetary policy and the ECB strategy. In recent years a growing consensus has emerged for price stability as the overriding, long-run goal for monetary policy. A nominal anchor is a constraint on the value of domestic money and in some form it is a necessary element in a successful monetary regime. A public price stability target thus can serve as a nominal anchor and coordinate in¯ation expectations.3 This provides the central bank with a direct link to households and ®rms expectations, which goes beyond the traditional interest rate and exchange rate channel of monetary policy. An important consequence of the convergence of in¯ation rates in Europe during the 1990s has probably been its effect on in¯ation expectations. In practice, price stability targeting is a package including the following components: ®rst, announcement of a numerical price stability target which anchors expectations; second, speci®cation of the target horizon for ful®lment of the target; third, a clari®cation of the way in which the central bank intends to meet the target on the relevant horizon; fourth, announcement of any escape clauses that may warrant departures from the general rule; ®fth, communication of the strategy to the public using monthly or quarterly reports, speeches, etc.
300 Claes Berg
When the central bank discusses how it intends to meet the price stability target there are several different ways to do it. First, the central bank could publish the in¯ation forecast for the target horizon, given an unchanged monetary policy stance, and then adjust the interest rate accordingly. This is done by in¯ation targeting central banks in New Zealand, the United Kingdom and Sweden. The second and most common way is to regularly publish a report which discusses price developments, indicators of economic activity and in¯ation and in¯ation expectations, in a more general way, but not publish a speci®c in¯ation forecast. This is done by many central banks around the world, including the ECB.4 In my view, the in¯ation forecast is a crucial input in monetary policy decisions and will always be very important for a monetary policy directed at price stability. By publishing the forecast, the central bank will increase the understanding of monetary policy. It will also contribute to increased transparency and accountability, as the public at large will be able to assess whether the stance of monetary policy is appropriate or not.5 However, the ECB is a new central bank and I guess that it will take time and experience to organise working routines. I expect that the Eurosystem will develop its methods in order to be prepared to publish explicit forecasts some time in the future.6 As was pointed out in Chapter 2 in this volume by Lars Svensson, the formulation of the monetary policy strategy of the Eurosystem could be clari®ed. Given the presentation of the reference value for M3 for 1999, it seems that the implicit price norm is 1±2 per cent. It could be announced that the ECB aims at a point estimate (for example, 1.5 per cent). In any case, a symmetric in¯ation target, either a point target or a target range with equal emphasis on the upper and lower bound, has some obvious advantages in order to anchor expectations. Another issue in relation to the ECB strategy ± which can be discussed ± is the role of gradualism versus activism in interest rate setting. It is widely accepted that policy makers facing uncertainty about the structure of the economy should be more cautious when implementing policy. This is the celebrated conservatism principle due to Brainard (1967). When there is uncertainty about some structural parameters the policy response usually becomes more cautious. In Svensson's Chapter 2, this no longer is always true. When there is non-additive uncertainty and certainty equivalence no longer holds, some covariance patterns for the parameters may make the optimal response more sensitive when uncertainty increases. However, Svensson is not very speci®c, in Chapter 2 at least, for which variables this may be the case in practice. One possibility that comes to mind is when there is uncertainty about the persistence of in¯ation. Then, somewhat paradoxically, it may be optimal for the central bank to respond more aggressively to shocks than if the parameters were known with certainty. With persistence in in¯ation, a too small policy
Concluding Remarks 301
response this quarter will lead to deviations of the target not only next quarter but also in forthcoming quarters. Therefore it may be optimal to pursue a more aggressive monetary policy this quarter when the economy is hit with a shock, È derstro in order to avoid bad outcomes in several periods (see So È m, 1999). Another possibility is when policy makers are uncertain about the state of the economy and learn from the economy's reaction to policy. When the private sector anticipates systematic attempts to incorporate this information into policy, modest interest rate changes may prove ineffective. In a recession, gradual policy initatives may elicit very little reaction. Because small interest rate cuts are unlikely to end the recession, ®rms and consumers feel safe waiting for rates to fall again before considering investing/consuming. A vicious circle may develop in which the expectation that the policy could fail leads investors/consumers to delay investment/consumption, thereby promoting failure (see Caplin and Leahy, 1996). This type of reasoning may have some bearing on the situation in part of the Euro Area, for example in Germany, although nominal interest rates are very low ± the IFO business expectations and business climate index has been falling since the end of 1997. Notes 1. I would like to thank Magnus Jonsson, Yngve Lindh, Peter Sellin and Lars Svensson for their comments. The views expressed are solely the responsibility of the author and should not be interpreted as re¯ecting the views of the Executive Board of Sveriges Riksbank. 2. Regarding the second point, the RBC (or SDGE, stochastic dynamic general equilibrium) literature have built sound theoretical models and confronted simulated moments from these models with the data for many years. Even formal statistical tests have been developed ± for example, SMM. McCallum's Chapter 1 in this volume gives support for and follows this modelling tradition. Given this, one question naturally arises (which is not discussed in Chapter 1): why is the central bank not optimising? There seem to be two dominating strategies for modelling monetary policy at the moment. On one hand, we have the optimising dynamic general equilibrium approach where monetary policy usually only follows an (ad hoc) monetary policy rule. On the other, we have a class of models where the central bank is optimising but the agents' behaviour is not modelled in any detail. It would be interesting to see a combination of these two modelling strategies. That is, models where the central bank is optimising welfare, taking into account its effects on the agents' behaviour, and at the same time, agents are optimising given the central banks' behaviour. Of course, this is not an easy task since it involves modelling dynamic games. But, nevertheless, it would most certainly give new and interesting insights on how optimal monetary policy should be conducted. 3. A nominal anchor can be implemented in order to avoid both in¯ation and de¯ation (see Berg and Jonung, 1999) for an account of the early Swedish experience, 1931±7. 4. Another possibility would be to announce the monetary policy reaction function, the policy rule or instrument rule, the central bank intends to follow. The rule
302 Claes Berg should explain how the interest rate should react, given certain deviations of, for example, actual in¯ation from target, deviations of actual output from potential output and changes in real exchange rates. However, at least to my knowledge, no central bank has announced such a reaction function. The reason, I guess, is that there may be situations when monetary policy will have to deviate from the announced reaction function. 5. An interesting task for future research is the following: how should the forecast be aggregated over members when there is an executive board or a governing council? Voting over forecasts may be very dif®cult. This problem has aspects related to the problem ®rst discussed by Condorcet in the 18th century. In modern terms, it is often referred to as the 'jury problem': to decide whether the accused is guilty or not requires con¯ating the opinion of several experts, with varying competence, into a single judgement. In practice, central banks ± be them in¯ation targeters or monetary targeters ± solve this problem. But how is it done? 6. In the Swedish case, in¯ation forecasts have been gradually introduced (see Berg, 1999). The implementation and communication of monetary policy since 1993 can be divided into three phases. In the ®rst phase, 1993±5, the in¯ation target strategy was announced and established. However, during this period, in¯ation forecasts were not published by the central bank; in the reports, the risks for future in¯ation were stated in a more general way. In the second phase, 1996±7, an in¯ation-forecast targeting was introduced. The Riksbank's own in¯ation forecasts were given more weight in the communication of monetary policy. Forecasts for future in¯ation were gradually introduced. In the third phase, from 1998, 'distribution forecast targeting' (see Svensson's Chapter 2 in this volume) was introduced and explicit paths for future in¯ation were published, surrounded by uncertainty intervals.
References Berg, C. (1999) `In¯ation Forecast Targeting: the Swedish Experience', paper presented at the seminar on 'In¯ation Targeting' in Rio de Janeiro, May, Sveriges Riksbank, Working Paper. Berg, C. and L. Jonung (1999) `Pioneering Price Level Targeting: The Swedish Experience 1931±37,' Journal of Monetary Economics, 43 (3), 525±51. Brainard, William (1967) `Uncertainty and the Effectiveness of Policy', American Economic Review, 57, Papers and Proceedings, 411±425. Calmfors, L. (1998) `Macroeconomic Policy, Wage Setting and Employment ± What Difference Does the EMU Make?' Institute for International Economics Studies, Stockholm University, Seminar Paper, 657. Caplin, A. and J. Leahy (1996) `Monetary Policy as a Process of Search', American Economic Review, 86 (4), 689±702. Jonsson, M. and P. Klein (1997) `Tax Distortions in Sweden and the United States', in M. Jonsson, Studies in Business Cycles (IIES, Stockholm University, Monograph Series, No. 34). È derstro
So È m, U. (1999) `Monetary Policy with Uncertain Parameters', Sveriges Riksbank, Working Paper.
Ignazio Visco
1 Introduction In spring 1998 many commentators would have predicted, in some cases almost as a fait accompli, that the launch of the Euro would be beset with problems. It was nonetheless considered that economic prospects for the Euro Area in 1999 were highly favourable, with predictions of solid growth, low and steady in¯ation and reduced levels of unemployment. In short, few imbalances were evident and prospects were rather bright. Indeed, growth was expected to slow down signi®cantly in the United States, in connection with the crisis in the emerging economies and the increasing stress in the ®nancial markets. In the fall, while some reduction in growth was also seen as a possibility in the Euro Area, the overall prospects were still considered to be rather positive. This led the OECD to recommend in its December 1998 Economic Outlook a cautious easing of monetary conditions in the Euro Area, and a sharper reduction of policy interest rates in the United States distributed over 1999. The implications for the Euro were quite obvious, ruling out, under these conditions, the case for its weakening. With the bene®t of hindsight, the launch of the Euro was accomplished without any major hitches, but there have been mounting concerns that prospects in the Euro Area have deteriorated. If anything, the Euro has been weaker than expected. The decision by the Federal Reserve to lower interest rates in three discrete steps of a 14 point between end-September and midNovember 1998 had the unexpected effect of bringing back substantial buoyancy in the US stock market. Furthermore, the strength of the US economy continued to be remarkable in the last quarter of 1998, leading many to consider a possible tightening of monetary conditions as a new possibility. At the same time, the substantial deterioration of business con®dence in the Euro Area and the associated sharp deceleration in current and expected economic activity revived the calls for further cuts in interest rates by the ECB, notwithstanding the non-negligible easing associated with the completion of the process of monetary uni®cation. Such a characterisation provides an apposite reminder of the hazards of foretelling even the near-term future. I would add that equally evident are the large uncertainties which surround the monetary policy transmission mechanism in the Euro Area and, as a most important consequence, the basis for clear-cut monetary policy decisions. I will not pretend that I have the answers to the many challenges which confront monetary policy makers, but rather raise certain issues and questions which are pertinent to a better understanding of how monetary policy operates in the Euro Area. My remarks 303
304 Ignazio Visco
will be structured around three themes. First, I shall give a brief summary of current economic conditions and prospects in the Euro Area. Then I shall sketch out what I see as some of the key issues and constraints which confront the design of macroeconomic policy and monetary policy in particular and, ®nally, highlight some of the key questions concerning monetary policy in the Euro Area.
2
Recent developments and prospects in the Euro Area
The Euro Area experienced a slowdown in growth towards the end of 1998. On a year-on-year basis GDP growth eased to 2.4 per cent in the ®nal quarter of 1998 (with almost no growth with respect to the third quarter) compared with about 4 per cent in the ®rst quarter and a solid 3 per cent in the central part of the year. Leading indicators suggest that activity remained sluggish in the ®rst half of 1999. The slowdown in activity primarily re¯ected weaker external demand, associated with recession in many emerging markets and Japan, and low business investment. While still substantially higher than during the trough of the 1993 recession, utilised capacity was declining and, while consumer sentiments remain sustained, business con®dence was rapidly deteriorating through the second half of 1998. However, weaker growth was not common to all countries in the Euro Area. In 1998 France recorded its best GDP growth performance since 1989 and periphery countries such as Ireland and Finland continued to grow at a pace close to double that for the Euro Area as a whole. Divergent in¯ation rates within the Euro Area have also surfaced. In¯ation measured by Eurostat, the Harmonised Index of Consumer Prices (HICP) is less than 0.5 per cent in France, Germany and Austria and above 2 per cent in Ireland and Portugal. The gap between the highest and lowest annual in¯ation rate in January 1999 rose to 2.3 percentage points,1 compared with 1.4 percentage points a year earlier. One might ask whether a disparity of this size might pose a signi®cant problem for setting monetary policy. While some persistent differences in in¯ation rates are plausible to the extent that they re¯ect a catch-up process and short-term differentials could be signi®cant, owing to differences in the relative cyclical position or tax changes, the appraisal would certainly be more circumspect if the disparity in in¯ation were sustained as a permanent difference. Differences in regional price movements within long-established monetary unions provide a gauge of the potential divergence of in¯ation rates in the Euro Area. Large in¯ation differentials have seldom been observed in other monetary unions and when they did happen, they were very short-lived episodes. For instance, the differential between East and West Germany following uni®cation was initially almost 10 per cent per annum, but fell quickly to less than 1 per cent. In the United States, modest differentials of
Concluding Remarks 305
between 0.7 and 1.3 per cent per annum among state consumer price indices have been apparent over the last thirty years. Likewise, the average differential between the lowest and highest in¯ation rate in the major Australian cities over ten-year periods is well below 1 percentage point; the same holds true considering the differential across Spanish regions. While in¯ation differentials in economically homogeneous areas are unlikely to accumulate to large price level differences, in¯ation differentials could persist between the more and less advanced Euro Area economies. As is well known, productivity differentials between traded and non-traded goods might not only generate sectoral in¯ation differentials, but differences in in¯ation rates between countries at different stages of development. One might consider this to be a possibility even for the Euro Area: indeed GDP per capita ranges between 12 000 Euros in Portugal and 29 700 Euros in Luxembourg (17 400 Euros for the area as a whole), even if the variation is much less if one adjusts for PPP differences. Other indicators of living standards ± such as the number of doctors, passenger cars, telephones and television sets per 1000 inhabitants ± also vary to a non-negligible extent. While some evidence of divergent economic conditions and disparate levels of development is apparent across the Euro Area, much progress towards structural and cyclical convergence has been achieved. Indeed, among the largest three economies in the Euro Area ± Germany, France and Italy ± which account for some three-quarters of total output, the level of economic development and living standards are very similar. The Maastricht criteria also suggest that closer convergence has been obtained over more recent years. Actually one of the most striking developments has been the rapid pace at which Euro Area countries reduced their ®scal de®cits and accomplished the nominal convergence criteria (in¯ation and interest rates) required by 1997 (the reference year). This achievement was con®rmed in 1998. Indeed, in spring 1999 only Ireland, the Netherlands and Portugal (accounting for less than 10 per cent of the Euro Area GDP) have in¯ation rates slightly higher than 2 per cent. With Finland, they are the only Euro Area countries with a positive and growing output gap. In general terms, a monetary union is less susceptible to asymmetric disturbances the more its regions are integrated with each other and diversi®ed within themselves. Apart from the Maastricht convergence criteria prior to the launch of the Euro, there are three key forces that promote closer integration. These are the degree of trade interdependence, the degree of intra-industry trade and closer income linkages, such as increased foreign direct investment or closer ®nancial market interactions. Over the past twenty years each has risen sharply and promoted closer synchronisation of business cycles across Euro Area countries.2 This process of integration is expected to continue in the years to come, and is possibly accelerated by the introduction of the Euro.
306 Ignazio Visco
3
Some monetary policy challenges and issues
Although disparities at the periphery of the monetary area raise a number of issues, the main problem in the current conjuncture is the risk of a signi®cant slowdown in the area at large. The combination of slower growth in economic activity and weaker prospects, falling in¯ation and the constraints on expansionary ®scal policy to absorb slack in the Euro economy have led to calls for an easier stance of monetary policy. Most importantly, it is necessary to raise the Euro economy's adaptability, and reduce its susceptibility to shocks and policies with uneven impacts across the monetary area. This will require further progress in implementing structural reforms, particularly enhancing the ¯exibility of labour markets. The calls for a softer monetary policy stance are (obviously) not uniformly accepted. The main arguments against a reduction in the policy interest rate include actual monetary conditions, which some already consider expansionary. Many Euro Area countries experienced fairly sharp declines as shortterm interest rates converged in the second half of 1998, with falls amounting to a 75 basis-points reduction for the area, on average. Yields at the longer end of the maturity spectrum also fell sharply, to around 4 per cent in 1999 compared with 512 per cent a year earlier. Overall, real interest rates, both short and long, were substantially lower than in the ®rst half of the 1990s (between 2 and 3 per cent now, compared with levels between 6 and 7, if not more, in the early 1990s). Furthermore the Euro real (and nominal) effective exchange rate depreciated by just over 5 per cent between October 1998 and March 1999. Moreover, monetary and credit aggregates are growing solidly. The lagged effects of these conditions should lend some support to domestic demand. Based on OECD simulations of the Interlink model, the 5 per cent depreciation of the Euro could raise the level of GDP by about 12 a per cent and only marginally boost in¯ation after two years. It should be borne in mind, however, that the level of the Euro in real effective terms is little changed from its level at the beginning of 1998 until spring 1999. Furthermore, other indicators of the overall stance of monetary policy suggest some scope for easing. The output gap in the Euro Area has not changed much since 1995 and is unlikely to do so in 2000±2001. Indeed, the low and fragile levels of business con®dence could signi®cantly tilt growth prospects downwards and result in greater spare capacity and further downward pressure on output prices. Moreover, prospects beyond the Euro Area are expected, on the whole, to remain relatively sluggish, which should limit external in¯ationary pressures. A simple relation between EU-wide in¯ation, the output gap and import price in¯ation suggests that pressures on prices are currently more on the downside than the upside. But how much faith can we place on all these indicators of economic and monetary conditions? We are inexperienced in analysing and interpreting Euro Area data. In particular, past observed relations may have changed (or are
Concluding Remarks 307
changing) with the launch of the Euro. Moreover, few historical series exist on which to base fairly conclusive analysis. To give an example, how accurate is the claim that real interest rates are near historical low levels in the Euro Area? This appears to be the case when real rates are measured based on past movements in the consumer price index (CPI), but quite a different picture emerges when alternative de¯ators or in¯ation expectation measures are used. For example, the short-term (long-term) real rate for the Euro Area using headline HICP in¯ation is 2.3 (3.2) per cent. But, if a business investment de¯ator were adopted, the measured real rate might be up to a full percentage point higher (and even more, if producer prices were used). On the other hand, excluding commodity and energy prices from headline in¯ation might lead to measures of short-term real interest rates below 2 per cent. In addition, it is well known that all CPIs have some inescapable positive bias. In the United States estimates of the bias range between 0.5 and 1.0 per cent per annum. The design and compilation of the HICP index, used by the European Central Bank (ECB) to monitor in¯ation, attempts to limit the source of known biases, but it is still believed to be positive. This implies that the level of in¯ation at 0.8 per cent may in fact be closer to the lower end of the range implicit in the ECB's interpretation of price stability, and this would raise important questions concerning the stringency and symmetry of the ECB's monetary policy objective. While monetary policy in a single currency area, by de®nition, can respond only to area-wide in¯ation pressures, sovereign ®scal policy can respond to limit the contractionary or expansionary impacts of regional shocks. This response is, however, limited by the provisions of the Stability and Growth Pact. Based on a variety of techniques, further progress on budget consolidation is required in order to avoid `excessive' de®cits and to provide adequate scope for ®scal stabilisers to fully operate in the event of a slowdown in economic activity. Most estimates suggest that a ®scal position on average close to balance would be necessary to this end. This compares with structural general government ®nancial balances that still exceed, for the average of the Euro Area, 1.5 per cent of GDP. Fiscal policy plans over the ensuing two to three years are targeting declines in de®cits (from 2.3 per cent of GDP in 1998, to about 1 per cent of GDP). This is in line with the principles of the Stability and Growth Pact. Furthermore, most countries are still near or exceed the limit of the general government debt to GDP ratio of 60 per cent speci®ed in the Maastricht Treaty. Although the ®scal plans embody tight controls on public spending, they are also based on somewhat optimistic growth projections and in some cases `backloaded'. A key risk to their realisation is a larger than expected and prolonged slowdown in activity, which constrains automatic stabilisers from operating without breaching the Pact's 3 per cent de®cit to GDP limit. Such a scenario, however, would imply substantially lower GDP growth. OECD simulations suggest that a 1 percentage-point reduction in output growth would result in an
308 Ignazio Visco
approximately 12 percentage point increase in the de®cit to GDP ratio. On this basis, GDP growth in the Euro Area would need to be almost 2 percentage points lower than the projections contained in the December 1998 OECD Economic Outlook. Scenarios of this kind still appear somewhat unlikely. But some Euro Area countries, notably the three large economies, have de®cit levels in 1998 closer to 3 per cent, and thus there is clearly little room for discretionary ®scal policy. And the risk that relatively depressed levels of economic activity may persist for some time, in the face of further declining levels of business con®dence and insuf®cient ®xed investment, calls for attention on the part of policy makers. Ultimately the issue consists in how well policies will respond to limit these downside risks. Recourse to monetary policy, obviously, cannot be expected to serve as a panacea for all problems ± and, in particular, asymmetric disturbances. Looking at recent experience, the reduction in con®dence (and associated low investment levels and growth prospects) cannot be attributed to a restrictive monetary policy stance. Given that use of ®scal policy is rather limited, structural reforms need to be pursued so that adjustment can take place without putting undue strain on economic activity. It is in particular important to promote greater overall labour market ¯exibility that would support job mobility and accelerate the pace at which real wage adjustments take place. Euro Area countries will undoubtedly continue to experience economic disturbances that affect a single country or sub-sets of regions. Such country- or region-speci®c supply and demand disturbances could be sources of tension within EMU unless they can be easily absorbed. It is in all parties interest to avoid such tensions. One approach towards this end is to strengthen the coordination of policies (structural, monetary and ®scal). A rapid and transparent discussion between the monetary authorities and the Euro Area governments on the appropriate policy mix ± inclusive of the structural reform requirements ± while preserving the full independence of the ECB would be a valuable step forward.
4
Conclusions
These remarks highlight the many uncertainties and much uncharted water ahead concerning the framing of monetary policy in the Euro Area. To complicate matters further we are still in a learning process about the transmission channels and the effectiveness of monetary policy in the new monetary area. Answers to, or at least a better and deeper understanding of, the following questions might help reduce some of these uncertainties: 1. How effective would a small interest rate cut be in the current context? In particular, what can be learned from US experience ± that is, what can be said about the effects on economic activity stemming from con®dence, expectations and asset price movements?
Concluding Remarks 309
2. Is the transmission mechanism different when sizeable slack exists in an economy? Or, more speci®cally, is there a range over which the use of monetary policy has no direct effect on prices, so that interest rates can be reduced without in¯ationary consequences and with positive effects on the real economy? 3. How does one interpret monetary conditions and what are the implications for setting policy? With respect to the ®rst point, what would seem the most important is the need to avoid possible cuts in interest rates ending up reducing rather than improving con®dence. This calls for improved policy coordination efforts which would also be helpful on the communication side: this would prevent the perception that the easing of monetary policy would either re¯ect the fact that the ECB had given up on the expectation that governments were ready to act rapidly and effectively on the structural front, or signal a state of current and prospective economic conditions much worse than expected. The second point is, perhaps, even more dif®cult to deal with. This is simply because there is still not much information on how the Euro Area economy operates ± i.e. how economic decisions might differ under a common monetary policy compared with the previous state of affairs. What might be relevant here is probably the need to avoid the ECB's concerns about price stability being perceived as biased towards less favourable monetary conditions than would be needed in a situation of substantial current and prospective slack in the Euro Area as a whole. To this end, the (perhaps evolving) identi®cation of a lower bound for the de®nition of price stability might be helpful (at least operationally). Moreover, it is important that the ECB should not be perceived as feeling equally at ease with price changes permanently situated at the bottom and at the centre of the range. Finally, while many measurement problems must still be sorted out, both with respect to price indices and monetary aggregates, a wide spectrum of interest rates should be considered. In fact, one can always expect to ®nd con¯icting signals concerning the stance of monetary conditions, depending on the indicators used. In addition the Euro Area has virtually no historical perspective against which benchmarks can be established. Consideration should then also be ± and undoubtedly will be ± given not only to the levels and spreads prevailing in the Euro Area member countries in the course of the 1990s, but also to those prevailing in other economies at different points of the cycle, as well as over a longer time perspective. Notes 1. Portugal has the highest in¯ation rate at 2.5 per cent and Luxembourg the lowest at ±1.4 per cent. The reported differential, however, excludes Luxembourg since the rate in January has been in¯uenced by a one-off effect almost entirely owing to the inclusion of the January discount sales in the price index for the ®rst time.
310 Ignazio Visco 2. The OECD has just published an in-depth assessment of the challenges facing the European Single Currency Area, including an evaluation of the forces promoting closer economic integration (OECD, 1999).
Reference OECD (1999) EMU: Facts, Challenges and Policies, Paris, OECD.
Manfred J. M. Neumann
I will brie¯y take up two issues: (1) What type of ®nancial sector structure is preferable? (2) Should the European Central Bank (ECB) be concerned about differences in regional or national transmission mechanisms?
1 What type of ®nancial sector structure is preferable? Structure of the ®nancial sector A central theme of this conference has been that the effectiveness of monetary policy, as regards impact as well as speed of transmission to the real economy, depends on the structure of the economies' ®nancial sectors. Though the available empirical evidence is not overwhelming, most economists ± and certainly the supporters of the `credit view' ± share the opinion that monetary policy is transmitted to the real sector quicker and more strongly if an economy's ®nancing relations are dominated by a large and not too healthy banking sector. Thus, it seems that, from the monetary control point of view, bank dominance is the preferable state. While this is a probably important message for central banks, provided they desire to contribute to cyclical stabilisation, it is not obvious that bank dominance is equally preferable from the long-run view of economic development. The long-run success of an economy as regards real growth and the creation of jobs depends, apart from other factors, on the ¯ow of innovation. From this more long-term point of view one might question the value of bank dominance. As a rule, banks are relatively risk averse institutions, as they have to be in the interest of depositors. As a result, banks are not too keen on ®nancing young, innovative companies. The only asset of a young enterprise in most cases is just a new business idea. Given the lack of assets that could be used as collateral for bank loans, the lack of pro®ts during the ®rst years of operation, the extremely high uncertainty about the success of a new company's attempt at establishing a market for the product innovation, young companies ®nd it dif®cult to attract the ®nancial means for rapid expansion if the ®nancial markets are dominated by commercial banks. As a rule, in such economies venture capital markets are underdeveloped. Thus, from the point of view that the ¯ow of innovation is an important factor for strong long-term development, it seems that a preferable ®nancial sector structure is one where large, liquid markets for the participation in venture capital and equity ®nance exist. It is instructive to compare on this account Germany with the United States. In the United States, it is easier for young companies to attract capital given that there equity and venture capital markets play a much more important 311
312 Manfred J. M. Neumann
role. This is highlighted by the following facts: (1) in Germany it takes 42 years, on average, until a ®rm is admitted to the share market by a regulatory committee that is dominated by banks; in the United States, in contrast, it takes 14 years on average; (2) in the United States market capitalisation, as a percentage of GDP, is three times of what it is in Germany; (3) the share of bank lending in total ®nance is much smaller in the United States than in Germany. In line with these differences in ®nancial structure are the differences in legal protection. In the United States the private investor in equity ®nance is better protected by more extensive rights for controlling company management than in Germany. As a corollary, creditors enjoy better protection under German than under US law. How have the ®nancial cultures come to develop so differently? A very probable explanation is the different role of universal banks. Since the 1870s universal banks have dominated the German culture of ®nance and this has in¯uenced German law while in the United States the banking reform of the early 1930s put an end to the existence of universal banks and promoted the fragmentation of the banking system. But note that we lack systematic knowledge about the relative advantages as regards the long-run impact of alternative structures of the ®nancing system. All in all, it seems that bank-dominated Germany has not fared worse than the more equity ®nance-oriented United States. While the long-term development of economies obviously does not solely depend on the quality and structure of the ®nancial system but on a host of other factors, too, it is interesting to note that the German economy has not grown less, but faster than the Anglo± Saxon economies. On average over the past ®fty years, real per capita GDP has grown by 3.5 per cent in Germany but by about 2 percent in the United States and in the United Kingdom. Impact of EMU on the convergence of national transmission mechanisms The launch of the Euro will speed up the process of ®nancial market integration in Europe and this, in turn, will promote the convergence of the transmission mechanisms. As the Euro promotes the establishment of uniform market standards in the Euro Area, the national ®nancial markets will rapidly integrate into a larger market that will be able to provide a wider maturity range of instruments and an increase in market liquidity. This, in turn, is likely to induce domestic and foreign institutions to expand the issuance and the investment in Euro instruments. A Euro-wide market for commercial paper is emerging and the issuance of corporate bonds is expanding. Similarly the fragmentation of equity trade is vanishing and it is very likely that the national regulation of equity issuance will be harmonised. The rising integration and size of the European ®nancial markets provides bene®ts of scale, and the increase in the degree of competition will reduce transaction costs. The national banking industries are involved in these developments and they are likely to demand that national governments harmonise banking
Concluding Remarks 313
regulation. Banks will become more similar across the Euro Area and move into new areas of ®nance. As the Banking Report of the ECB (ECB 1999) shows, banking intermediation is losing out to institutional investors, like investment funds, insurance companies and pension funds. Thus, I conclude that the transmission mechanism in the Euro Area will indeed become more similar to that of the United States. We can therefore expect that monetary policy will lose its strength to some degree in the longer run.
2 Should the ECB be concerned about differences in regional or national transmission mechanisms? The Treaty of Maastricht assigns two objectives to monetary policy. The ®rst is maintaining price stability. Provided that this objective is not endangered, the European System of Central Banks (ESCB) is expected to support the general economic policies in the Community. Though this second stipulation is open to interpretation, it probably aims at business cycle stabilisation. The ECB puts emphasis on the objective of maintaining price stability by calling it the overriding objective. With respect to this objective differences in regional or national transmission mechanisms are of no particular relevance. Consider that EMU starts from a state of price stability and let us assume that the ECB has adopted a mediumrun strategy of supplying a stable monetary expansion path ± relative to the evolution of money demand. Under those equilibrium conditions all member countries are affected to the same degree. Asymmetries in the transmission mechanisms will not matter, given that the ECB's forecast errors as regards the appropriate medium-run path are likely to be comparatively small. But note that we cannot be sure that the ECB has adopted a medium-run strategy. The ECB has announced that it will take into consideration its reference value for the growth of the money stock M3 and at the same time a host of other economic indicators. In contrast, differences in the transmission mechanism will matter if the ECB produces monetary shocks which by de®nition are unanticipated. This will become particularly relevant should the ECB try to ®ght a business downswing or upswing. The regional economies will be hit differently, implying that their relative positions as regards competitiveness in the single market will be affected. This applies especially to small and medium-sized ®rms who traditionally rely on bank ®nance. Moreover, it seems that those member economies where the bank lending channel is strongest will be hit the harder if their business cycles run counter to the average cycle showing up in the EU-11-wide aggregates. This will be a problem for the smaller members of EMU whose data enter the aggregate data with small weight. The question is: what lesson should the ECB draw from this perspective? It goes without saying that national business cycles are none of its business. Monetary policy has to be based on the EU-11 aggregates and nothing else.
314 Manfred J. M. Neumann
However, the ECB should not completely disregard differences between national cycles because the larger the differences become the more asymmetric will be the impact of its policy. In view of this I conclude that the existence of differences in the transmission mechanism lends additional support to the monetarist prescription that the ECB should avoid larger policy swings but try to run a stable monetary policy tied to the medium run. Reference ECB (1999) Possible Effect of EMU on the EU Banking System in the Medium to Long Term, Frankfurt/Main.
Name Index
Note: page numbers in bold refer to main contributions in this book; page references to ®gures and tables are given in italics. Cukierman, A. 265 Currie, D. 72
Agell, J. 266 Akerlof, G. A. 67, 93±4 n14, 115, 259 Andersen, T. M. 28 Angeloni, I. 201±7, 298
Danthine, J.-P. 172, 265 de Bondt, G. J. 181 DeFina, R. 285 Dickens, W. T. 67, 93±4 n14, 115 Dohse, D. 257, 267 Dornbusch, R. 172, 181, 209, 257 Dotsey, M. 38 Drif®ll, J. 264, 265
Ball, L. 65, 66 Bank for International Settlements 1, 206, 249 Barro, R. 265 Batini, N. 77 Bean, C. 266 Benhabib, J. 115 Berg, C. 8, 298±302 Bernanke, B. S. 11, 31, 33, 38, 78, 132, 175 Blanchard, O. 256 Blinder, A. S. 51, 72, 175 Brainard, W. 85, 300 Brouwer, H. J. 7, 276±8, 299 Brunner, K. 115 Bruno, M. 255 Burda, M. C. 7, 252±73, 276±8, 279±82, 298 Buti, M. 257
Ehrmann, M. 183 Eichenbaum, M. 28, 31, 33, 47, 175 Eichengreen, B. 257 Eliasson, A.-C. 27 Evans, C. L. 28, 31, 33, 175 Favero, C. 159±63, 172, 181, 209, 257, 295 Feldstein, M. 257 Fisher, M. E. 140 Flabbi, L. 295 Friedman, M. 255, 283, 291 Fuhrer, J. C. 24, 27, 28, 31, 54
Calmfors, L. 260, 264, 265 Campos e Cunha, L. 7±8, 279±82, 299 Carlino, G. 285 Cecchetti, S. G. 5, 63, 170±93, 195±200, 201±7, 242, 298 Chari, V. V. 28, 47 Christiano, L. J. 2, 28, 31, 33, 37, 44±9, 174±5 Clarida, R. 12, 48, 131, 132, 135, 137, 137t, 140 Cooley, T. F. 45 Corbett, J. 222, 223
Gali, J. 12, 48, 131, 132, 135, 137, 137t, 140 Gerlack, S. 181 Gertler, M. 11, 12, 38, 48, 131, 132, 135, 137, 137t, 140, 175 Ghironi, F. 55 Giavazzi, F. 8, 172, 181, 209, 257, 294±7 Goodfriend, M. S. 31, 168 Goodhart, C. A. E. 6, 241±6 Gordon, D. 265
Index compiled by Sue Lightfoot. 315
316 Name Index Greenspan, A. 73, 111 Gust, C. 48±9 Hackethal, A. 222, 241 Haldane, A. G. 77 Hall, R. E. 28, 64 Hansen, G. D. 45 Hansen, L. 47 Herrmann, H. 1±9 HM Treasury 66 Hunt, J. 265 Isard, P. 27 Issing, O. 8, 89, 90, 91, 283±92, 298 Jeanne, O. 267 Jenkinson, T. 222, 223 Kashyap, A. K. 170, 176, 204 Kehoe, P. J. 28 Kieler, M. 181, 288 King, M. 3, 103±6 King, M. A. 65, 88, 94 n26 King, R. G. 47, 132, 139, 140, 142, 143, 144, 154±5 Koenig, E. F. 125±6 Krieger-Boden, C. 257, 267 Kuttner, K. N. 137, 138, 138t Kydland, F. E. 47, 51 La Porta, R. et al. 172, 185, 186, 187 Lafontaine, O. 270 Laxton, D. 27 Leamer, E. 166 Leeper, E. M. 37 Levin, A. 72 Levine, P. 72 Lippi, F. 265 Lothian, J. R. 139 McCallum, B. T. 2, 11±40, 44±5, 46±9, 52, 53, 54±8, 72, 122, 138, 143, 298 McCauley, R. N. 172 McConnell, M. M. 183 McGrattan, E. R. 28 McKinnon, R. 252 Maclennan, D. J. 285 Mankiw, N. G. 259 Mayer, C. 221±2
Meltzer, A. H. 3±4, 11, 112±29 Mihov, I. 31, 33 Mishkin, F. S. 11, 78, 139 Monticelli, C. 289 Moore, G. R. 24, 27, 28, 31 Moutot, P. 4, 164±9, 298 Muellbauer, J. 285 Mundell, R. 252, 253, 279 Nelson, E. 13, 14, 16±17, 24, 28, 34 Neumann, M. J. M. 5±6, 8±9, 195±200, 298, 311±14 Obstfeld, M. 11 Organization for Economic Cooperation and Development 303, 307±8 Orphanides, A. 67, 94 n14, 115, 153 Peersman, G. 289 Perez Quiros, G. 183 Perry, G. L. 67, 93±4 n14, 115 Posen, A. S. 137, 138, 138t Prescott, E. C. 47, 51 Rich, R. W. 183 Rogoff, K. 11 Romer, C. 36±7 Romer, D. 36±7, 256, 259 Rotemberg, J. J. 13, 46, 175 Rudebusch, G. 77, 137 Saarenheimo, T. 181, 288 Sachs, J. 255 Sapir, A. 257 Sargent, T. 47 Schmidt, R. H. 6, 208±36, 241±3, 246, 247±50, 298 Schmitt-Grohe, S. 115 Schuberth, H. 183 Seater, J. J. 140 Shapiro, M. D. 37 Shiller, R. J. 164, 169n2 Sims, C. A. 37 Smets, F. 181, 289 Smets, G. 249 Stein, J. C. 170, 176, 204 Stephens, M. 285 Summers, L. H. 115, 139 Svensson, L. 2±3, 60±97, 103±6, 113, 137, 160, 300
Name Index 317 Taylor, J. B. 11, 28, 57
Tobin, J. 211
Tristani, O. 289
Uhlig, H. 2, 51±9
Uribe, M. 115
Vickers, J. 78
Vienney, A. 6±7, 247±51
Ä als, J. 3, 107±11
Vin
Visco, I. 8, 303±10
Vlaar, P. J. G. 183
Wallis, K. F. 78, 79, 96 n38
Walsh, C. E. 27
Watson, M. W. 38, 47, 132, 139, 140,
142, 143, 144, 154±5
Weber, A. A. 4, 131±56, 161,
164±8
White, W. R. 172
Wieland, V. 67, 72, 94 n14, 115
Williams, J. C. 72
Williamson, O. E. 218
Wolman, A. L. 94 n14
Woodford, M. 13, 46, 80, 132, 168
Wyplosz, C. 151
Yellen, J. L. 259
This page intentionally left blank
Subject Index
Note: page references to ®gures and tables are given in italics, e.g. 20f, 32t. channels of monetary policy 210±11 channel of relative prices 211±12 credit channel see credit channel exchange rate channel 211 interest rate channel 211, 212, 248±9 collective bargaining 264±5, 267, 277, 280 Consumer Price Index 66±7 consumption 125±7, 126±7f credit channel 125, 204±6, 212±15, 228±9, 247±8, 249, 294 balance sheet channel 213, 214, 228 bank lending channel 176, 213±14, 228, 295 cross-border competition 205 France 249, 295 Germany 227±8, 228f, 241±2, 243f, 247, 249, 295 Italy 295 Spain 295 strength 178, 181t, 186, 187, 196, 197, 198±200, 199f United Kingdom 227±8, 228f, 241, 243f, 247, 249
aggregate demand and supply 113, 124 asset price channel see channel of relative prices Bank of Canada 64, 77 Bank of England forecast targeting 76, 78±9, 82, 86, 92, 106 in¯ation targeting 66, 67, 86, 95±6 n37, 96 n38, 105, 106 bank reserves 174±5 banking industry change 206, 290±1 consolidation 294, 295±7, 296t, 312±13 cross-country differences 171±2, 180t, 214, 295 France 291, 295 Germany 219, 220, 222, 295, 312 (see also Deutsche Bundesbank) health 178, 179, 179t Italy 295 lending channel 176, 213±14, 228, 295 liability diversi®cation 204±5 size and concentration 177±8, 177t, 181t, 196±7, 197t, 198 Spain 295 United Kingdom 219, 222 (see also Bank of England) United States 295, 296t, 312 (see also Federal Reserve System (USA)) see also Bank of Canada; credit channel; European System of Central Banks (ESCB); New Zealand, Reserve Bank of; Sveriges Riksbank
de¯ation 68, 115±24 Deutsche Bundesbank 61, 106, 135, 146, 153 ECB see European System of Central Banks EMU see European Monetary Union ESCB see European System of Central Banks European Monetary Institute 67, 73 European Monetary Union (EMU) 195, 210, 303, 306 economic development 305
capital mobility 254, 254t channel of relative prices 211±12
Index compiled by Sue Lightfoot. 319
320 Subject Index European Monetary Union (continued) effects on labour market 252±3, 258±71, 276±8, 280±2
®scal policy 307±8
growth 304
in¯ation rates 304±5
integration 305
interest rates 306, 307
monetary policy 306±9
and national ®nancial systems 223±31,
249±50
and optimal currency areas 279
transmission mechanisms 284±8,
286t European System of Central Banks (ESCB)
challenges 170, 188, 230±1
credit channel 206
de®ning price stability 67, 88±9,
104±5
indicators of risks to price stability 81,
82
in¯ation 124, 259
M3 de®nition 205
maintaining price stability 89±92,
290
money 89±90, 105
monetary policy 195, 282, 289t
price stability 66, 88, 89, 104, 105,
313
strategy 283±4, 290, 291±2, 299±300,
313±14
transparency 92
exchange rates 202±4, 255, 256, 259f,
279, 291, 306
exports 204t
external ®nance premium 213
Federal Reserve System (USA) 58±9,
115±16, 122, 124, 285
history 116±24
®nancial markets 290±1, 312
see also banking industry
®nancial sectors 209, 214±15, 231t
national differences 285, 286t, 287,
311±12
®nancial systems 209±10
complementarity and
consistency 217, 218, 221f,
222, 241
corporate governance 232t
differences 215±19, 216f
®nancing patterns 217, 221±3, 222t,
224f, 233t
France 220, 222t, 223, 224f, 241, 247,
250
Germany 219, 220, 222±3, 222t, 224f,
225±9, 228f, 241, 243f
and legal structures 172±4, 173t, 185,
186, 187t, 188, 191 n5, 202
national differences 201±2, 209,
214±15, 285
non-bank ®nance 178, 179, 180t, 197,
198t, 214
reactions to common monetary
policy 188, 223±31, 249±50
sub-systems 216±18, 216f, 221±2, 247
United Kingdom 219±20, 222±3, 222t,
224f, 225±9, 228f, 241, 243f, 250
see also banking industry
Fisher effect 139±40, 166
forecast targeting see targeting
France
banking industry 291, 295
corporate sector 285
credit channel 249, 295
®nancial systems 220, 222t, 223, 224f,
241, 247, 250
interest rate channel 248
Germany banking industry 219, 220, 222, 295,
312 (see also Deutsche
Bundesbank)
credit channel 227±8, 228f, 241±2, 243f, 247, 249, 295
®nancial sector structure 311±12
®nancial systems 219, 220, 222±3,
222t, 224f, 225±9, 228f, 241, 243f
interest rate channel 227±8, 228f,
241±2, 243±5f, 246, 246t, 248
goods markets 284±5, 286t
Haberler±Pigou±Patinkin wealth
effect 124, 127
housing sector 285
in¯ation 48±9, 63, 65, 132, 134f,
259±61, 260t, 261t, 278, 304±5
costs 124
Subject Index 321 and interest rates 133±4f, 136f,
138±48, 141±2t
econometric framework 143±4, 159,
161±3, 165±6
empirical evidence 144±8, 145t, 147f,
148f, 149f
Fisher effect 139±40, 166
time series properties 140±2
in¯ation targeting see monetary policy interest rate channel 211, 212, 248±9
France 248
Germany 227±8, 228f, 241±2, 243±5f,
246, 246t, 248
United Kingdom 227±8, 228f, 242,
243±5f, 246, 246t, 248
interest rate targeting see monetary
policy
interest rates
Euro Area 306, 307
European links
empirical evidence 151±3, 152t,
167
time series properties of
differentials 148±9, 167
VAR model 149±51, 161±2, 164±5
exchange rate effects 202±4
and housing sector 285
and in¯ation 133±4f, 136f, 138±48,
141±2t econometric framework 143±4, 159, 161±3, 165±6 empirical evidence 144±8, 145t, 147f, 148f, 149f
Fisher effect 139±40, 166
time series properties 140±2
policy 131±5, 168±9, 242 (see also
monetary policy)
smoothing rules 135±8, 137t, 138t,
142, 146, 154
wealth effect 124±5
intermediate targeting see monetary
policy
internal devaluations 299
investment 53±4
Italy 295
labour markets
Euro effects on 258±71, 276±8,
280±2
institutions 265, 278, 281, 287, 299
macroeconomic impact 253±8 structure 286t, 287
labour mobility 253±4, 270, 280
labour unions 264±5, 267, 277, 280
legal systems
and impact of monetary policy 173,
173t, 186±8, 187t, 198±200, 205±6
national differences 183, 185±6, 185t,
188, 202
see also ®nancial systems and legal
structures lending channel see credit channel liquidity traps 113±15 Marshall±Hicks rule 264, 272 n15
models of monetary policy 54, 298
diagnostics 31±6, 32±3f, 32t, 34f, 35f,
36f, 54±5
limited participation model 48
maximum likelihood 47
`price puzzle' 55±6
second moments 47
simple analytical 13±15
speci®cation 19, 23±31, 23t, 25±6f,
27t, 29±30f, 54, 55
structural 15±19, 18t, 20±2f, 48±9,
53±5, 56±7f
testing 47±9
VAR systems 14±15, 33, 52±3, 54,
249±50 monetary policy
communication 62, 92, 108, 109, 111
decision framework 61±2, 72±3, 92,
103±4, 106, 108±9, 110±11
economic theory 53, 58
effects 15±19, 56±8, 170±1, 195,
208±9 Euro Area 306±9 impact on output and prices 173t, 181±3, 182f, 184t, 187t, 198±200,
199f, 206
in¯ation targeting 61, 63±6, 67±8, 88,
103, 104, 105, 115, 300±1
intermediate targeting 61, 86±8, 103,
104, 110
interest rate targeting 69±73
legal systems and 173, 173t, 186±8,
187t, 198±200, 201±2, 205±6
monetary targeting 61
price-level targeting 63±5, 68, 104
322 Subject Index monetary policy (continued) rules 51±2, 56±8, 62, 71±3, 103±4, 109±10, 168±9 tightening 288, 294 see also channels of monetary policy; ®nancial systems; forecast targeting; models of monetary policy; shocks to monetary policy; single monetary policy; transmission mechanisms monetary unions 304±5 money 89±90, 105, 294 money growth indicator 82, 90, 95 n32, 136f New Zealand, Reserve Bank of 67, 77, 92 nominal frictions 255±6 optimal currency areas 279 price rigidity 256, 256f, 258±9, 259f, 266, 281 price stability de®ning 60, 62±8, 88±9, 104±5, 107 indices and levels 66±8 instrument assumptions 76±7 loss function 62, 65±6, 68, 69, 71, 73, 78 versus low in¯ation 62±5 maintaining 60, 68±9, 89±92, 107±8, 109, 290 price-level targeting 63±5, 68, 104 risks 81±2, 90 role of indicators 79±82 targeting 299±300 price-level targeting see monetary policy product market integration 255, 265 rigidities nominal 255±7, 256f, 258±61, 266, 267, 281 real 253±7, 256f, 262, 264±5, 267 shocks to monetary policy analysis 45±6 identi®cation procedures 31 impulse responses 20±2f, 25±6f, 28, 29±30f versus systematic policy 12±15, 18±19 VAR systems 14±15, 33, 52±3
single monetary policy 188, 223±31, 249±50, 290 Spain 295 Stability and Growth Pact 307 Sveriges Riksbank forecast targeting 76, 78, 79, 82, 86, 92, 96 n38 in¯ation targeting 66 Sweden 64 see also Sveriges Riksbank targeting forecast targeting 60±1, 73±9, 109±11; distribution forecast targeting 85±6, 104; judgemental adjustments 77±8; mean, median or mode 78±9, 105±6; non-additive uncertainty 84±5; non-linearities 82±4; optimality criterion 75±6 in¯ation targeting 61, 63±6, 67±8, 88, 103, 104, 105, 115, 300±1 interest rate targeting 69±73 intermediate targeting 61, 86±8, 103, 104, 110 monetary targeting 61 price-level targeting 63±5, 68, 104 taxation 266, 299 transmission mechanisms 11, 224 asymmetry 188, 298, 313 channel of relative prices 211±12 channels classi®cation 210±15 consumption 125±7, 126±7f credit channel see credit channel econometric evidence 124±8 effects of single currency 288, 290±1 in Euro Area countries 284±8, 286t exchange rate effects 202±4 ®nancial structures 170±4, 173t, 176 (see also ®nancial systems) historical evidence 115±24 institutional features 284±8, 286t interest rate channel 125, 211, 248±9 lending channel see credit channel lending view 171, 175±6, 201 macroeconometric analyses 288 methods of analysis 1 models 68±9, 112±15, 125, 160±1 objective function 122, 124 strength 177±81, 181t
Subject Index 323 theories 170±1, 174±6
traditional money view 175
unemployment 265, 277, 281, 287, 299 United Kingdom banking industry 219, 222 (see also Bank of England) credit channel 227±8, 228f, 241, 243f, 247, 249 ®nancial systems 219±20, 222±3, 222t, 224f, 225±9, 228f, 241, 243f, 250 interest rate channel 227±8, 228f, 242, 243±5f, 246, 246t, 248 monetary policy 242, 246, 249±50
United States banking industry 295, 296t, 312 (see also Federal Reserve System) ®nancial sector structure 311±12 wage behaviour nominal wages 260, 261, 262t, 263t, 264, 266±7, 276 real wages 267, 268t, 269f, 269t, 278 and transmission process 287 wealth effect 124±5, 127, 248