Economic Integration: Limits and Prospects

Economic Integration Limits and Prospects

Edited by

George M. Korres and George C. Bitros

Economic Integration

Also by George M. Korres TECHNICAL CHANGE AND ECONOMIC GROWTH PRODUCTIVITY AND US GROWTH INDUSTRIAL ORGANISATIONS ADVANCED MACROECONOMICS

Also by George C. Bitros MICROECONOMIC THEORY

Economic Integration Limits and Prospects Edited by

George M. Korres Professor in the University of Aegean and Institute of Computational Mathematics, Crete Institute of Research and Technology

and

George C. Bitros Professor of Economics Athens University of Economics and Business

Editorial matter and selection © George M. Korres and George C. Bitros 2002 Chapters 11 and 15 © George M. Korres 2002 Chapters 1 and 16 © George C. Bitros 2002 Chapters 2–10, 12–14 and 17 © Palgrave Publishers Ltd 2002 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 W1T 4LP. 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 2002 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 0–333–72472–0 (outside North America) ISBN 0–312–22664–0 (in North America) 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 Economic integration : limits and prospects / edited by George M. Korres and George C. Bitros. p. cm. Includes bibliographical references and index. ISBN 0–312–22664–0 1. Europe—Economic integration. I. Korres, George M. II. Bitros, George. HC241 .E2927 2001 337.1’4—dc21 2001032787 10 11

9 10

8 09

7 08

6 07

5 06

4 05

3 04

2 03

1 02

Printed in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire

Contents List of Figures

vii

List of Tables

ix

List of Abbreviations

xiii

List of Contributors

xv

Preface

xvii

1 Introduction George C. Bitros PART 1

1

MACROECONOMIC ISSUES OF ECONOMIC INTEGRATION

2 A Dynamic Analysis of Structural Asymmetries in the European Union Laurence Boone

13

3 Testing for Common Features in the European Union Hong Bai, Stephen Hall and David Shepherd

45

4 The Demand for Money in the European Union: the Role of Germany's Interest Rate Yonghao Pu and George Zis

58

5 Credibility and In¯ation Expectations in the European Monetary System Nicholas Sarantis

77

6 Exchange-Rate Regimes, In¯ation and Credibility: Evidence from Greece George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos

91

PART 2

MICROECONOMIC ISSUES OF ECONOMIC INTEGRATION

7 The Effects of European Integration on Greek Manufacture Yannis Katsoulakos and Nicholas Tsounis

v

105

vi Contents

8 Economic Integration in Europe and the Pattern of German Foreign Direct Investment Florence Hubert and Nigel Pain

135

9 Spanish Unemployment ± a Hiring Function Approach: GARCH-M Model Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos

157

10 Economic Growth, Labour Effort and Fertility Choice in Greece: Evidence and Implications for Economic Integration George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou

176

PART 3 INSTITUTIONS AND POLICIES OF INTEGRATION 11 Institutional Development and the Harmonisation of Technological Policy in the European Union George M. Korres

197

12 Reform of the EU Institutional Framework for Coastal Shipping: Concepts Concerning Greek Coastal Shipping Policy George Vlachos and Maria Lekakou

217

13 Regional Inequalities and the Allocation of EU Funding: The Case of Greece Dimitrios Giannias

230

PART 4 CHALLENGES FOR AN INTEGRATED EUROPE 14 Economic Integration and the Future of the Welfare State in the European Union Eleni Paliginis

245

15 Technological Change, Productivity and Economic Integration of the European Union George M. Korres

261

16 Privatisation: Recent Experiences and Prospects in a World Without Borders George C. Bitros

280

17 The Balkans and the European Union Nicholas V. Gianaris

293

List of Figures 2.1 2.2 2.3 2.4 A2.1 A2.2 4.1 4.2 4.3 5.1 5.2 5.3 5.4 9.1 9.2 9.3 10.1 10.2 13.1 13.2 14.1 14.2 14.3 14.4 14.5

Supply shocks, Germany and France, 1963±93 26 Estimation of  (Equations 2.5 and 2.6), 1974±92 27 Supply shock differentials for Germany and the United 28 States, 1963±93 Estimated (Equations 2.5 and 2.6), 1974±94 28 Supply shocks, 1974±94 38 Demand shocks, 1974±94 39 M1 and M2 sequence of break-point Chow test, 1983(2)±91(4) 66 Dm1t, Dm2t ®tted and actual values from the 72 conditional dynamic models System stability and constancy tests 73 Domestic in¯ation rates relative to Germany's 78 in¯ation, pre-ERM Domestic in¯ation rates relative to Germany's 79 in¯ation, pre-ERM Domestic in¯ation rates relative to Germany's 79 in¯ation, ERM Domestic in¯ation rates relative to Germany's 80 in¯ation, ERM Hazard function and Spanish unemployment 165 The hazard/unemployment ratio and the Spanish 169 economy Hazard rate of unemployment 170 Live births per 1000 inhabitants, Greece, 1960±95 178 Response to one standard deviation innovations 188 The relationship between utility and regional index 234 The relationship between utility and regional indexes 235 Rates of unemployment in the EU, 1970±93 248 Part-time employment as a percentage of total 249 employment Male and female part-time employment as a 249 percentage of total employment, 1993 Single women with children 250 Total social expenditure as a percentage of GDP, 252 1970±93

vii

viii List of Figures

14.6 15.1

Total growth of social expenditure, annual percentage change, 1980±93 Per capita GDP and patenting in EU member states, 1973±90

253 267

List of Tables 2.1 2.2 2.3 2.4 2.5 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 4.1 4.2 4.3 4.4

Standard deviations of GDP growth rates in Europe, 1964±94 Supply shock convergence on Germany as opposed to the rest of the world, 1974±94 Supply shock convergence on the core as opposed to the rest of the world, 1974±94 Demand shock convergence on Germany as opposed to the rest of the world, 1974±94 Demand shock convergence towards the core as opposed to the rest of the world, 1974±94 Output growth correlation coef®cients, 1980±89 Unit root tests for quarterly GDP with respect to Germany Bivariate VAR(1) for GDP, 1980±89 The common features test for GDP growth, 1980±89 2SLS common features test for GDP growth, 1980±89 LIML common features test for GDP growth, 1980±89 Unit root tests for nominal and real exchange rates LM tests for bivariate VAR(1) features in nominal exchange rates LM tests for bivariate VAR(1) features in real exchange rates The common features test for nominal exchange rates The common features test for real exchange rates LM tests for bivariate VAR(1) features in nominal interest rates The common features test for nominal interest rates LM tests for bivariate VAR(1) features in real interest rates Residual correlations, goodness of ®t and evaluation Cointegration analysis of M1 and M2 Restricted cointegration vector * and loadings * Signi®cance of ECMs in the dynamic system of M1 and M2 ix

22 30 31 32 33 50 50 51 51 52 52 53 53 54 54 55 55 56 56 65 68 69 70

x List of Tables

4.5 5.1 5.2 5.3 5.4 6.1 7.1 7.2 8.1 8.2 8.3 8.4 9.1 10.1 10.2 10.3 10.4

11.1 11.2 11.3 11.4 11.5

Dynamic equations of European money (M1 and M2) demand function Forecasting equations for price in¯ation (®xed-parameters model) Tests for unbiasedness and ef®ciency (®xed-parameters model) Diagnostic statistics for the time-varying parameters model Tests for unbiasedness and ef®ciency (time-varying parameters model) Restricted SUR estimates of (6.12a)±(6.12b)±(6.12c) subject to (6.11), Greece, 1960±94 Patterns of change of the shares of expenditure in apparent consumption (sign of share change) Changes in the RCA, Grubel±Lloyd and specialisation indexes over the two periods Geographical and industrial distribution of German FDI stocks The determinants of German FDI: dependent variable 1n(FDI)ijt, sample period 1977±94 The impact of the internal market on FDI, by country and sector The impact of the internal market on German FDI in the EU Regression results Tests of the unit roots hypothesis Johansen±Juselius cointegration test: employment, fertility choice and real output, 1960±95 Summary causality results based on vector error-correction models of fertility choice, labour effort and output Percentage of forecast variance explained by innovations in variance decomposition of labour growth Eureka projects, 1992 EU framework research programmes, 1984±98 Framework programmes for R&TD activities: a comparison of budget share, 1984±94 Framework programmes of EU activities in R&TD, 1987±98 Fourth framework programme, 1994±98

71 84 86 86 87 97 119 124 138 148 151 152 167 182 183 185 186

200 203 204 204 205

List of Tables xi

11.6 11.7

Planned budgetary expenditures, 1992±97 Categories of EU expenditures in member states (average 1985±89) 11.8 Status of European technologies vis-aÁ-vis those of the United States and Japan 11.9 Distribution of direct action projects, 1977±80 11.10 Major European programmes promoting new technology 13.1 Results and ranking of regional indexes 13.2 Results and ranking of per capita EU funding and regional indexes 13.3 Results and ranking of actual and planned spending, 1994±5 15.1 Patterns of growth 15.2 GERD, GFCF (percentage growth rates), RDI and external patent applications 15.3 Productivity: average percentage changes at annual rate 15.4 Relationship between productivity and innovation for twelve EU member states, 1973±90 15.5 The basic model tested for twelve EU member states, 1973±90 15.6 Model with exports and terms of trade for twelve EU member states, 1973±90 15.7 Tested model for the eight advanced member states (category 1), 1973±90 15.8 Tested model for the four less advanced member states (category 2), 1973±90 16.1 The diffusion of privatisation policy 16.2 Estimated percentage of state-owned assets in selected sectors for various OECD countries, 1990±91 16.3 Central government budget surplus/de®cit as percentage of GNP 17.1 Economic indicators for the Balkan countries, 1995 17.2 Economic indicators for the EU countries, 1995

206 208 209 209 210 238 239 240 263 264 265 266 269 272 273 274 282 284

287 295 297

This page intentionally left blank

List of Abbreviations 2SLS ADF BERD COST CRH CSFs DF EC ECSC ECT Ecu EFTA EPO EMS EMU ERM ESA ESF ESRO EU EUF EURATOM FAST FDI FIML GDP GERD GDPPC GFCF GNP IFS IIT ILO IM IMF

Two stage least squares Augmented Dickey±Fuller Business Expenditures for Research and Development Community Organisation of Science and Technology Constant returns hypothesis Community structural funds Dickey±Fuller European Community European Coal and Steel Community Error-correction term European currency unit European Free Trade Association European Patent Of®ce European Monetary System European Monetary Union Exchange Rate Mechanism European Space Agency European Social Funds European Space Research Organisation European Union European Union per capita funding European Atomic Energy Community Forecasting Assessment in the ®eld of Science and Technology Foreign direct investment Full-information maximum likelihood Gross domestic product Government Expenditures for Research and Development Gross domestic product per capita Gross ®xed capital formation Gross national product Institute of Fiscal Studies Intra-industry trade International Labour Of®ce Internal market International Monetary Fund xiii

xiv List of Abbreviations

JRC LIML LFRs ML MMSE NAFTA OCA OLS PP RCA R&D RDI RI R&TD ROW SEA SF SI SIC SITC SMEs SMP SUR VAR VECM WS

Joint research centres Limited-information maximum likelihood Less favoured regions Maximum likelihood Minimum square error criterion North American Free Trade Association Optimal currency area Ordinary least squares Phillips±Perron Revealed comparative advantage Research and development Research and development intensity Regional index Research and technological development Rest of the world Single European Act Structural funds Specialisation index Standard Industrial Classi®cation Standard International Trade Classi®cation Small and medium enterprises Single Market Programme Seemingly unrelated residuals Vector autoregressive representation Vector error-correction modelling Welfare state

List of Contributors George S. Alogoskou®s: Athens University of Economics and Business, and CEPR. Hong Bai: Centre for Economic Forecasting, London Business School, United Kingdom. George C. Bitros: Athens University of Economics and Business, Greece. Laurence Boone: CEPII (Centre d'Etudes Prospectives et d'Informations Internationales), France. Isabel Figuerola-Ferretti: London School of Economics, United Kingdom. Nicholas V. Gianaris: Fordham University, USA. Dimitrios Giannias: Department of Economics, University of Crete, Greece. Stephen Hall: Management School, Imperial College, University of London, and Centre for Economic Forecasting, London Business School, United Kingdom. George Hondroyiannis: Economic Research Department, Bank of Greece, and Harokopio University, Athens, Greece. Florence Hubert: National Institute of Economic and Social Research, United Kingdom. Yannis Katsoulakos: Athens Economic and Business University, and Centre for Economic Research and Environmental Strategy, Greece. George M. Korres: Department of Geography, University of Aegean, and member of the Institute of Computational Mathematics, Crete Institute of Research and Technology, Greece. Maria Lekakou: Piraeus University, Greece. Sarantis Lolos: Economic Research Department, Bank of Greece, and Panteio University, Athens, Greece. Nigel Pain: National Institute of Economic and Social Research, United Kingdom. xv

xvi List of Contributors

Eleni Paliginis: Middlesex University, United Kingdom. Evangelia Papapetrou: Economic Research Department, Bank of Greece, Greece. Yannis G. Paraskevopoulos: Queen Mary and West®eld College, United Kingdom. Apostolis Philippopoulos: University of Essex, and Athens University of Economics and Business, Greece. Yonghao Pu: East London University, United Kingdom. Nicholas Sarantis: The London Guildhall University, United Kingdom. David Shepherd: Management School, Imperial College, University of London, United Kingdom. Nicholas Tsounis: Centre for Economic Research and Environmental Strategy, Greece. Vanghelis Vassilatos: Athens Institute for Economic Policy Studies, Greece. George Vlachos: Piraeus University, Greece. George Zis: Manchester Metropolitan University, United Kingdom.

Preface Economic integration is the most noteworthy issue in international economic policy at the beginning of the twenty-®rst century. The recent successful examples of the European Union (EU) and the North American Free Trade Association (NAFTA) have raised some important questions about the economic integration process and the possible establishment of economic unions in other parts of the world. This book makes an important contribution to the investigation of this topic. The book is intended to provide a basic understanding of the current issues in and the problems of economic integration, and it examines many aspects and consequences of this integration that are obscure or as yet unexplored. After addressing general issues in the ®eld of economic integration, the discussion turns to empirical and theoretical aspects of monetary union, social policy reform and social union, public ®nance and technology policy. In particular, with its wide range of topics, methodologies and perspectives, the book offers stimulating and wide-ranging analyses that will be of interest to students, economic theorists, empirical social scientists, policy makers and the informed general reader. We have used the abbreviation EC to refer to the European Community and EU to refer to the European Union: historical accounts of policy developments prior to the Treaty on European Union refer mainly to the EC, whilst references to more recent and current policy refer to the EU. The volume comprises four parts. Part 1 is devoted to macroeconomic issues and the problems of economic integration. The chapters in this part contain theoretical and empirical analyses of economic integration, the European Union and the European Monetary System. Part 2 investigates the microeconomic implications of economic integration with regard to manufacturing, foreign direct investment, unemployment and fertility. Part 3 deals with institutional matters and the policies of integration, including technology policy, shipping policy, the distribution of EU funds, regional development and productivity problems. Finally, Part 4 discusses the challenges for an integrated Europe, with emphasis on social policy, the welfare state, political reforms and privatisation. I would like to thank all the contributors to this volume for their great enthusiasm and willingness to cooperate. The publication of this book has been a truly worthwhile project, involving 26 authors. I would also like to xvii

xviii Preface

thank the anonymous reviewer of the volume, and above all my publisher for his encouragement and support. GE O R G E M. KO R R E S

1

Introduction George C. Bitros

1.1

The background

Suppose that there are two completely independent countries that are similar in most respects but have different currencies and different economic structures. Suppose also that these countries wish to move from their present regime of bilateral trading to complete uni®cation. The fundamental question that arises is: under what conditions would uni®cation be economically bene®cial for both countries? After uni®cation the currency of each country would be abolished and all transactions within their uni®ed area would be carried out in a common currency. Hence, all activities that under bilateral trading were associated with the conversion of one currency to the other would cease and the countries would experience some savings in transactions costs. This would be a plus. But at the same time the two countries would lose the freedom to conduct their own independent monetary policies and this would deprive them of the ability to adjust their exchange rates to counteract external demand and supply shocks. Hence, they would be forced to adjust through changes to nominal wages and employment, which would incur higher adjustment costs. This would be a minus. Therefore, the answer to the question posed above is that uni®cation would be bene®cial for both countries if their transaction savings were at least equal to the increase in their adjustment costs. In the light of this proposition, which was ®rst presented in the early 1960s when Robert Mundell laid down the foundations for the theory of optimal currency areas, it has been natural for researchers interested in worldwide policies of uni®cation to turn their attention to the determinants of transaction savings and adjustment costs. As a result, and under the impetus of the prospect of a United States of Europe, a voluminous

1

2 Introduction

literature has developed with at least three branches: one trying to identify the forces and processes through which transaction savings and adjustment costs are determined; another trying to identify the functional forms and estimate the size of the parameters involved; and the third trying to devise policies and institutions through which the costs and savings from uni®cation could be brought into politically acceptable alignment. The main motivation when planning the present volume was to contribute towards a wider understanding of the dif®culties that lie ahead for European Monetary Union (EMU) and the policies that will be required at the national and supranational levels in order for it to succeed. However, the decision to proceed was also encouraged by the coincidence of several other factors. First, in late summer 1996 one of the editors attended two conferences and ascertained that a select number of papers presented there might be available for the volume.1 Second, these papers addressed a balanced range of key issues using different methodologies and sets of data, and most were of exceptional quality in terms of conception, execution and ®nish, so if put together under a single cover they would appeal to a wide audience of specialists, policy makers and practitioners. Third, it was considered that the few gaps left by the conference papers could be ®lled by solicited papers from the local academic community. With the volume now in hand, we are pleased to say that we are satis®ed with the ®nal result and thank all the participants for their support and encouragement.

1.2

The issues considered in this volume

A conventional approach in introductions like this is to provide a brief summary of each chapter, describing the issue with which it deals and what it does to the enhance the understanding of the issue. But in this case the approach would be cumbersome, and possibly not useful, due to the large number of chapters and their diversity. Therefore, in order to guide the reader in a meaningful way, we shall focus on the main issues discussed in this volume and the different points of view expressed by the authors. This implies that the sequence in which the chapters are considered may not coincide with the order in which they appear.

1.3

The nature and degree of convergence in Europe

The economic integration of Europe can be separated into three stages. In the ®rst stage the member states joined together in a customs union and

George C. Bitros 3

all tariff and non-tariff barriers to trade were set to be abolished. During the course of this the participating economies were expected to converge in the sense of attaining comparable degrees of ¯exibility in wages and prices, and comparable degrees of mobility in labour and capital. Then, when an adequate degree of convergence had been achieved and adjustment to supply and demand shocks through exchange rate changes became super¯uous, the customs union was expected to evolve into a fully ¯edged monetary union. The preparations for this second stage began in 1995 with the laying down of the Maastricht criteria for convergence. For those countries that could meet the criteria, economic and monetary union was set to start in 1999 with the adoption of the euro as the single European currency. Finally, in the third stage, while the remaining countries struggled to meet the Maastricht criteria, the process of economic integration would be precipitated by the abolition of whatever legal and institutional constraints remained to curtail price ¯exibility and the mobility of resources among the various regions. By implication, as we are only a few years from EMU, there is an urgent need to obtain a better understanding of the degree of convergence achieved so far and the nature, scale and time pro®le of the shocks that member states might experience after uni®cation. Relevant for some of these issues are the results reported in the ®rst two chapters in Part 1. In Chapter 2 Laurence Boone applies the time-varying parameter estimation method to supply and demand disturbances, obtained from a VAR decomposition of standard OECD data on annual GDP and GDP de¯ators, to investigate the extent to which the European economies have converged. From his analysis of supply shocks he ®nds that, even though France, Belgium, Austria, the Netherlands, Spain and Italy have converged signi®cantly towards Germany, this `core' group of countries is not acting as an attractive force for the rest of the European Union, including Germany itself. When demand-related shocks are considered, his analysis fails to indicate any progress in the convergence process since at least the mid-1980s. As a result the author is led to conclude that there is a strong asymmetry in the way that European economies are converging. In contrast Hong Bai, Stephen Hall and David Shepherd (Chapter 3), using a different set of data, a different methodology and a slightly different group of countries, arrive at a very different conclusion. In particular they use data on real GDP and nominal and real foreign exchange and long-term interest rates from Germany, the United Kingdom, Austria, Spain, France and Italy to test for the presence of synchronised business cycles between each of these countries and

4 Introduction

Germany. From their tests it emerges that, with the exception of real interest rates, all the series investigated show common cyclical features. This, they infer, suggests that these countries have converged signi®cantly and may not face serious shocks from the design of monetary policy after the adoption of the single currency. Nevertheless, as their tests are quite sensitive to the particular speci®cations adopted in the estimations the authors suggest that their results should be viewed with caution. Overall, therefore, the evidence from these two chapters shows that: (1) to the extent that some convergence has been achieved by a core group of countries, Germany and many countries of the so-called periphery have converged asymmetrically; (2) all countries remain far apart in their response to demand shocks; and (3) after monetary uni®cation, a great deal of coordination will be necessary in order for monetary and ®scal policies to confront the asymmetrical response of member states to demand shocks.

1.4

Controlling the European rate of in¯ation

When partial monetary uni®cation took effect in 1999 the supply of base money for the euro was placed under the control of the European Central Bank. In order for the latter to maintain price stability it will need to have precise information on the form and stability of the demand for money within the uni®ed currency area. To prepare for this eventuality, for some years researchers have been estimating demand for money functions on the assumption that all, or a core group of, countries participating in the European Exchange Rate Mechanism (ERM) would join the monetary union from its inception. In Chapter 4 Yonghao Pu and George Zis follow this tradition. Their main ®nding is consistent with the view that the M1 and M2 money demand functions may be stable and well behaved. Owing to the innovations introduced by the authors with respect to the level of aggregation, the speci®cation of the equations, the estimating methodology and so on, the likelihood that this will be so at the level of both the individual member states and the single currency area is strengthened. So, even though the evidence obtained is tentative because the empirical puzzle of currency substitution has not yet been resolved, the nature of its implications is hard to ignore. For it suggests that, if adopted, a Germantype monetary policy would enable the European Central Bank to control the rate of in¯ation quite ef®ciently in the single currency area. This policy prescription leads naturally to the following question. To the extent that European governments have been unable to guarantee price stability in the past, is there any evidence to suggest that they might

George C. Bitros 5

obtain a ®rmer grip on their in¯ation rates by tying their monetary policies to an in¯ation-averse country such as Germany? An attempt to answer this question is made by Nicholas Sarantis (Chapter 5) and George Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos (Chapter 6). The former uses annual data over the period 1952±92 for eight ERM countries and estimates ®xed and time-varying parameter models, whereas the latter employ annual data from Greece for the period 1960± 94 and estimate a simultaneous model for wage in¯ation, price in¯ation and unemployment. Despite the differences in their data speci®cations and estimating approaches, the evidence obtained in both chapters points to the same conclusion. Even if countries were to employ the Bundesbank model they could not expect to experience signi®cant credibility gain, the reason being that the rate of in¯ation is in¯uenced by economic fundamentals such as unit labour costs, demand conditions and world commodity and oil prices. In summary, these chapters conclude that the European rate of in¯ation could be ef®ciently controlled by a mixture of three policies: German-type monetary policy, a ®scal policy aimed at controlling aggregate demand, and a structural policy to address the rigidities of markets.

1.5

Real convergence due to economic integration

Turning next to the structural effects of economic integration, a convenient starting point is with the predictions that accompanied the European Commission's 1985 white paper Completing the internal Market. According to the Single Market Programme, as this policy initiative came to be known, the abolition of all barriers to trade and factor ¯ows would have positive static and dynamic effects on welfare and economic growth, particularly to the less developed countries of the EU. These effects were expected to result from the signi®cant allocation, accumulation and location effects that the policy would provoke, and from acceleration of the diffusion rate of innovation and technical change. So the question that arises is whether and to what extent these predictions materialised. To obtain an answer, Yannis Katsoulakos and Nicholas Tsounis (Chapter 7) calculate several relevant indices with the help of data at the three-digit NACE manufacturing level for the 1981±92 period. Their ®ndings lead them to conclude that the Single Market Programme did indeed spur trade creation in most manufacturing sectors. But contrary to expectations, all the other indices show little improvement. The gains in the competitive position of most manufacturing sectors were minimal, the advances in specialisation were marginal and the expansion of intra-

6 Introduction

industry trade remained very limited and con®ned to a small number of sectors. On balance, and given that the authors assume that Greek manufacturing would not have progressed in the absence of the programme, one may be tempted to judge that it failed to bring about the intended effects. Yet the evidence needs to be interpreted with caution because in the period in question the Greek economy in general and Greek manufacturing in particular were operating under extremely volatile ®scal and monetary conditions. An additional reason to be cautious about using the results from Greece to judge the degree of success or failure of the Single Market Programme is the evidence furnished by Florence Hubert and Nigel Pain in Chapter 8. The authors use annual panel data on investment by German corporations in four economic sectors and ten EU and non-EU locations to estimate a model of foreign direct investment while controlling for most relevant variables and sector-speci®c ®xed effects. Their results reveal that the Single Market Programme had several positive effects: it stimulated the level of investment in all four industries, it diverted investment from the United States into the EU, and it promoted the diffusion of new ideas and technologies within Europe. In conclusion, if a single implication must be drawn from the results and analyses presented in Chapters 7 and 8, it is that the Single Market Programme contributed signi®cantly to real convergence in Europe. But certain member states failed to bene®t as much as expected, due mainly to their own shortcomings.

1.6

Labour force growth and unemployment

Unlike the United States and Japan, the EU in general and certain member states in particular have experienced very high and persistent rates of unemployment. As a result, many studies have attempted and many hypotheses have been advanced to explain the main factors that determine the growth of the labour force as well as the processes of job creation, job attrition and the matching of job seekers to vacancies. The present volume adds to this literature with the help of two chapters, the ®rst by Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos (Chapter 9) and the second by George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou (Chapter 10). In Spain, note the authors of Chapter 9, periods of robust economic growth and recession have been marked by high unemployment but ample job vacancies. To shed light on this puzzling phenomenon, they set up a GARCH-M model of the exit rate from the labour market and

George C. Bitros 7

estimated it using quarterly data from 1979±96. From their ®ndings they conclude that arrangements such as ®xed-term contracts and generous job protection and unemployment insurance schemes exert a positive in¯uence not only on the expected mean but also on the expected variance of the exit rate. However, their investigation stops short of providing workable policy instruments, so the usefulness of their results for application purposes is quite limited. The authors of chapter 10 use various high-power tests on data from Greece for the period 1960±95 to investigate the nature of the relationship among three variables: the weekly hours of work, the fertility rate and the growth of output. It emerges that an unanticipated increase in unemployment leads in the short term to a reallocation of time from labour effort to childbearing, thus increasing the fertility rate. In the long term this growth in fertility leads to an expansion of the labour force, which is hence unequal to output growth. Finally, the growth in output feeds back to the fertility rate, since the latter is found to be an endogenous variable in the labour market and the process of growth. In this concluding section the authors draw on this relationship to suggest various policy initiatives that could accelerate the convergence of the Greek to the major EU economies without jeopardising the achievement by Greece of the Maastricht criteria for nominal convergence.

1.7

Real convergence through active policies

Recognising the signi®cant disparities in the levels of economic development that existed among member states as well as among regions and key economic sectors, from the very beginning the EU authorities adopted various policies to speed up convergence. Their approach was to formulate EU-wide policies and encourage their implementation by member states by distributing of ®nancial assistance based on need, readiness to bene®t, adherence to rules and other criteria. In Chapter 11 George Korres gives one example of such a policy. Korres traces the stages through which the efforts to stimulate technological advancement have evolved and assesses the achievements. And asserts that over the years the EU has devoted only modest funds to bring about parity with the United States and Japan in the development and commercialisation of new technologies. Furthermore, before the socalled `framework programmes' were initiated in 1984 its endeavours were largely experimental, and even after their introduction the results fell short of expectations because the policy lacked proper orientation and coordination. But in more recent years the EU has shown new vigour. In

8 Introduction

order to improve policy coordination it took overall control of decision making on technology matters, it set new inward-looking criteria for the selection and implementation of projects, and, last but not least, it placed greater emphasis on joint research efforts, quicker diffusion processes and improved cooperation between theoretical and industrial research. Another illuminating example of the EU's attempt to bring about convergence in economic sectors where member-state practices are far apart is that of coastal shipping. This is presented by George P. Vlachos and Maria Lekakou in Chapter 12. Certain countries in the EU have large island areas that depend mainly on cargo and passenger/ferry ships for the transportation of people and goods. But there are many market formations, ranging from state monopoly to state-controlled competition, and the market structure of coastal shipping often violates the central tenet of EU Regulation 3577/92, which calls for unrestricted freedom in the provision of maritime services in the EU (maritime cabotage). To facilitate a smooth transition period, Regulation 3577/92 involves a four-pronged approach. First, the date at which all restrictions to entry will be terminated is set at 2004. Second, the upper age limit for ships entering a coastal ¯eet is set at 20 and no ship may be older than 35 years ± this is designed to reduce the average age of ships in the inter-EU ¯eet. Third, the regulation speci®es the provisions under which coastal shipping services may be subsidised by member states after the end of the transition period. Finally, the policy initiatives that are required to meet the provisions of the regulation are left to the discretion of the member states. Hence the member states are free to adjust at will, but in the knowledge that unrestricted entry must be in place by 2004. The latter does not mean that the EU authorities have adopted a handsoff approach, rather they have been closely monitoring the progress made by each country towards the objectives set and modifying the ®nancial assistance provided to them in order to induce cooperation. But the success of this approach is not easy to evaluate because there are no objective criteria against which to measure it. The value of the contribution made by Dimitrios Giannias in Chapter 13 lies in the fact that he offers and tests a methodological approach for evaluating optimality in the distribution of funding for any policy on purely technical grounds. In summary, to bring about real convergence, the EU authorities rely both on the powerful structural effects of economic integration and on the implementation of sectoral policies. In enforcing the latter, aside from moral and political inducements, they employ the carrot of ®nancial

George C. Bitros 9

assistance and the stick of deadlines. So far the approach has been quite effective and this is re¯ected in the determination of EU to proceed to the next phase ± economic uni®cation.

1.8

Policy options for economic growth and social cohesion

Globalisation and the need to be competitive in international markets pose several challenges to the EU, the most urgent of which are as follows. First and foremost is the challenge to reform the traditional role of the state, particularly in the production of goods and services. The second challenge is to reform the welfare state in order to encourage greater mobility and human capital accumulation, and the third is to foster a feeling of European identity among the peoples in Europe. The last four chapters of this volume discuss various aspects of these challenges and offer constructive ideas for policy options at the EU level. In Chapter 14 Eleni Paliginis traces the reasons for the establishment of welfare systems in Europe and explains why their maintenance is no longer viable. In the core countries, she conjectures, competition for a share of the international markets could lead to a downward spiral in social expenditure, whereas in peripheral countries whose welfare systems are in need of development, neither the government nor the private sector are able to fund it. In view of this impasse, and in the interest of creating a `true European community', Paliginis recommends the creation of a European social welfare net, which would imply a regular transfer of resources from the core to the periphery. However, since a net of this sort is not likely in the foreseeable future, other ways of upgrading the existing welfare systems should be considered. One well known method is to adopt policies aimed at promoting economic growth. More speci®cally, if the rate of European economic growth could be accelerated, social expenditure could increase, despite the negative developments cited by Paliginis. In this respect two relevant policy options are analysed in Chapter 15 by George Korres and in Chapter 16 by George Bitros. While Korres presents estimates to show that technical change is positively related to per capita gross national product, Bitros argues that privatisation contributes signi®cantly to economic growth. Finally, there is the expansion of the EU to countries of the former Eastern European communist bloc. Of course, this expansion cannot be viewed as a new frontier promising lucrative business opportunities and spectacular economic growth, mainly because these countries are at present endowed with values systems and human capital that will take

10 Introduction

many years to transform into the necessary ingredients for marketdriven economic development. But still, as Nickolas Giannaris argues in Chapter 17, these countries have a long history of high rates of investment and there is no reason why enlargement of the EU to encompass their territories cannot bring mutual advantages.

1.9

Conclusion

So far the steps taken to construct a United States of Europe along federalist lines has been successful. By slowly bringing the economies of the various countries into a state of relative harmonisation, the stage is now set to launch one of the biggest economic experiments in the history of mankind: the creation of an economic union with no borders, tariffs or other barriers to trade, and no restrictions on the movement of people and capital. When European uni®cation ®nally takes effect the event will mark one more triumph for the market. But numerous problems will continue to demand solutions. Perhaps the most pressing problem is that of unemployment. Given that the economic policies of the United States have managed for several years to secure signi®cant rates of economic growth with low in¯ation and historically low unemployment, we in Europe can look to the United States for solutions. Not unrelated to the problem of high unemployment are the personal disincentives emanating from the welfare systems that took hold in Europe in the postwar period. These systems are in need of major overhaul to raise the rate of European economic growth. By promoting investment in new technologies as well as in human capital and mobility, Europe will be able to bene®t from the current trends of globalisation and economic integration. Last but not least, a case can be made on purely economic grounds for the EU to look eastwards. The economic problems confronting the people of these territories are transient, and friends are for hard times as well as good. Providing economic assistance now, with the prospect of incorporating these countries into the EU when they are ready, could be a small price to pay for a Europe that may one day stretch from the Atlantic Ocean to the Urals and from the North Pole to the Mediterranean Sea. Note 1. The Conference on Economic Integration in Transition, held at the Athens University of Economics and Business Science from 22±24 August 1996, and the Conference on Money, Macro and Finance, held at the London Business School from 4±6 September 1996.

Part 1 Macroeconomic Issues of Economic Integration

This page intentionally left blank

2

A Dynamic Analysis of Structural Asymmetries in the European Union Laurence Boone1

2.1

Introduction

In the run-up to EMU there is still no agreement on the state of convergence of the EU economies, nor on the scale of costs and bene®ts of monetary union. In the literature some economists have discussed these issues based on the optimal currency area (OCA) theory developed by Mundell (1961) and extended by Kenen (1969) and McKinnon (1963). The main idea is that a (European) country should join a monetary union if the savings it will make in transaction costs are estimated to be greater than the costs it will incur by relinquishing national exchange rate and monetary policy. Adjustment costs depend directly on the asymmetry of the disturbances an economy is facing: if all the member countries of a monetary union face the same disturbances, there are no costs in having a common policy. In contrast, when faced with asymmetric shocks, countries will suffer higher adjustment costs induced by a common policy that may not be the most appropriate. The costs may be mitigated by ®scal policy, wage and price ¯exibility and labour mobility. However, early empirical work revealed a lack of labour mobility and the existence of price rigidity. Hence, it appears that asymmetries play a crucial part when assessing the costs of joining a monetary union. The measures of asymmetries that can be found in the literature are mostly static (Bayoumi and Eichengreen, 1993a, 1993b, 1994; Funke, 1995). Various economists have used different econometric methods to estimate series of the shocks that economies have faced, and then derive a correlation coef®cient, over a period that generally starts in the late 1960s and ends around 1994±95. The trouble with this approach is that it implicitly assumes that the correlation coef®cients are stable over the

13

14 Structural Asymmetries in the European Union

entire period in question, which depends on three assumptions. The ®rst is that there is no evolution in the structural similarities between European countries over the period. The second is that there is no extraordinary exogenous event that may disturb structural convergence if it is at work. Finally, this approach implicitly assumes that shocks are structural and not policy induced. Some authors have attempted to take account of regime changes by comparing the correlation coef®cients in the pre- and post-ERM period (Fatas, 1996), and in periods excluding and including German reuni®cation (Funke, 1995). However, the process of integration among European countries is likely to be an ongoing one. Hence, one needs a dynamic measure of this correlation coef®cient to assess the evolution towards greater similarity among European countries. This requires a methodology that not only allows a time-varying correlation to be estimated, but also takes into account exogenous events such as German reuni®cation. The Kalman ®lter offers such a methodology. Not only does it allow time-varying parameters to be estimated, but it also allows for structural breaks. On the other hand, an econometric method is also required that allows structural shocks to be estimated. The early literature generally distinguishes between monetary, supply and demand shocks (see Demertzis et al., 1996). As asymmetries in monetary shocks are likely to disappear with monetary union, we shall focus on supply and demand shocks. The distinction between the two rests on the following idea. Permanent shocks to the economy are assumed to be mostly caused by changes in technical knowledge. Hence, the movements in output due to permanent shocks are considered as evidence of the importance of movements in aggregate supply. Therefore, supply shocks are thought to re¯ect (a)symmetries in the structure of the economies, and are considered to be independent of policies. Hence, assessing these asymmetries will have crucial implications for assessing the costs induced by monetary union. On the other hand, the main effects of aggregate demand shocks on output are largely transitory. However, symmetry and asymmetry of demand shocks are of interest, to the extent that they may be induced and corrected by policy. Measuring asymmetries in demand disturbances will then enable us to assess whether monetary and budgetary policies are becoming or need to become more coordinated across countries. This chapter is organised as follows. In Section 2.2 the state of empirical knowledge about OCA theory will be brie¯y reviewed. Section 2.3 presents a dynamic method, used extensively by the convergence literature and based on time-varying parameter estimation via the Kalman ®lter. These two sections highlight how we can obtain a dynamic measure

Laurence Boone 15

of correlation of asymmetric shocks in Europe. Section 2.4 introduces the data that proxy structural European shocks and explains how they are obtained. Section 2.5 presents the results for 14 EU countries for both supply and demand shocks, emphasising the differences arising from dynamic (versus static) estimations. We also assess the power of attraction of Germany and the symmetry of convergence paths, namely: have countries converged towards Germany or have all countries, including Germany, made an effort towards convergence? The conclusion brie¯y discusses the policy implications of this exercise.

2.2 2.2.1

Motivation A word on the evidence about OCA criteria in Europe

Discussion about monetary union usually rests on the optimal currency area (OCA) theory, developed by Kenen (1969), McKinnon (1963) and Mundell (1961). This approach is based on the idea that when an economy is faced with a (negative) shock to its trade balance, the exchange rate is a more ef®cient adjustment tool than labour, wages or prices. This stems from the observation that nominal wages are sticky and labour is fairly immobile. Therefore, it is easier to adjust to external shocks by changing the real exchange rate or the terms of trade. Monetary union deprives the member states of the exchange rate tool, and therefore raises adjustment costs. At the same time it also saves the transaction costs incurred by countries due to the existence of multiple currencies. Thus, following Mundell (1961), the literature states that it is optimal to establish a monetary union if the savings in transaction costs are at least equal to the increase in adjustment costs. The savings in transaction costs are a function of an economy's openness, whereas adjustment costs depend on the asymmetry of disturbances and the interregional mobility of factors. These costs may be mitigated by ®scal policy, ¯exibility of prices, and wages and labour mobility. On the other hand, asymmetries may be enlarged or reduced depending on the degree of regional specialisation. The empirical work has focused on these elements to assess the costs of giving up the exchange rate instrument. There seems to be global consensus on the following points:

. European countries are fairly open to trade. . Capital is rather mobile, but labour is immobile. However, as long as labour is as immobile across countries as it is across regions, there is no implication in terms of monetary union (Gros, 1996).

16 Structural Asymmetries in the European Union

. Wages and prices do not appear to be ¯exible. Yet if real wages are rigid, they are unaffected by changes in nominal variables. Therefore monetary union will not affect the costs arising from this type of rigidity (Buiter, 1995). But the origin of the rigidity (nominal or real) is still subject to discussion.

Calculations by Bayoumi and Prasad (1995) for the period 1970±89 for the United States and 1970±87 for the EU show that the share of variance in output explained by region-speci®c shocks is slightly higher in Europe than in the United States (31 per cent versus 26 per cent, the rest of the variability being explained by global and industry-speci®c shocks). Hence, industries tend to be more concentrated in the United States than in Europe. The effect of this is ambiguous, depending on the level at which EU concentration takes place, that is, at the national or the regional level. Indeed, an increase in the degree of specialisation at the national level implies a more pronounced national business cycle (and therefore may require national policy actions), while an increase in the degree of specialisation at the regional level re¯ects increasing interdependence between regions that may belong to different countries. In the latter case, the meaning of national borders is reduced and European monetary union will make little more difference than national monetary union. Fatas (1996) shows that cross-regional correlations of the business cycle (as proxied by the unemployment rate) in the EU rose in the post-ERM period (1980±92) compared with the pre-ERM period (1966±79), whereas they diminished at the national level during the same periods. Hence, from this point of view, monetary union might have a lower economic cost than expected. The relative importance of each of the above points depends on the degree of asymmetry between European countries. Indeed, factor mobility is needed if one sector and/or one region is affected by a shock that does not affect the rest of the EU. Similarly, the degree of regional specialisation is said to have an impact on the degree of symmetry of shocks: the higher it is, the more asymmetrical the shock and the transmission of the shock at the regional level, (which might be different from the national or the EU level). Furthermore, asymmetries enhance the persistence of the disequilibrium induced by a shock. The relevant information to assess the costs of monetary union, then, would seem to be (1) the magnitude of the asymmetries and (2) their relative importance across countries. The ¯exibility of wages and prices is important insofar as an economy has no other way of adjusting to a shock (®scal transfer or one of the other points), whether these are symmetrical or not.

Laurence Boone 17

Demertzis et al. (1996) formalise the notion that the costs of monetary union are directly related to the degree of asymmetry between national shocks and transmission mechanisms. If labour is immobile and wages and non-wage labour costs are not ¯exible, asymmetrical shocks will lead to signi®cant and persistent disequilibrium under a common currency (since the exchange rate channel cannot be used any more).2 Hence, to test whether Europe, or a subset of countries, could function well enough as a single currency area, it is necessary to examine the symmetry of shocks. The next step is to de®ne which types of shock are of interest. In particular, monetary shocks are meant to disappear with a single currency and might therefore be less relevant.3 Bayoumi and Eichengreen (1994) argue that supply shocks that have a permanent effect on economic aggregates are crucial in assessing intrinsic economic similarities between the countries of the EU, as they are more independent of policy. On the other hand, the analysis of demand shocks is relevant to the extent that they may have a temporary effect on the economy that may stem from or be corrected by monetary or ®scal policy.4 In the absence of ¯exibility in factor costs and quantities, the similarity of supply shocks will condition the costs of adjustment induced by monetary union. Furthermore, the degree of similarity between demand shocks will condition the extent to which these adjustment costs are painful with respect to the loss of the monetary instrument. 2.2.2

The measurement of OCA criteria

Econometric methodology has been fairly speci®c about the criterion being analysed. There is virtually no empirical work on the bene®ts of monetary union5 and the empirical literature has focused nearly exclusively on the costs of such a union. As discussed above, there is global consensus on only some criteria (openness to trade, mobility of factors and ¯exibility of factor prices). The focus here will be on the size and magnitude of asymmetries in supply and demand shocks, and how they have evolved.6 The ®rst problem the literature has had to deal with has been the identi®cation of shocks, since shocks are not variables for which data are readily available. Several methodologies have been used. The decomposition of series into trends and cycles leads to the measurement of business cycle similarities or differences across countries as a measure of the similar or dissimilar structure of shocks. The VAR methodology goes into more detail by decomposing shocks into `supply' and `demand' components. Finally, cross-country regression methods are used to distinguish countryspeci®c shocks from others (global and industry-speci®c shocks).

18 Structural Asymmetries in the European Union

The importance of a shock will then depend on two factors: the size and magnitude of its impact on macroeconomic aggregates (measured by the variability or variance of the aggregate), and the correlation of these shocks across countries as measure of their symmetry. The correlation coef®cient is interpreted in the following way:

. Sign of the correlation: a negative (positive) correlation of a variable in two countries characterises asymmetry (symmetry);

. Magnitude of the correlation: the lower or less signi®cant the

correlation, the less important the issue of (a)symmetry between the considered variables. A benchmark is needed to assess the relative size of the asymmetries between the EU countries. The usual reference is the existing correlation for the US states (Gros, 1996). Bayoumi and Eichengreen (1993a, 1993b, 1994) use the correlation between the regions of Canada from 1968±88 for output and prices. Finally, Funke (1995) uses as a benchmark the correlation between German Lander, using the same data as Bayoumi and Eichengreen but for the period 1968±92.

In the next subsection we brie¯y survey the evidence provided by these methods. 2.2.2.1

Measurement based on business cycles

Several studies (for example, Artis and Zhang, 1995; Christodoulakis et al., 1995) reveal similarities in the business-cycle stylised facts for the EU countries. Artis and Zhang (1995) study the impact of the ERM by contrasting such facts over the pre-ERM period (1961±79) and from 1979±93. They consider the monthly industrial production cycles7 for most ERM countries, using the United States as the benchmark. Canada, Japan and the UK are used to distinguish ERM-speci®c phenomena from general tendencies in the business cycle. They conclude that the synchronisation and linkage between the ERM economies and Germany have grown strongly over the two periods, whilst the linkages with the US cycle have diminished. However, this methodology does not allow assessment of whether these increasing similarities come from the shocks themselves or from the responses to the shocks, since both are embodied in the short-run cyclical component. 2.2.2.2

Measuring asymmetries from variance decomposition

Another way of measuring (a)symmetries between the EU countries rests on the decomposition of GDP volatility. The idea is to analyse the share of

Laurence Boone 19

the variance of GDP due to common factors (that is, symmetrical shocks) and to country-speci®c factors (asymmetrical shocks). Using these methods over the period 1962±95, Melitz (1996) ranks countries according to the degree of asymmetry, as indicated by the variance of asymmetrical shocks to that of symmetrical shocks. The lowest ratio is for France, then said to be the most `typical' country, which means the most integrated country. France is followed by Italy and the UK. Germany does not appear to be very typical, according to this criterion. However Belgium, Austria, the Netherlands and, to a lesser extent, Sweden seem fairly typical. Hence, these results tend to con®rm the general evidence of a core of integrated countries similar to that obtained by the VAR methodology, and a periphery of less integrated economies. 2.2.2.3

Estimation and measurement of (a)symmetries from VAR models

VARs attempt to redress two criticisms of the above method:

. They allow shocks to be distinguished from responses; . They allow for a decomposition of the shocks to identify their origins (for example, monetary, supply or demand).

The general methodology followed in this literature is that developed by Blanchard and Quah (1989). The estimation of a VAR model leads to the estimation of residuals as a by-product. Then econometric theory helps to decompose these residuals into several structural components through the computation of some square roots of their variance/covariance matrix, after which the transmission of the shocks can be analysed through the construction of impulse response functions. These latter are representations of the impact of a shock on the macroeconomic aggregates of the model. Bayoumi and Eichengreen (1996) and Funke (1995), among others, use VAR models to identify aggregate supply and demand shocks. Let yt be a bivariate vector of variables, say differences in GDP and prices (as in Bayoumi and Eichengreen, 1994, 1996), which can have a general, dynamic structural representation, as follows: B0 yt ˆ k ‡ Bt yt

1

‡ B2 yt

where: yt ˆ …
GDPt ;
Pt † gdp

ut ˆ …ut ; uey †

2

‡


‡ B p yt

p

‡ uI

…2:1†

20 Structural Asymmetries in the European Union

B0 ˆ

  1 . . . B21 B21 . . . 1

k ˆ …k1 ; k2 † and ut is white noise. Equation 2.1 summarises Bayoumi and Eichengreen's (1994) representation of the economy. Intuitively, it links the bilateral in¯uence of GDP and prices to each other, with suf®cient dynamics to ensure that the ut are serially uncorrelated. Since this representation follows from an economic representation, ut is said to be structural. To estimate this dynamic system we need to invert the matrix B0 ; that is, to put the above system into the following form: yt ˆ c ‡ 1 yt

1

‡ 2 yt

2

‡ . . . ‡  p yt

p

‡ "t

…2:2†

where: c ˆ B0 1 k s ˆ B0 1 Bs "t ˆ B 0 1 u t Equation 2.2 is a vector autoregressive representation (VAR) of the dynamic structural system (2.1). In other words a VAR can be viewed as a reduced form of a general, dynamic structural system. If ut is white noise, then "t is also white noise. However, it is quite easy to see that the "t are not structural disturbances because they are a linear combination of all the shocks affecting the economy; hence, it is not possible to distinguish the origin of a shock with this representation. Furthermore, there is no reason to believe that the elements of "t are not contemporaneously correlated with one another. However, we are interested in the origin of the shock, as we would like to assess the impact of different types of economic shock on the variables of the system. For that we need to proceed in two steps:

. First, orthogonalising the disturbances "t will enable us to get disturbances that are not contemporaneously correlated;

. Second, some extra identi®cation rules, derived directly from economic theory, will allow us fully to identify the system.

Orthogonalisation can be done through an MA representation. For full identi®cation, structural VARs use economic theory. Bayoumi and Eichengreen (1994, 1996) argue that demand disturbances affect only

Laurence Boone 21

prices in the long term, while supply disturbances may have long-term effects on both prices and output. These economic hypotheses allow the long-term multipliers from the MA representation to be identi®ed, and allow the series [ut ] = [(u1t , u2t )] to be estimated. These two series of structural disturbances are labelled demand and supply disturbances, according to the hypotheses made above to identify them. They represent exogenous shocks that were not anticipated by the agents. Bayoumi and Eichengreen then compute the correlation coef®cient of these shocks to provide evidence of their similarities. They identify a core group of EU countries (Austria, France, Denmark and the Benelux countries) whose disturbances are highly correlated with Germany. This is con®rmed by Funke (1995), who uses the same data and methodology but over a time period running to 1994. He notes a lowering of the correlation coef®cients, which is attributed by Bayoumi and Eichengreen (1996) to German reuni®cation.8

2.3

A need for dynamics

A general critique of the above presentation is that all the methodologies considered are static measures of asymmetry. They are merely a coef®cient computed over a period of time. The only attempt to introduce dynamic factors into the model is through the measurement of correlation and variances pre- and post-ERM. However, this will not help to assess a move towards greater symmetry if this requires long-term adjustment lags or is still an ongoing process. Yet the degree of integration between the EU countries could not be expected to be the same in the 1980s as it was in the 1970s, or even in the early 1990s with respect to the 1980s. Furthermore, these methods make it dif®cult to account for structural breaks, such as German reuni®cation. At best the sample period can be split in two parts, or 1991 may be deleted from the sample considered (see Bayoumi and Eichengreen, 1996). As an illustration of these problems, Table 2.1 gives the standard deviation of GDP growth rate for the EU countries. Table 2.1 highlights in a striking way the problems that traditional econometric methods cannot account for. First, the dramatic reduction in the standard deviations suggests that a convergence process has been at work since the beginning of the 1980s for the whole of the EU. Second, there is divergence in the speed of convergence from one country to another, with the standard deviation evolution being quite different for the core and the EU as a whole. Third, German reuni®cation appears as a strong abnormal event, which may have disturbed the convergence path(s).

22 Structural Asymmetries in the European Union Table 2.1 Standard deviations of GDP growth rates in Europe, 1964±94

Core Reextended core EU 14

1964±72

1973±83

1984±94

0.99 1.29 1.90

1.37 1.52 1.98

0.68 0.99 1.46

Note: Excluding 1991. Source: Pisani-Ferry (1996), calculations based on OECD data.

This con®rms that it is necessary to capture the dynamics of the process affecting the symmetry of shocks between countries, which requires measurement of the (speed of) convergence of the various shocks between the EU countries. We also need to take into account exogenous structural breaks. This requires a modelling strategy. The next section presents a dynamic method of measuring convergence that also takes into account potential structural breaks. 2.3.1

Measuring convergence

The empirical literature on convergence focuses on different issues from those treated in the empirical literature on structural asymmetries. Indeed, whereas works on asymmetry mostly focus on the structure of EU economies in terms of the similarity of the shocks they face, works on convergence look at the variables of these economies. A large number of authors work with the nominal variables that are required for convergence under the Maastricht Treaty (in¯ation, nominal interest rates). Others use the methodology developed in the empirical literature on endogenous growth for studying convergence amongst world economies (see, amongst others, Barro, 1991; Blanchard and Quah, 1989) and use it to measure convergence of GDP in the EU economies. Little work seems to have been done on the convergence of economic structures.9 Methods of measuring convergence vary from a simple measure of the dispersion of the variables (that is, a comparison of the variance of the two variables at the beginning and at the end of the sample period) to a more sophisticated approach using the concept of cointegration. In this latest framework, the necessary and suf®cient conditions for two series to converge is that the differences between the series have ®nite variances, and that the cointegrating vector between the two series is (1, ±1).10 However, these tests will only provide evidence of convergence if it was realised before the beginning of the sample period. If convergence started or was completed within the sample period, it will not be detected. We

Laurence Boone 23

saw in the previous section that it is very likely that the convergence process is still ongoing in the EU. Furthermore, as mentioned previously, German reuni®cation probably induced a structural break in these relationships. The Kalman ®lter is an econometric tool that allows time-varying parameters to be estimated. Haldane and Hall (1991) and Hall et al. (1992) combined the concept of cointegration with the Kalman ®lter to estimate time-varying parameters in a methodology that gives a dynamic measure of convergence. We brie¯y present this method below. The Kalman ®lter estimation process, which allows time-varying parameters to be estimated, is explained in Appendix 2.1. Consider three countries i, j and k. To measure the evolution of convergence of a variable X in country i towards country j, one can estimate the following measurement equation: j

Xit ˆ t ‡ bt Xt

…2:3†

where t and bt are time-varying coef®cients, de®ned in the following transition equations as autoregressive processes: t ˆ t

1

‡ 1t

bt ˆ bt

1

‡ 2t

…2:4†

These are pretty arbitrary structural forms but it is quite easy to generalise them. It is relatively straightforward to see that if b tends toward zero, j then ¯uctuations Xt do not help to explain ¯uctuations in Xit . Yet if b tends towards 1, then there is convergence of Xit towards Xit . In this case, at most we would ®nd a cointegrating vector (1, ±1) between Xi and Xj (assuming the residuals are white noise). The originality of the method relies on the estimation of the timevarying parameters. If a process of convergence is at work, but has not yet been achieved or will be achieved at some point in the sample, then the coef®cients b will be allowed to take a value close to zero at the beginning of the period, but tending towards 1 as the process of convergence takes place. However, this ®rst measure is fairly rough: it is possible that the entire world is converging for variable X, and with this method we cannot distinguish between a global trend towards convergence and a more speci®c convergence between two countries. Therefore, we need to contrast convergence on one country with convergence towards the rest of the world, which is formalised in the following measurement equation: j

…Xt

Xit † ˆ t ‡ t …Xit

Xt †

…2:5†

24 Structural Asymmetries in the European Union

where  and are characterised by the following transition equations: 1 ˆ t

1

‡ lt

1 ˆ t

1

‡ 2t

…2:6†

Hence, when tends towards zero, the movements of X for country i are explained by the ¯uctuations in X in country j; and when tends towards 1, the spread on X for countries i and j is explained by the spread between countries j and k; namely, there is no convergence,  is normally interpreted as autonomous convergence and will embody any explanatory effect that is not in the other variables.11 Hall et al. (1992) looked at convergence of nominal exchange rates, in¯ation rates and interest rates, using cointegration methods and the Kalman ®lter. They concluded that, on average, there was no cointegration over the period 1970±91 for any of the EU currencies with respect either to the dollar or to the Deutschmark. By splitting their sample in 1980, they were able to provide evidence of a structural break around that time, which triggered convergence of the exchange rates in the post-1980 period. Indeed, using the time-varying parameter method, they found that the convergence of nominal exchange rates was well established by the mid-1980s; something they could not reveal with traditional cointegration methods.

2.4

Methodology

This study applies the time-varying estimation method to measure in a dynamic way the correlation of supply and demand shocks. First the data is presented, and then the above econometric methodology is brie¯y explained. 2.4.1

The data

The data used here are taken from the series of supply and demand disturbances obtained from a VAR decomposition by Bayoumi and Eichengreen (1996). They proxy for structural similarities between the countries of the EU over the period 1963±94. They are annual, in order to avoid a direct interpretation of the results being prevented by too much noise. The series needed for the VAR decomposition are standard OECD data with an annual GDP and GDP de¯ator. 2.4.2

The econometric methodology

The econometric methodology is the time-varying parameter method described in Section 2.2. We focus on the more accurate method

Laurence Boone 25

(Equations 2.5±2.6) since we are interested in the convergence of the EU economies independently of a general move towards global convergence. Two potential convergence hypotheses are used: (1) convergence on Germany, in opposition to the rest of the world (proxied by the United States), and (2) convergence on the so-called core group (Belgium, the Netherlands, Austria and France). This is based on the hypothesis that while German reuni®cation might have triggered a divergence effect, there might still be convergence towards the core or within the core, which we hope to capture by such a regression. It is important to note that when we study the convergence of a core country towards the core, it is taken out of the core as the ensuing correlation between the two sides of the equation would otherwise affect the results. We apply this methodology for 14 EU countries (Luxembourg is not included) for both supply and demand shocks. The time-varying parameter methodology developed by Haldane and Hall (1991) and Hall et al. (1992) requires a maximum likelihood estimation via the Kalman ®lter (see Appendix 2.1). The estimation process involves careful speci®cation of the measurement equation and requires the use of diagnostic testing. Hence, each estimation procedure is checked with the usual statistical diagnostic tests, and corrected accordingly. For example, in some cases the results of the Jarque±Berra normality tests reveal the presence of outliers that correspond to particular, extraordinary, exogenous economic events of a country. Thus, it appears necessary in the case of most countries to correct for German reuni®cation by including a dummy variable. Indeed, German reuni®cation was a one-off exogenous event. Therefore, it might have altered the path of convergence on Germany, and temporarily disturbed the longterm convergence patterns. German reuni®cation was also an extraordinary event, whose characteristics should be distinguished from the supply shocks that occur in the normal course of an economy. As an illustration, we present in Section 2.5.1 a careful implementation of the method for French supply shocks, so as to underline all the steps taken and emphasise the advantages of a modelling strategy over simple, static coef®cients. For the others, a note is added when statistical corrections are required, emphasising the likely impact on the estimated coef®cients.

2.5

The results

In order to illustrate the methodology of estimation with time-varying parameters, the modelling of the estimation of the dynamic correlation coef®cient between French and German supply shocks will ®rst be

26 Structural Asymmetries in the European Union

explained in detail. The rest of the section then presents the results for other European countries. These results are synthesised into tables of averages over certain periods for all countries.12 Graphs of the evolution of the various supply and demand shocks for the period 1974±94 are presented in Appendix 2.2. The analysis will emphasise the emergence of a group of countries characterised by signi®cantly similar shocks, and will facilitate comparisons with the previous ®ndings in the literature. 2.5.1

French, German and US supply shocks

The estimated equation is YtGE

YtFR ˆ t ‡ t …YtGE

YtUS †

…2:7†

where Yi represents the supply shocks of country i, as identi®ed in the VAR decomposition by Bayoumi and Eichengreen (1996), and  and are as de®ned in Equation 2.6. Figure 2.1 shows a relatively important similarity between the French and German supply shocks, except at the time of German reuni®cation. This can be quite clearly identi®ed on the graph, and was the biggest supply shock experienced by Germany since the beginning of the sample period. This was followed by a relatively important negative supply shock in France, but about two years later. We therefore do not expect to see convergence taking place until the late 1980s, followed by a temporary 3 2.5 2 1.5 1 0.5 0 –0.5 –1 –1.5 –2 –2.5

Source: Bayoumi and Eichengreen (1996).

Figure 2.1

Supply shocks, Germany and France, 1963±93

Laurence Boone 27

Figure 2.2 Estimation of  (Equations 2.5 and 2.6), 1974±92

break and the resumption of convergence thereafter (though the size of the sample might prevent us from seeing it). A careful diagnostic check of the ®rst estimation of the time-varying parameters between France and Germany reveals quite clearly that it is necessary to introduce a dummy variable to allow for the impact of German reuni®cation (the Jarque±Berra normality test is excessively high, which usually denotes the presence of an outlier). This dummy takes the value of 1 in 1991 and 0 elsewhere. The likelihood ratio of this equation is much higher than is the case with one that does not take into account the German exogenous shock in 1991. Furthermore, diagnostic tests become reasonably good. Graphs of the estimated coef®cients  and between 1974 and 1994 are presented in Figures 2.2 and 2.4. In Figure 2.2 the coef®cient  is converging towards zero, although it has not yet reached that value. However, it appears rather stable from the beginning of the 1980s, at around ±0.1. This can be explained as follows. Over the period of estimation (1963±94) the spread of supply shocks between France and Germany is 0 on average, but this is not the case for Germany and the United States over the same period. Hence, the stochastic constant might be capturing the fact that the difference between the German and US shocks is not null over the sample (Figure 2.3).

28 Structural Asymmetries in the European Union

1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993

Figure 2.3 Supply shock differentials for Germany and the United States, 1963±93

Figure 2.4 shows quite clearly the convergence process that has taken place since the beginning of the 1980s: declines from a quite high value of 0.56 at the beginning of the period to 0.33 at the end of the period. However, this ongoing process seems to have bottomed out and stabilised in the late 1980s. The coef®cient is not equal to zero, so there is no complete convergence at the end of the period. However, we can see that there are convergence forces at work. This is quite an important result as it would not have been possible to detect it with a classical method. The advantage of using time-varying parameters is that we can conclude in favour of a dynamic process, whereby supply shocks between France and Germany are becoming more symmetrical. But the stabilisation of the

Figure 2.4

Estimated (Equations 2.5 and 2.6), 1974±94

Laurence Boone 29

process also raises questions. Might recent policies be responsible for this halt in the convergence process? What factors condition convergence, and how could they be helped by policy? 2.5.2

Supply shocks

The results are presented below. First we assess convergence on Germany and then on the core group. Table 2.2 con®rms that the core group (Belgium, Austria, the Netherlands and, to a lesser extent, France) have converged on Germany, even when German reuni®cation in 1991 is taken into account, although this convergence process is not completed. But more importantly, Table 2.2 shows that Italian supply shocks converged on German ones throughout the whole period. It is also remarkable that Spain and Portugal attained an equal amount of convergence (with at around 0.3) to that of the `core'. In the Scandinavian countries the situation is quite diverse. A strong process of convergence seems to be at work in Denmark, but Sweden and Finland remain at a fair, although very stable, level of convergence, above the core countries. Only Greece and the UK show no evidence of convergence. This is quite different from Bayoumi and Eichengreen's (1996) results. Table 2.2 seems to suggest that for some countries the attractive power of Germany has stabilised since German reuni®cation, which is con®rmed by the ®gures in Appendix 2.2. The question that immediately follows is: could the power of Germany as an attractor have declined and/or been mitigated by an increase in the attractive power of the other core countries? To measure such an effect, we estimated Equations 2.5 and 2.6 with the core as the attractor rather than Germany (Table 2.3).13 Table 2.3 shows no evidence of convergence towards the core, with the exception of Belgium, whose approaches 0.5 at the end of the period. However, it is worth noting that appears to be signi®cantly below 1 for the core countries, Germany, Italy and Ireland, while it is signi®cantly above 1 for Greece. To sum up, this method provides new evidence of the direction some countries are taking. Hence, Italy shows as much convergence towards Germany, with respect to supply shocks, as does France. Furthermore, Spain, Portugal, Ireland and Denmark appear to have converged very signi®cantly too, especially in the recent past. Yet Greece and the UK clearly differ from the rest of Europe regarding the convergence of supply shocks. The ERM may have played a role in triggering greater integration, because for those countries that ®rst joined the ERM in 1979 the coef®cient appears signi®cantly lower after 1979 than it was before the

30 Structural Asymmetries in the European Union Table 2.2 Supply shock convergence on Germany as opposed to the rest of the world, 1974±9414 Bayoumi & Eichengreen (1963±93) France

0.69

United Kingdom

0.82

Italy

0.78

Spain

0.80

Belgium

0.65

Netherlands

0.71

Austria

0.68

Portugal

0.92

Greece

1.02

Finland

1.03

Ireland

1.03

Denmark

0.64

Sweden

0.85

1974±79 0.53 (0.02) 0.53 (0.03) 0.55 (0.04) 0.38 (0.04) 0.43 (0.01) 0.51 (0.01) 0.44 (0.06) 0.39 (0.04) 0.81 (0.04) 0.27 (0.03) 0.23 (0.05) 0.93 (0.10) 0.36 (0.03)

1980±90 0.37 (0.04) 0.46 (0.02) 0.31 (0.11) 0.37 (0.10) 0.34 (0.14) 0.30 (0.10) 0.29 (0.07) 0.31 (0.08) 0.50 (0.18) 0.37 (0.05) 0.19 (0.11) 0.45 (0.33) 0.33 (0.04)

1991±94 0.34 (0.00) 0.46 (0.00) 0.25 (0.00) 0.29 (0.00) 0.26 (0.00) 0.25 (0.00) 0.23 (0.00) 0.27 (0.00) 0.45 (0.01) 0.38 (0.01) 0.10 (0.01) 0.32 (0.03) 0.36 (0.01)

Notes: The coef®cient is estimated from regressions (Equations 2.5 and 2.6) over the period 1963±94, where the attractor is Germany and the alternative is the rest of the world, represented by the United States. Standard deviations are in brackets. The Bayoumi and Eichengreen coef®cient is 1 minus the correlation coef®cient so as to make the comparisons between the results more easily readable. So the lower the coef®cient, the higher the correlation.

ERM existed. With regard to the other countries, it is probably too early to see any signi®cant impact yet.15 This analysis also con®rms the anchor role played by Germany, despite reuni®cation, as without Germany the core does not seem to display any attractive power. It also means that the EU countries have made an effort to converge on Germany, but the reverse does not appear to the case.

Laurence Boone 31 Table 2.3 Supply shock convergence on the core as opposed to the rest of the world, 1974±9416

Germany France United Kingdom Italy Spain Belgium Netherlands Austria Portugal Greece Finland Ireland Denmark Sweden

1974±79

1980±90

1991±94

0.95 (0.02) 0.77 (0.02) 0.93 (0.05) 1.06 (0.05) 1.17 (0.40) 0.92 (0.07) 0.87 (0.02) 1.03 (0.06) 0.97 (0.05) 1.18 (0.13) 0.99 (0.03) 1.20 (0.07) 1.28 (0.08) 0.90 (0.04)

0.76 (0.08) 0.77 (0.02) 0.84 (0.04) 0.78 (0.12) 0.71 (0.31) 0.55 (0.27) 0.78 (0.04) 0.66 (0.10) 0.93 (0.04) 0.75 (0.21) 0.93 (0.03) 0.82 (0.19) 0.69 (0.29) 0.88 (0.02)

0.77 (0.05) 0.76 (0.01) 0.81 (0.02) 0.71 (0.06) 0.89 (0.26) 0.52 (0.13) 0.76 (0.01) 0.73 (0.15) 0.91 (0.01) 1.01 (0.15) 0.91 (0.01) 0.75 (0.08) 0.93 (0.24) 0.88 (0.01)

Notes: the coef®cient is estimated from regressions (Equations 2.5 and 2.6) over the period 1963±94, where the attractor is the core (France, Belgium, Austria and the Netherlands) and the alternative is the rest of the world, represented by the United States. Standard deviations are in brackets.

2.5.3

Demand shocks

We now turn to the results for demand shocks. Table 2.4 provides contrasting information on convergence towards Germany. The situation is quite different from that described by Bayoumi and Eichengreen (1996). Hence, France, Belgium and the Netherlands seem to have been quite close during the 1980s but there is evidence of a structural break in the early 1990s that led to wide divergence, with reaching some quite high levels, especially for France and Belgium. Among the core countries, only

32 Structural Asymmetries in the European Union Table 2.4 Demand shock convergence on Germany as opposed to the rest of the world, 1974±9417 Bayoumi & Eichengreen (1963±93) France

0.72

United Kingdom

0.88

Italy

0.63

Spain

0.99

Belgium

0.80

Netherlands

0.81

Austria

0.65

Portugal

0.76

Greece

0.83

Finland

1.05

Ireland

1.03

Denmark

0.76

Sweden

0.75

1974±79 0.29 (0.13) 0.46 (0.06) 0.36 (0.05) 0.70 (0.09) ±0.11 (0.15) 0.46 (0.03) 0.34 (0.02) 0.60 (0.03) 0.62 (0.01) 0.50 (0.01) 0.65 (0.03) 0.78 (0.01) 0.40 (0.04)

1980±90 0.13 (0.25) 0.45 (0.12) 0.33 (0.01) 0.34 (0.08) 0.30 (0.37) 0.37 (0.02) 0.28 (0.03) 0.49 (0.04) 0.57 (0.02) 0.51 (0.02) 0.56 (0.02) 0.80 (0.02) 0.36 (0.02)

1991±94 0.69 (0.00) 0.83 (0.02) 0.35 (0.01) 0.61 (0.05) 0.87 (0.08) 0.40 (0.01) 0.32 (0.02) 0.47 (0.01) 0.61 (0.01) 0.55 (0.01) 0.58 (0.00) 0.76 (0.00) 0.37 (0.01)

Notes: the coef®cient is estimated from regressions (Equations 2.5 and 2.6) over the period 1963±94, where the attractor is Germany and the alternative is the rest of the world, represented by the United States. Standard deviations are in brackets. The Bayoumi and Eichengreen coef®cient is 1 minus the correlation coef®cient so as to make the comparisons between the results more easily readable. So the lower the coef®cient, the higher the correlation.

Austria remains with a low over the last subsample. This seems to suggest that the impact of the reuni®cation shock on the overall German economy might have had quite a strong in¯uence on the rest of Germany's close partners with respect to demand, leading to policy action that was different from that of Germany's. The more northern countries present a different picture ± there is no evidence of convergence and the coef®cient remains at a high level throughout the period (apart from Sweden and, to a lesser extent, Finland).

Laurence Boone 33 Table 2.5 Demand shock convergence towards the core as opposed to the rest of the world, 1974±9418

Germany France United Kingdom Italy Spain Belgium Netherlands Austria Portugal Greece Finland Ireland Denmark Sweden

1974±79

1980±90

1991±94

0.73 (0.00) 0.68 (0.06) 0.92 (0.01) 0.58 (0.07) 0.79 (0.03) 0.47 (0.06) 0.77 (0.01) 0.68 (0.01) 0.85 (0.01) 0.94 (0.00) 0.67 (0.04) 1.36 (0.90) 0.98 (0.01) 0.84 (0.01)

0.71 (0.01) 0.61 (0.07) 0.88 (0.02) 0.76 (0.15) 0.73 (0.03) 0.60 (0.14) 0.72 (0.01) 0.66 (0.02) 0.79 (0.02) 0.90 (0.01) 0.90 (0.14) 0.72 (0.31) 0.98 (0.02) 0.82 (0.01)

0.73 (0.01) 0.67 (0.03) 0.88 (0.01) 0.87 (0.04) 0.74 (0.01) 0.84 (0.07) 0.73 (0.01) 0.68 (0.00) 0.79 (0.01) 0.89 (0.00) 1.20 (0.16) 0.96 (0.37) 1.00 (0.00) 0.83 (0.00)

Notes: the coef®cient is estimated from regressions (Equations 2.5 and 2.6) over the period 1963±94, where the attractor is the core (France, Belgium, Austria and the Netherlands) and the alternative is the rest of the world, represented by the United States. Standard deviations are in brackets.

In the south, Spain started a convergence process in the early 1980s, but this ended in the mid-1980s. Portugal's convergence path has been more persistent, although it stagnated in the last period. However, the s remain at too high a level to talk about convergence. Finally, Greece shows no sign of convergence. The results in Table 2.5 are very close to the ones obtained from the analysis of supply shocks. The core countries (except, surprisingly, Belgium) are closer to each other than to the rest of the EU ( is around 0.7, whereas it is never below 0.8 for any other country), but this is not

34 Structural Asymmetries in the European Union

enough to consider them as a global attractor as is far too high to allow one to talk about convergence. Overall, there appears to be less evidence of convergence in terms of demand shocks, and the situation does not seem to have evolved very much during the period in question. As demand shocks are supposedly induced or potentially corrected by monetary and ®scal policy, this conclusion is quite worrying. In EMU there will be a common monetary policy and the use of ®scal policy will be restricted.

2.6

Conclusion

This chapter has tried to assess whether there is increasing symmetry in the shocks faced by the EU economies. The originality of this research is that it uses a dynamic measure that allows structural changes to be taken into account. Hence, we can assess whether there is a process of integration at work, and how it has been affected by German reuni®cation. Furthermore, instead of looking at nominal versus GDP convergence, we have focused on the structure of the asymmetries; the variables of interest being supply and demand disturbances obtained through VAR decomposition. The results are in line with previous research in the sense that they con®rm that the so-called core countries (France, Belgium, Austria and the Netherlands) display greater symmetry in the supply shocks they face. But the analysis in this chapter goes beyond this simple evidence in several ways. Regarding supply shocks, we have shown that all countries but Greece and the UK display a pattern of supply disturbances that look increasingly similar to that of Germany. On the other hand, we have also shown that the direction of the integration process has always been focused on Germany alone. Hence, there is a strong asymmetry in the way that the EU economies are converging. Finally, the analysis suggests that the ERM has been a factor in integration. This has several policy implications. First, all countries but Greece and the UK should have low adjustment costs when they enter into monetary union with Germany. Thus, according to OCA theory, they probably have an incentive to join the single currency. On the other hand, there seems to be a disequilibrium in the balance of power between these economies and Germany. However, this could disappear after monetary union, as the convergence process is likely to be accelerated and deepened (see De Grauwe, 1995).

Laurence Boone 35

With respect to demand shocks, this chapter has shown that, although the distinction between the core and the periphery is still relevant, it is very weak. Furthermore, there does not seem to have been any progress in the convergence process since at least the mid-1980s. As demand shocks may be driven and corrected by monetary and ®scal policy, this is worrying in two senses: (1) monetary policy will be common and ®scal policy constrained, and (2) it reveals a dramatic lack of policy coordination between the countries of the EU. This chapter suggests some direction for future research. In particular, while we have provided evidence of the dynamic evolution of the asymmetry of shocks, we have not analysed structural asymmetries in the transmission of these shocks. This might be interesting not only in terms of assessing the persistence of the effects of shocks, but also for suggesting suitable policy reactions.

Appendix 2.1 The Kalman ®lter and the estimation of time-varying parameters The roots of the Kalman ®lter in economics lie in the rational expectations hypothesis, augmented by Friedman (1979). Friedman suggests that, given the true model yt = xt A + ut (ut = white noise), agents may not know it but use information as it becomes available to update the estimate of A. Hence they effectively estimate a time-varying parameter. Cuthberson, et al. (1992) extend this framework to include the case where (1) agents have some prior information about the initial value of A, and (2) this coef®cient is allowed to vary stochastically. Such a model may be analysed with the Kalman ®lter. In this context it can be interpreted as a form of adaptive expectation, where in each period the adjustment parameter is updated, based on new information. Let us consider model (2.3) from the main text (the measurement equation): j

…Xt

j

Xit † ˆ t ‡ …Xt

Xt † ‡ "t †

…2:3†

and the transition equations: t ˆ t

1

‡ 1t

t ˆ t

1

‡ 2t

These can be rewritten in matrix terms as follows: Measurement equation: Yt ˆ At Zt ‡ "t

…2:4†

36 Structural Asymmetries in the European Union

Transition equation: At ˆ TAt 1 ‡ t where Yt ˆ Xjt

8 9 >. . . 1 . . .> > Zt ˆ > : j ;; Xt Xt

Xit ;

At ˆ …t ; t †;

8 9 1 . . . 0> : ;and Tt ˆ > 0...1

it ˆ …1t ; 2t †: We assume that "t is normal with mean 0 and variance Vt , and that t is also normal with mean 0 and variance/covariance matrix Qt . We also assume that "t and t are uncorrelated (for all t), that At 1 is uncorrelated with "t and independent of the error term t . The thinking behind the Kalman ®lter, then, is fairly simple. Given an initial estimate for A0 and for its covariance matrix P0 (assuming we know Vt and Qt , and that we have observations for Yt , the Kalman ®lter provides optimal forecasts of the unobserved At (t = 1 . . ., T), in the sense of the minimum mean square error (MMSE) criterion. Assuming further that all disturbances are normal, then the Kalman ®lter provides the maximum likelihood estimator of At . Let us assume that we have guessed a value for At 1 and Pt 1 , and that we know Vt and Qt . At time t±1, it is then possible to infer value of A at t and its covariance matrix Q, based on the information available at t ± 1 (this holds for any t = 1, . . ., T). These are called prediction equations: A1t Pt=t

1

ˆ Tt At

1

1

ˆ Tt Pt

1

‡ Qt

From these we can infer a value for Yt=t 1 and compare it with the true value, Yt . The difference between them is the one-step-ahead prediction error: t ˆ Yt

Yt=t

1

ˆ Yt

Zt At=t

1

its covariance matrix being: Ft ˆ Zt Pt=t

1 Zt

‡ Vt

When new information becomes available at time t it is possible to update the above predictions: At ˆ At=t Pt ˆ Pt=t

1 1

Pt=t

0 1 0 1 Zt Ft Zt …yt

Zt0 At=t

Pt=t 1 Zt0 Ft 1 Zt Pt=t 1

These are called the updating equations.

1†

Laurence Boone 37

The ®lter works like this, recursively through time. Once the predictions are updated, the updated estimate can be used to work out predictions for t + 1, and so on until the end of the sample. The Kalman ®lter, then, is a simple algorithmic procedure if we know the variance/covariance matrices of the error processes (Vt and Qt ) and the parameters (Tt ) in the state space form. However, in practice we are ignorant of the value of the so-called hyperparameters, Vt and Qt . But in using the Kalman ®lter we can assume any initial value for these matrices, and derive recursive values for At , Pt and vt conditional on these initial guesses. These can then be fed into a likelihood function, which will give us new initial estimates to be plugged into the Kalman ®lter algorithm. The process goes on recursively until the likelihood function reaches its maximum. Then we obtain the ML estimate of all these parameters. It can be shown (see Cuthberson et al. 1992; Harvey, 1992) that the likelihood function is a function of the prediction errors: log L ˆ

T 1X logjFt j 2 tˆ1

T 1X u t Ft 1 y t 2 tˆ1

where v is the prediction error de®ned above. The conditional mean Yt=t 1 is also the MMSE of Yt , while the N  1 vector vt can be interpreted as a vector of prediction errors. Hence, the likelihood function above is said to be the prediction error decomposition form of the likelihood.

Appendix 2.2

Representation of for supply and demand shocks, 1974±94 38

Figure A2.1 Supply shocks, 1974±94

39

Figure A2.2 Demand shocks, 1974±94

40 Structural Asymmetries in the European Union

Notes 1. I wish to thank T. Bayoumi and B. Eichengreen, who very kindly provided the results of their 1996 analysis. This chapter has bene®ted from valuable discussions with Jean Pisani-Ferry, Claire Lefebvre, Benoit Mojon, Stephanie Guichard, Loic Cadiou, Isabelle Bensidoun, Virginie Coudert and participants at an internal seminar at the CEPII at the XIV Journees d'Economie Mon Taire et Bancaire, Orleans, 1977, and at the ESEM and EEA Annual Conference in Toulouse, September 1997. The usual disclaimers apply. 2. Note that a symmetrical shock will also generate persistent disequilibria if the degree of labour mobility or wage and price ¯exibility varies between countries; that is, in the presence of structural asymmetries. 3. However, the transmission channels of monetary shocks might still be of interest since they could differ from one country to another, even under a single currency, see Barran et al. (1996) and Demertzis et al. (1996). 4. A (negative) demand shock will trigger a fall in output, which causes unemployment to rise. Three types of action are then possible: (1) a price adjustment: wages and prices are ¯exible and fall far enough to trigger a new equilibrium; (2) quantity adjustment: the mobility of labour leads the newly unemployed to migrate elsewhere; and (3) policy action: a ®scal policy is implemented to offset the demand shock. 5. The only potential bene®ts for which measurements are available are the transaction cost savings, which were evaluated by the European Commission (1990) and amounted to 0.1 per cent of GDP for both France and Germany (0.4 per cent for the whole of the EU). Pisani-Ferry (1996) reckons that the lack of empirical analysis on the bene®ts of EMU is due to the complexity of the task. In effect, while studying costs only involves macroeconomic analysis, bene®ts may involve both macroeconomic (monetary) and microanalysis. 6. The diversity in adjustment channels to a shock is still subject to debate. A representative study has been conducted by Erkel-Rousse and Melitz (1995). 7. They used three ®ltering methods to extract the business cycle components and found that their results were insensitive to the ®ltering method. 8. For reference, see also Chamies et al. (1994) and Erkel-Rousse and Melitz (1995). 9. Empirical works on structural convergence focus on the structure of wage formation. See for example, Artis and Nachane (1990), Artis and Ormerod (1991), Barrel et al. (1990), Anderton et al. (1992) and Malgrange (1995). 10. Campbell and Mankiw (1989), and Bernard and Durlauf (1991) use such tests to study the question of GDP convergence, while Hakkio and Rush (1989) analyse the convergence of nominal variables (exchange rates, interest rates and in¯ation rates) with these methods. 11. A more detailed interpretation of  is given in Section 2.5, as this chapter studies the special case of shocks, which is peculiar compared with the usual economic time series. 12. A technical explanation of each equation is not provided here, but is available from the author on request. 13. When convergence of one of the core countries was being measured, it was removed from the core in order to avoid any correlation problems. 14. A dummy was introduced, taking the value of 1 in 1991 and zero otherwise, to allow for German reuni®cation.

Laurence Boone 41 15. All countries entered the ERM in 1979 except for Spain (1989), the UK (1990± 92), Portugal (1992), Austria (1995) and Finland (1996), which joined after the sample period. Sweden and Greece are still outside the ERM. 16. For the equations to pass the usual diagnostic tests, a dummy was introduced, taking the value of 1 in 1987 for Spain and Portugal, in 1986 for Austria, in 1991 for Finland and in 1992 for Sweden, and zero otherwise. However this did not signi®cantly affect the value of the coef®cient , nor the standard deviation. 17. As for the tables above, it was necessary to introduce dummies, which took the value of 1 in 1987 for Spain, in 1993 for Portugal, in 1981 for Denmark and in 1991 for Greece. The value of the coef®cient was not signi®cantly altered. 18. Similarly to the above tables, and without affecting the convergence paths, some dummies took the value of 1 (in 1987 for Spain and the UK, and in 1991 for Greece).

References Anderton, R., R. Barrel and J. McHugh (1992) `Nominal convergence in European Wage Behaviour: Achievements and Explanations', in R. Barrel (ed.) pp. 31±57. Ardeni, P. G. (1992) `On the Way to EMU: Testing Convergence of the European Economies', Economic Notes di Monte des Paschi di Sienna, vol. 21, pp. 238±57. Artis, M. and D. Nachane (1990) `Wages and Prices in Europe, a Test of German Leadership Hypothesis', Weltwirtschaftliches Archiv, vol. 126, pp. 59±77. Artis, M. and P. Ormerod (1991) `Is there an EMS Effect in European Labour Markets?', CEPR Discussion Paper, no. 598 (London: CEPR, December). Artis, M. and W. Zhang (1995) `International Business Cycles and the ERM: Is There a European Business Cycle?', CEPR Discussion Paper, no. 1191 (London: CEPR, August). Barran, F., V. Coudert and B. Mojon (1996) `The Transmission of Monetary Policy in the European Countries', Financial Market Group Special Paper, no. 86. Barrel, R., J. Darby and C. Donaldson (1990) `Structural Stability in European Wage and Price Systems', NIESR Discussion Paper, no. 188. Barro, R. J. (1991) `Economic Growth in a Cross-Section of Countries', Quarterly Journal of Economics, vol. 106, no. 2 (May), pp. 407±43. Barro, R. and X. Sala-i-Martin (1990) `Economic growth and convergence across the United States', NBER Working Paper, no. 3419 (Cambridge, MASS: NBER). Bayoumi, T. and B. Eichengreen (1993a) `Shocking Aspects of European Monetary Uni®cation', in F. Torres and F. Giavazzi (eds), Adjustment and Growth in the European Monetary Union (Cambridge: Cambridge University Press), pp. 193±230. Bayoumi, T. and B. Eichengreen (1993b) `Is There a Con¯ict between EC Enlargement and European Monetary Uni®cation?', Greece Economic Review, vol. 15: no. 1, pp. 131±54. Bayoumi, T. and B. Eichengreen (1994) `One Money or Many? On Analysing the Prospects for Monetary Uni®cation in Various Parts of the World', Princeton Essays in International Finance, no. 76 (Princeton, NJ: Princeton University, International Finance Section). Bayoumi, T. and B. Eichengreen (1996) `Operationalizing the Theory of Optimum Currency Areas', mimeo (London: CEPR). Bayoumi, T. and P. R. Masson (1992) `Fiscal Flows in the United States and Canada: Lessons for Monetary Union in Europe', mimeo (Washington, DC: IMF).

42 Structural Asymmetries in the European Union Bayoumi, T. and E. Prasad (1995) `Currency Unions, Economic Fluctuations, and Adjustment: Some Empirical Evidence', mimeo (Washington, DC: IMF). Bayoumi, T. and A. Thomas (1995) `Relative Prices and Economic Adjustment in the US and EU: A Real Story About EMU3', IMF Staff Papers, no. 42 (Washington, DC: IMF, March), pp. 108±33. Beine, M. (1995) `Asymmetric Shocks Inside Future EMU: Going Beyond Real Exchange Rate Variability', mimeo (Services des EÂtudes et de la Statistique, MinisteÁre de la Region Wallone). Bernard, A. B. (1990) `Empirical implications of the convergence hypothesis', mimeo (Stanford, CA: Stanford University). Bernard, A. B. and S. N. Durlauf (1990a) `Convergence in international output movements', mimeo (Stanford, CA: Stanford University). Bernard, A. B. and S. N. Durlauf (1990b) `A Test for Convergence across National Economies', mimeo (Stanford, CA: Stanford University). Blanchard, O. J. and D. Quah (1989) `Dynamic Effects of Aggregate Demand and supply Disturbances', American Economic Review, vol. 79, pp. 655±73. Blanchard, O. J. and L. F. Katz (1992) `Regional Evolutions', Brookings Papers on Economic Activity, no. 1 (Washington, DC: Brookings Institute). Buiter, W. H. (1995) `Politique MacroeÂconomique dans la Transition vers l'Union Monetaire', Revue d'Economie Politique, no. 5 (Sept.±Oct.) Campbell, J. Y and N. G. Mankiw (1989) `International Evidence on the Persistence of Fluctuations', Journal of Monetary Economics, vol. 23, pp. 319±33. Campbell, R. T. and T. Bollersey (1989) `Common Stochastic Trends in a System of Exchange Rates', Journal of Finance, vol. XLIV, pp. 167±81. Chamies, N., A. Deserrdes and R. Lalondes (1994) `Optimum Currency Areas and Shock Asymmetry: A Comparison of Europe and the United States', Bank of Canada Working Paper, no. 94/1 (Bank of Canada). Christodoulakis, N., S. P. Dimelis and T. Kollintzas (1995) `Comparisons of Business Cycles in the EC: Idiosyncracies and Regularities', Economica, vol. 62, pp. 1±27. Commission of the European Communities (1991) `The Economics of EMU', European Economy, Directorate General for Economic and Financial Affairs, Special Edition, no. 1 (Brussels: European Commission). Cuthberson, K., S. G. Hall. and M. P. Taylor (1992) Applied Econometric Techniques (Brighton: Harvester Wheatsheaf). De Grauwe, P. (1992) The Economics of Monetary Integration (Oxford: Oxford University Press). De Grauwe, P. (1994) `Alternative Strategies Towards Monetary Union', mimeo (London: CEPR). De Grauwe, P. (1995) `Monetary Union and Convergence Economics', mimeo (London: CEPR and University of Leuven). Demertzis, M., A. Hughes Hallett, and O. Rummel (1996) `Is a 2-speed System in Europe the Answer to the Con¯ict between the German and the Anglo-Saxon Models of Monetary Control?', mimeo (Glasgow: University of Strathclyde). Erkel-Rousse, H. and J. Melitz (1995) `New Empirical Evidence on the Costs of European Monetary Union', INSEE D.T, no. 9516. European Commission (1990) various publications. Fatas, A. (1996) `EMU: Countries or Regions?', Paper presented at the EEA meeting in Istanbul, August 1996.

Laurence Boone 43 Funke, M. (1995) `Europe's Monetary Future: One Market, One Money', CEF, LBS Discussion Paper, no. 1±95. Giavazzi, F. and A. Giovannini (1996) Exchange Rates and Prices in Europe, pp. 593±605. Gros, D. (1996), `A Reconsideration of the Optimum Currency Area Approach: The role of External Shocks and Labour Mobility', mimeo (London: CEPR). Hakkio, C. S and M. Rush (1989) `Market Ef®ciency and Cointegration: An Application to the Sterling and Deutschmark Exchange Markets', Journal of International Money and Finance, vol. 8, pp. 75±88. Haldane, A. G. and S. G. Hall (1991) `Sterlings' Relationship with the Dollar and the Deutschmark: 1986±89', Economic Journal, vol. 101, pp. 436±43. Hall, S. G., D. Robertson and M. Wickens (1992) `Measuring Convergence of the EC economies', CEF Discussion Paper, no. 1±92 (London: London Business School). Hall, S. G., D. Robertson and M. Wickens (1992) `Measuring convergence of the EC Economies', The Manchester School, LX supplement (June), pp. 99±111. Hallet, A. J. H and Y. Ma (1995) `Economic Cooperation within Europe: Lessons from the Monetary Arrangements in the 1990s', CEPR Discussion Paper, no. 1190 (London: CEPR). Hallet, A. J. H. and D. Vines (1990) `Adjustment dif®culties within a European Monetary Union: an Analysis and a Comparison with Alternative Regimes', mimeo (London: CEPR). Harvey, A. C. (1992) Forecasting, Structural Time Series Models and the Kalman Filter (Cambridge: Cambridge University Press). Henin P. Y and C. Aigueperse (1994) `La convergence reÂelle des eÂconomies europeÂennes: EÂleÂments d'eÂvaluation et contribution de la politique budgetaire' (CEPREMAP, January). Henin, P. Y and Y. Le Pen (1994) `Les pisodes de la convergence EuropeÂenne' (CEPREMAP, September). Italianer, A. and J. Pisani-Ferry (1992) `Regional Stabilization Properties of Fiscal Arrangements: What Lessons for the Community', mimeo (CEPS conference). Kenen, P. B. (1969) `The Theory of Optimum Currency Areas: An Eclectic View', in R. Lou®r and L. Reichlin (1993), `Convergence nominale eÂte reÂelle parmi les pays de la Communaut europeÂenne et de l'AELE', Observations et Diagnostics EÂconomiques, vol. 43, pp. 69±92. Malgrange, P. (1995) `Y a-t-il convergence des modes de formation des salaires en Europe? Une eÂtude comparative de l'Allemagne, des Pays-Bas et de la France', Document presented at l'AFSE, Nantes, May. Mankiw N., D. Romer and D. Weil (1992) `A Contribution to the Econometrics of Economic Growth', Quarterly Journal of Economics, vol. 107. Masson, R. P. and S. Symansky (1992) `Evaluating the EMS and EMU using Stochastic Simulations', IMF working paper (Washington, DC: IMF). McKinnon, R. I. (1963) `Optimum Currency Areas', American Economic Review, vol. 53, pp. 717±25. Melitz, J. (1996) `The Evidence about the Costs and Bene®ts of EMU', mimeo (London CEPR). Mundell, R. A. (1961) `The Theory of Optimum Currency Areas', American Economic Review, vol. 51, pp. 657±65.

44 Structural Asymmetries in the European Union Pisani-Ferry, J. (1996) Comment on Bayoumi and Eichengreen (forthcoming). Stockman, A. C. (1988) `Sectoral and National Aggregate Disturbances in Industrial Output in Seven European Countries', Journal of Monetary Economics, vol. 21, pp. 387±410. Thygesen, N. (1992) `Wage Determination and ¯exibility: the Requirements for EMU?', Note for the CEPS Economic Policy Group. Vinals, J. (1996) `European Monetary Integration: a Narrow or a Wide EMU', mimeo (Banco de Espan Ä a and CEPR). Vinals, J. and J. F. Jimeno (1996) `European Unemployment and Monetary Union', mimeo. Whitley, J. D. (1992) `Common Policy Simulations', mimeo (Warwick: University of Warwick).

3

Testing for Common Features in the European Union Hong Bai, Stephen Hall and David Shepherd

3.1

Introduction

As the prospect of full monetary integration in the European Union (EU) draws closer to reality, attention has focused on the dif®culties that individual member states are likely to face after the adoption of a single currency and a common monetary policy. The implications of economic and monetary union in Europe have been discussed at length by numerous authors (see, for example, Eichengreen, 1993; Goodhart, 1995; Shepherd, 1990). As far as macroeconomic policy is concerned, one of the key questions is whether the various economies have achieved, or can achieve, a degree of convergence suf®cient to ensure the smooth operation of a uni®ed monetary system. The answer to this question hinges on the nature of the shocks that affect each country and how they are absorbed in the absence of exchange rate changes and independent monetary policy adjustments. If wages and prices were perfectly ¯exible across all member states, and if labour and capital were highly mobile, the adoption of a single currency would in itself pose few problems for macroeconomic policy. Under these conditions it would, in principle, be possible to achieve the appropriate adjustment to any demand or supply shocks via domestic wage and price movements and labour and capital ¯ows, rather than exchange rate realignments and independent monetary policy changes. In practice it seems implausible to suppose that labour and product markets in the EU exhibit the above degree of ¯exibility, so the important question to be considered is whether recurring macroeconomic shocks are broadly symmetrical or asymmetrical across the member states. Generally speaking, the pressures that lead to con¯icts over the design and implementation of macroeconomic policy in a common currency

45

46 Testing for Common Features in the EU

area are greatly reduced if the participating countries face shocks that are broadly symmetrical, in the sense that they share a common time pro®le. Even in this case it is important to recognise that the relative magnitude of the shocks may differ across countries, and that there are always likely to be disagreements about the required strength of any policy response ± for example, the extent to which monetary policy should be tightened in response to rising in¯ation. Nevertheless, the presence of symmetry does at least imply an opportunity to secure general agreement on the appropriate direction of policy. In contrast, if the underlying shocks are asymmetrical (affecting different groups of countries at different times), it becomes extremely dif®cult to design policies that are uniformly appropriate to the needs of all and the potential for con¯ict is that much greater. An obvious example is the con¯ict over monetary policy that could arise if some countries entered a recession while others entered a boom. When we consider the shocks that affect any group of economies, we are concerned not so much with the precise nature of the shocks themselves but with how they are manifest in the form of time movements in the key macroeconomic variables, such as output, employment, interest rates and the price level. The degree of convergence between the economies in question can then be identi®ed by examining the degree of comovement across the relevant time series. This problem can be approached from a long-term perspective by considering the relationship between trend movements in the relevant variables. The degree of convergence over the long term can then be examined with the aid of cointegration tests or, more ¯exibly, time-varying parameter techniques that allow for underlying changes in structure. For example, Hall et al. (1993) use the time-varying parameter formulation, plus estimation by Kalman ®lter methods, to examine the degree of long-term convergence exhibited by nominal exchange rates, nominal interest rates and in¯ation rates across the EU countries. An alternative approach is to consider how trend deviations in the variables are correlated across countries (Backus et al. 1995). The correlation coef®cients then give some indication of the degree of comovement in the series in the short term, but say little about the underlying dynamics. In this chapter we concentrate on the short-term dimension of the convergence issue by examining the extent to which the larger EU economies exhibit a common cyclical response to macroeconomic shocks. Our analysis is based on the work of Engle and Kozicki (1990) and Vahid and Engle (1993a). Starting with the idea that many economic time series exhibit features such as serial correlation, heteroscedasticity

Hong Bai, Stephen Hall and David Shepherd 47

and seasonality, they develop a test to determine whether the identi®ed features are common across any particular series. Although this common features test was initially used to examine the synchronisation of output growth over the cycle for the G7 countries, it can also be applied to different sectors within an economy. For example, Lippi and Reichlin (1994) and Vahid and Engle (1993b) examine the existence of common trends and cycles to determine whether permanent and transitory components of income and consumption are positively related to the rate of output growth. Returning to our own contribution, we begin by describing the nature of the common features test and then use it to investigate whether some of the key macroeconomic variables across the larger EU countries exhibit common cyclical features, even in the absence of full convergence. Our results suggest that a common cycle can be identi®ed for GDP, nominal exchange rates, nominal interest rates and real exchange rates. In contrast, real interest rates do not appear to share this common feature. The chapter ends with a brief discussion of the implications of our results.

3.2

The common features test

The common features test provides a means of determining the degree of comovement between a set of variables that may or may not exhibit cointegration. If a set of I(1) variables are related in a stochastic macroeconomic model, it is well known that the trend component can be written as the sum of an I(1) trend and an I(0) cycle in an in®nite number of ways. These non-stationary variables are said to be cointegrated when a linear combination of them generates a stationary (white noise) error term. In this case the cointegrating vector effectively represents the common feature that explains the comovement in the variables that renders the combined series stationary. The strength of the common features test is that it can also be applied to vector autoregressive systems that are not cointegrated. The test then gives an indication of the degree of comovement in the short term, even in the absence of full cointegration. For example, the variables in question may not be cointegrated, but they may each exhibit a feature such as serial correlation. The common features test can then be used to determine whether the pattern of serial correlation is similar, indicating a degree of short-term comovement in the variables. Vahid and Engle (1993b) suggest that a `serial correlation common feature exists when a linear combination of the series is unpredictable relative to the past history of the variables'. Note, however, that while serial correlation is used as the

48 Testing for Common Features in the EU

example, the test is quite general and the common feature could be one of the other statistical properties of the series or a common exogenous factor. The important point is that the common features test provides a measure of the comovement of sets of time series that is less stringent than cointegration. To see the rationale behind the common features test, let us suppose that we have two time series, y1t and y2t and that a time-varying feature, Wt , is present in both series. We assume that the error terms are independent and that Wt represents some form of serial correlation. 8 9 8 9 8 9 e y1t > >> > > : 1t > ; …3:1† : > ; ˆ : ;Wt ‡ > y2t e2t 1 The ®rst step in the common features test is to establish whether or not a particular feature is present in both sets of data (if the feature is present in only one of the series, there is clearly no point in testing for commonality). Having established the existence of a particular feature, the next step is to determine whether it is common. The obvious way to do this is to see whether there exists some combination of the original series that removes the identi®ed feature. In our example, Wt represents a form of serial correlation and the test for a common feature is essentially a test of whether it is present in some linear combination of the original series, ut = y1t ± y2t . In other words, if the pattern of serial correlation is the same in both series (if it is a common feature) it should be cancelled out in the linear combination, leaving the combined series ut serially uncorrelated. On the other hand, if the pattern is signi®cantly different (if the feature is not common) the combined series ut is likely to exhibit serial correlation. A simple test procedure is to implement a version of the standard LM test using an auxiliary regression. We can test for the presence of a feature W in a given variable y by estimating a model of the form yt = c + Wt + et , where c is a constant and Wt may include the lag of yt or some other variables. If the series yt does not exhibit the feature we would expect to be insigni®cantly different from zero. The null hypothesis is then: H0 : ˆ 0 ! no feature Hl 6ˆ 0 ! feature exists If s(y) is the LM test statistic for the presence of a feature in a data series {yt }, the null hypothesis of no feature is rejected when PH0 [s(y) > c]  5 per cent. In other words, we accept that the feature exists if s(y) > c in the critical region for c at the 5 per cent level. Note that the LM test statistic for this auxiliary regression can be calculated as LM = TR2 .

Hong Bai, Stephen Hall and David Shepherd 49

The above procedure can be used to determine whether the feature is present in both y1 and y2 . We can then test whether there is some combination of the series that eliminates the feature. The test proceeds by constructing a variable u, where u = y1 ± y2 , and  is chosen so as to minimise the test statistic, s(u). The distribution of s(u) over  satis®es a simple inequality: ^  S…yI S…u†

^ z †  min S…y1 y

^ > cŠ  P‰S…e1 PHt1 ‰S…u†



y2 †  S…y1

y2 † ˆ S…e1

e2 †

e2 † > cŠ  5%

^ is asymptotically distributed as chi square and  is The statistic s(u† approximated by , which can be estimated via limited-information maximum likelihood (LIML), two-stage least squares (2SLS) or direct minimisation of s(u). The signi®cance (or insigni®cance) of the estimate of  then indicates the presence (or absence) of the common feature. Of the three methods for estimating the signi®cance of , the most accurate is the direct minimisation of s(u). The 2SLS approach involves the regression of y1 on y2 or y2 on y1 and we would expect the resulting estimates to differ slightly from those obtained via LIML because of the different normalisations imposed in the 2SLS regressions (according to whether y1 or y2 is on the left-hand side). Although Vahid and Engle (1993a) suggest that 2SLS and LIML in practice give similar results, and that they have the advantage of computational ease, both of these approaches provide only an approximation of the true minimum obtained via direct minimisation. To get some idea of how the estimates from these different methods compare, we have used all three in our test procedures.

3.3

Testing for a common cycle in GDP

We begin by examining the behaviour of output relative to the trend for the six largest EU countries (Germany, France, Italy, Spain, Austria and the UK). Our objective is to determine whether common business cycle features can be identi®ed for these countries. The analysis uses quarterly real output data (seasonally adjusted) and we measure the trend deviations as deviations around a simple linear trend. There are of course other ®ltering methods available and it is worth noting that the results of this kind of exercise may be sensitive to the precise method employed (King and Rebelo, 1993). Table 3.1 shows the output growth correlation between the six countries. Although the correlation coef®cients are positive, on their own they do not provide suf®cient information to determine whether the

50 Testing for Common Features in the EU Table 3.1 Output growth correlation coef®cients, 1980±89

France Italy Spain UK Austria

Germany

France

Italy

Spain

UK

0.64 0.68 0.37 0.61 0.67

± 0.60 0.66 0.49 0.80

± ± 0.47 0.27 0.73

± ± ± 0.49 0.53

± ± ± ± 0.28

cycles in growth are common or unsynchronised. Before we consider whether a set of variables share a common feature, it is important to know something about the long-term structure of the data, and in particular whether the variables are cointegrated (whether they move together in the long term). In the present context, a simple procedure is to calculate the difference in real GDP between each country and Germany and then test for the presence of a unit root in the residual. Table 3.2 reports the relevant ADF statistics. The results suggest that there are no cointegrating relationships with respect to Germany, although France is close to the test borderline. The presence of a unit root in the residuals suggests that we can estimate a bivariate VAR(1) model in ®rst difference form. Common serial correlation features then indicate the degree of persistence and comovement in the system and can be interpreted as evidence of common business cycle features. The test for serial correlation is essentially a test of whether the past growth rates of GDP are signi®cant. In other words, if y1t and y2t represent the log of GDP for two countries, the presence of serial correlation is indicated by a signi®cant LM test statistic, calculated as TR2 in the regression
y1t = c + 
y2t (±1) +
y1t (±1) + et . The joint LM tests for a set of Bivariate VAR(1) models are presented in Table 3.3. Table 3.3 shows that, from the 30 least square regressions, 25 suggest evidence of a cycle by rejecting the hypothesis that both coef®cients together are insigni®cantly different from zero at the 5 per cent level. Ten of the possible 15 pairs show the presence of serial correlation in both

Table 3.2 Unit root tests for quarterly GDP with respect to Germany

ADF(1)

France (1967±89)

Italy (1970±89)

Spain (1970±89)

UK (1965±89)

Austria (1972±89)

±3.35

±2.72

±1.63

±1.33

±3.17

Note: Critical value for ADF(1) = ±3.40.

Hong Bai, Stephen Hall and David Shepherd 51 Table 3.3

Germany France Italy Spain UK Austria

Bivariate VAR(1) for GDP, 1980±89 Germany

France

Italy

Spain

UK

Austria

± 21.60 20.28 65.52 9.36 3.70

9.18 ± 24.25 77.35 9.44 3.62

8.58 15.04 ± 77.26 10.20 7.06

4.84 17.48 24.84 ± 10.40 6.81

7.45 4.35 30.08 77.35 ± 5.24

6.65 8.63 24.07 69.72 9.60 ±

Note: 5 per cent critical value with 2df = 5.99. The dependent variables are in the ®rst column.

countries. Given that there appears to be some sort of cycle in each country, the next step is to examine whether the cycles are a common feature. Following the procedure described by Equation 3.2, we derive a new variable (u) given from the function ut =
y1t + 
y2t , where  is chosen so as to minimise the statistic s(u) derived as TR2 from the auxiliary regression ut = c + ut 1 . The presence of a common serial correlation feature is then indicated when  is insigni®cantly different from zero (that is when s(u) < 3.84, which is the 5 per cent critical value of the chi-squared distribution for 1df). Table 3.4 presents the common feature estimations where  is obtained by direct minimisation of s(u). The insigni®cance of the  coef®cient in the ®rst column of Table 3.4 indicates the presence of a common business cycle feature for each country with respect to Germany (in the sense that there are common serial correlation features). In addition, the negative sign attached to the  coef®cient suggests that real GDP growth for these countries moves in a Table 3.4

The common features test for GDP growth, 1980±89 Germany

UK Austria Spain France Italy

^ s(u)  ^ s(u)  ^ s(u)  ^ s(u)  ^ s(u) 

0.00 0.46 0.00 ±0.14 0.00 ±0.59 0.00 ±3.04 0.00 ±1.69

Note: 5 per cent critical value = 3.84

UK ± ± 0.00 ±1.33 3.55 ±0.99 0.02 ±3.82 0.01 ±2.84

Austria

Spain

France

± ± ± ± 0.00 ±1.73 0.00 ±2.28 0.00 ±3.96

± ± ± ± ± ± 0.00 ±1.30 0.00 ±1.28

± ± ± ± ± ± ± ± 0.14 ±0.63

52 Testing for Common Features in the EU Table 3.5 2SLS common features test for GDP growth, 1980±89

Germany UK Austria Spain France Italy

^ s(u)  ^ s(u)  ^ s(u)  ^ s(u)  ^ s(u)  ^ s(`u') 

Germany

UK

Austria

Spain

France

Italy

± ± 1.67 ±0.03 0.02 ±0.06 2.66 ±0.007 0.31 ±0.06 0.22 ±0.09

0.05 ±0.24 ± ± 0.19 ±0.24 1.18 ±0.11 0.003 ±0.02 0.25 ±0.28

0.04 ±0.05 1.71 ±0.07 ± ± 1.86 ±0.05 0.02 ±0.02 0.12 ±0.03

0.10 ±0.58 0.91 ±0.29 0.20 ±0.71 ± ± 0.05 ±0.42 0.13 ±0.49

0.02 ±0.21 2.01 ±0.12 0.09 ±0.02 2.39 ±0.03 ± ± 0.13 ±0.03

0.14 ±0.52 2.09 ±0.11 0.01 ±0.04 2.51 ±0.01 0.04 ±0.09 ± ±

Note: 5 per cent critical value = 3.84.

procyclical manner with respect to Germany. The signi®cance levels reported for the other cross-country pairs also suggest the presence of common serial correlation (business cycle) features of a procyclical form. The results in Table 3.4 were obtained via direct minimisation of the s(u) function. By way of comparison, Tables 3.5 and 3.6 show the results of the common features test using the LIML and 2SLS approximations, where the instruments are the constant and lags of GDP growth in the two countries. Although the results in Tables 3.5 and 3.6 point in the same direction as those obtained via direct minimisation, the 2SLS and LIML approximations give much larger test values for s(u) and generally lower signi®cance Table 3.6 LIML common features test for GDP growth, 1980±89 Germany UK Austria Spain France Italy

^ s(u)  ^ s(u)  ^ s(u)  ^ s(u)  ^ s(u) 

1.90 ±0.23 0.99 ±0.19 2.17 0.50 0.027 ±2.87 0.53 ±3.33

Note: 5 per cent critical value = 3.84.

UK ± ± 3.38 ±0.46 1.86 ±0.82 1.18 ±0.33 0.80 ±1.76

Austria ± ± ± ± 4.18 ±0.56 0.61 ±0.07 1.41 ±66.57

Spain

France

± ± ± ± ± ± 0.57 ±0.42 0.18 ±0.50

± ± ± ± ± ± ± ± 0.42 ±0.82

Hong Bai, Stephen Hall and David Shepherd 53 Table 3.7

Unit root tests for nominal and real exchange rates

France (1972±94) Italy (1980±93) Spain (1972±93) UK (1972±94) Austria (1972±94)

Nominal exchange rate DF ADF

Real exchange rate DF ADF

±1.88 ±1.82 ±1.61 ±1.84 ±2.23

±2.63 ±2.35 ±1.79 ±1.63 ±

±1.89 ±2.21 ±1.98 ±2.17 ±2.46

±2.89 ±2.43 ±1.96 ±2.06 ±

Note: Critical value for ADF(1) = ±3.40

levels for . It also appears that the different normalisations do have a signi®cant impact on the 2SLS estimates. The implication is that the results from the LIML and 2SLS approximations are far less robust than those obtained via direct minimisation.

3.4 The common features test for exchange rates and interest rates In this section we consider whether common cyclical features in exchange rates and interest rates can be identi®ed for the larger EU countries. Starting with exchange rates, we examine the behaviour of the real and nominal exchange rate against the US dollar for each country. Following the same procedure as before, Table 3.7 shows the stationarity tests for each country with respect to Germany. The relevant ADF statistics suggest that nominal and real exchanges exhibit no cointegrating relationship with respect to Germany. The next step is to examine whether VAR(1) features are present in the bivariate models for nominal and real exchange rates. The relevant LM test statistics are reported in Tables 3.8 and 3.9. Although there are some marginal cases, particularly with respect to the Table 3.8 LM tests for bivariate VAR(1) features in nominal exchange rates

Germany France Italy Spain UK Austria

Germany

France

Italy

Spain

UK

Austria

± 7.92 7.23 7.22 6.23 6.96

14.82 ± 14.82 13.85 14.88 13.50

10.58 10.78 ± 9.74 10.69 8.68

11.05 11.66 11.23 ± 11.73 11.20

4.29 4.68 4.99 3.71 ± 3.36

7.03 7.31 6.10 7.14 6.11 ±

Note: 5 per cent critical value = 5.99 (10 per cent = 4.61). Dependent variables are in the ®rst column.

54 Testing for Common Features in the EU Table 3.9 LM tests for bivariate VAR(1) features in real exchange rates

Germany France Italy Spain UK Austria

Germany

France

Italy

Spain

UK

Austria

± 8.73 8.01 7.83 5.65 6.29

6.72 ± 12.51 12.78 4.71 6.61

12.04 6.71 ± 7.28 8.82 5.25

12.60 12.69 12.20 ± 13.18 6.86

2.61 3.34 2.43 2.39 ± 5.38

4.76 10.44 4.57 9.13 2.18 ±

Note: 5 per cent critical value = 5.99 (10 per cent = 4.61). Dependent variables are in the ®rst column.

UK, the LM test statistics reported in Tables 3.8 and 3.9 suggest that serial correlation features are generally present in both nominal and real exchange rates. To see whether these features are common, we apply the common features test to the logarithmic ®rst differences of the nominal and real exchange rate series for the period 1972 (®rst quarter) to 1993 (third quarter). The results obtained via the direct minimisation approach are shown in Tables 3.10 and 3.11. In general, there appear to be common cyclical features in the nominal exchange rate relationships, with the exception of Italy in relation to France. For real exchange rates, ®ve of the test statistics reject the presence of a common cycle, all involving a pair-wise comparison with Italy. These results suggest that both the nominal and real exchange rate in Italy appear to be subject to cyclical patterns that are signi®cantly different from those of the other main European countries.

Table 3.10 The common features test for nominal exchange rates Germany UK Austria Spain France Italy

s(u Ã)  s(u Ã)  s(u Ã)  s(u Ã)  s(u Ã) 

1.80 ±0.63 0.23 ±1.03 1.88 ±1.29 0.16 ±1.25 3.42 ±0.79

Note: 5 per cent critical value = 3.84

UK ± ± 1.57 ±0.69 2.21 ±0.42 0.69 ±0.61 1.33 ±0.62

Austria

Spain

France

± ± ± ± 1.62 ±1.31 0.46 ±1.27 0.12 ±0.39

± ± ± ± ± ± 2.53 ±0.93 2.05 ±1.13

± ± ± ± ± ± ± ± 4.62 ±1.40

Hong Bai, Stephen Hall and David Shepherd 55 Table 3.11 The common features test for real exchange rates Germany UK Austria Spain France Italy

s(u Ã)  s(u Ã)  s(u Ã)  s(u Ã)  s(u Ã) 

2.16 ±0.39 0.00 ±0.79 0.19 ±0.11 0.28 ±1.18 4.20 -3.28

UK ± ± 0.87 ±0.59 0.95 ±1.64 0.22 ±0.67 0.62 ±1.53

Austria

Spain

France

± ± ± ± 2.26 ±1.01 0.33 ±0.95 5.55 ±0.68

± ± ± ± ± ± 1.07 ±0.91 5.54 ±0.21

± ± ± ± ± ± ± ± 5.39 ±0.77

Note: 5 per cent critical value = 3.84

Turning next to the behaviour of interest rates, we ®rst repeat the test procedures for the case of long-term nominal interest rates. The LM test results for the bivariate VAR(1) relationships in Table 3.12 suggest that serial correlation features are present in most of the series, but not all. The results in Table 3.13 indicate that in those cases where the feature is present, it appears to be a common feature. Finally, we complete our investigation of common features by considering the behaviour of long-term real interest rates. Table 3.14 shows the LM test statistics for the preliminary VAR(1) model. The results from the VAR(1) estimates suggest that real interest rates in most cases do not exhibit the required serial correlation features and hence, the common features test is not generally appropriate. The LM test statistics indicate that a serial correlation feature is present on both sides of the relationship only for the UK and Germany. Although we have not reported the results, the common features test for the UK and Germany indicates a common procyclical relationship between the long-term real interest rates for those two countries. Table 3.12 LM tests for bivariate VAR(1) features in nominal interest rates

Germany UK Spain France Italy

Germany

UK

Spain

France

Italy

± 3.01 1.95 2.30 1.99

11.64 ± 2.39 5.61 7.08

14.34 9.16 ± 13.05 11.70

26.35 21.35 8.55 ± 21.17

11.64 10.57 3.92 16.15 ±

Note: 5 per cent critical value = 5.99. Dependent variables are in the ®rst column.

56 Testing for Common Features in the EU Table 3.13 The common features test for nominal interest rates

UK

s(u Ã)  s(u Ã)  ^ s(u)  ^ s(u) 

Spain France Italy

Germany

UK

Spain

France

0.00 ±0.37 0.00 ±0.96 0.28 ±0.56 0.54 ±1.99

± ± 0.00 ±0.46 0.21 ±0.57 0.00 ±0.79

± ± ± ± 0.97 ±0.78 0.33 ±0.71

± ± ± ± ± ± 0.38 ±0.95

Note: 5 per cent critical value = 3.84

Table 3.14 LM tests for bivariate VAR(1) features in real interest rates Germany Germany UK Spain France Italy

± 8.46 1.30 1.62 5.77

UK

Spain

France

Italy

12.11 ± 11.04 17.22 20.24

2.14 0.27 ± 0.12 0.058

3.91 3.06 21.42 ± 2.21

1.86 1.12 23.11 2.21 ±

Note: 5 per cent critical value = 5.99. Dependent variables are in the ®rst column.

3.5

Summary and conclusions

The extent to which the EU is likely to face con¯icts in the design and implementation of monetary policy after the adoption of the single currency depends in part on the degree of economic convergence achieved by the member states. In this chapter we have used the common features test to examine the degree of short-run (cyclical) convergence exhibited by GDP growth, exchange rates and long-term interest rates. Although the precise nature of this cyclical convergence needs further investigation, our preliminary results suggest that common cyclical features (represented by common serial correlation features) can be identi®ed in most of the series we have examined with the important exception of long-term real interest rates. We have also considered the ef®ciency of the different methods available for constructing the common features test. Our results suggest that the approximations based on 2SLS and LIML can be misleading and that they are far less robust than those obtained via direct minimisation.

Hong Bai, Stephen Hall and David Shepherd 57

References Backus, D. K., P. J. Kehoe and F. E. Kydland (1995) `International Business Cycles: Theory and Evidence', in T. F. Cooley (ed.), Frontiers of Business Cycle Research (Princeton NJ: Princeton University Press). Eichengreen, B. (1993) `European Monetary Uni®cation', Journal of Economic Literature, vol. X X X I , pp. 1321±57. Engle, G. F and S. Kozicki (1990) `Testing for Common Features', Discussion Paper no. 90±23 (San Diego, CA: Department of Economics, University of California, June). Goodhart, C. A. E. (1995) `The Political Economy of Monetary Union', in P. B. Kenen (ed.), Understanding Interdependence: the Macroeconomics of the Open Economy, (Princeton NJ: Princeton University Press). Hall, S. G., D. Robertson and M. Wickens, (1993) `How to Measure Convergence with an Application to the EC Economies', Working Paper, no. 19±93 (London Business School, Centre for Economic Forecasting). King, R. G. and S. T. Rebelo (1993) `Low Frequency Filtering and Real Business Cycles', Journal of Economic Dynamics and Control, vol. 17, pp. 208±31. Lippi, M. and L. Reichlin (1994) `Common and Uncommon Trends and Cycles', European Economic Review, vol. 38, pp. 624±35. Shepherd, D. (1990) `Economic Issues in European Monetary Union', European Business Journal, vol. 2, pp. 8±16. Vahid, F. and G. F. Engle (1993a) `Common Trends, Common Cycles', Journal of Applied Econometrics, vol. 8, pp. 341±60. Vahid, F. and G. F. Engle (1993b) `Non-Synchronous Common Cycles', Discussion Paper no. 93±55 (San Diego, CA: Department of Economics, University of California, December).

4

The Demand for Money in the European Union: the Role of Germany's Interest Rate Yonghao Pu and George Zis

4.1

Introduction

Artis et al. (1993) Kremers and Lane (1990) and Monticelli and StrausKahn (1993) have investigated the existence of a stable EU demand for money function, while Cassard et al. (1994) have estimated the demand for money functions involving aggregation only across the core countries of the Exchange Rate Mechanism (ERM). All four studies' empirical ®ndings suggest that a stable function is identi®able either at the EMS or the core ERM level. However, aspects of these studies raise questions about the robustness of the statistical evidence presented and about the consistency of their theory. This chapter discusses some of the issues relating to the estimation of an EU money demand function and presents new empirical evidence to indicate the existence of stable EU demand functions for both M1 and M2. How and why does this study differ from the other investigations of the EU money demand function? Section 4.2 addresses this question. Section 4.3 brie¯y describes the data used, and the statistical ®ndings are presented and discussed in Section 4.4. Finally, our conclusions and their policy implications are considered in Section 4.5.

4.2 EU money demand: rationale, speci®cation issues and aggregation problems The study of the demand for money is motivated by an interest in ascertaining the potential ef®ciency of monetary policy to control the rate of in¯ation. Put simply, the necessary and suf®cient condition for control of the money supply growth rate to result in control of the in¯ation rate is that the demand for money is systematically and predictably determined

58

Yonghao Pu and George Zis 59

by a small number of economic variables. But the simplicity of this basic proposition conceals highly controversial issues that need to be addressed if viable and potentially effective policy prescriptions are to be advanced. A set of issues involves the speci®cation of the money demand function, the potential stability of which is to serve as the basis for policy recommendations. Artis (1994, p. 215) distinguishes `ingenious speci®cations and data mining from judicious application of economic theory'. An implication of this distinction is that highly complex speci®cations, successful though they are in yielding statistically well-behaved money demand functions, are of limited usefulness. For policy prescription purposes, the simpler the identi®ed stable money demand function, the greater its relevance. Dif®culties in specifying and, especially, in estimating the demand for money function for a `small' open economy have partly motivated an interest in investigating the stability of the demand for money at the EMS level. It has been suggested that such dif®culties may be responsible for the relatively poor performance of single-country money demand equations in recent years. Misspeci®ed functions and/or limitations in econometric estimation techniques may yield theoretically implausible coef®cient estimates and/or misleadingly indicate instability. Kremers and Lane (1992) have presented a formal analysis of the potential misspeci®cation implications for conventionally de®ned national money demand functions when currency substitution and international portfolio diversi®cation are ignored. Their analysis provides an explanation of the difference between EMS and single country money demand studies regarding the speed at which money holdings adjust to their equilibrium level. The latter have yielded estimates indicating implausibly slow speeds of adjustment. Kremers and Lane demonstrate how this may be the outcome of misspeci®cation, which need not be present or as severe if aggregation across countries is implemented. This, then, is an argument for estimating an EMS money demand function. But if member countries' money demand functions differ, then the estimation of an EMS function will involve an aggregation bias. Thus, Kremers and Lane identify a tradeoff between the speci®cation bias that is likely to be present in estimates of single-country money demand functions and the aggregation bias involved in estimates of the demand for money function across a group of countries. They suggest that, because of currency substitution and portfolio diversi®cation, aggregation at the EMS level is appropriate, but perhaps would not be so for a broader group of countries. This judgement rests on a particular interpretation of the empirical studies of the EMS money demand function. Kremers and Lane (1990, 1992), in line with

60 The Demand for Money in the EU

Artis et al. (1993), Cassard et al. (1994) and Monticelli and Straus-Kahn (1993) maintain that the identi®cation of a stable demand for money function at the EMS level is evidence of the signi®cance of currency substitution. However, only Artis et al. (1993) test directly for currency substitution and present some (inconclusive) evidence in favour of the hypothesis. Similarly, while they detect the presence of aggregation bias, this is contradicted by the ®ndings of Cassard et al. (1994). Be that as it may, the statistical evidence on currency substitution and aggregation bias is not suf®ciently robust as to provide empirical foundations for choosing or rejecting aggregation at the EMS or the core ERM level. The rationale for investigating the demand for money at the EMS level not only rests on speci®cation problems, it also re¯ects interest in policy issues associated with the provisions of the Maastricht Treaty. The existence or otherwise of a stable EMS money demand function is informationally signi®cant for the European Monetary Institute's attempts to promote the increasing coordination of member countries' monetary policies and for the design of the EU's monetary policy in Stage Three. The trade-off between speci®cation bias and aggregation bias and interest in policy issues may therefore be combined to identify the appropriate group of countries across which the demand for money is to be investigated. However, the level of aggregation thus determined may be of dubious operational validity. Johnson (1972a, 1972b), when asserting the existence of a stable world money demand function during the era of the Bretton Woods international monetary system, emphasised the signi®cance of ®xed exchange rates in effectively transforming individual countries into regions of a single monetary domain. That is, Johnson's thesis implies an alternative criterion for determining the appropriate level of aggregation, which is especially relevant when investigating the EMS demand for money. The empirical signi®cance of currency substitution and portfolio diversi®cation is necessary but not suf®cient for establishing the legitimacy of aggregation across the system's member countries. Two features of the EMS impart importance to the Johnson criterion. First, the EMS has not functioned as a system of ®xed exchange rates. Second, not all EU member countries took part in the ERM during the 1980s. Thus, the construction of aggregate EMS variables is highly problematic. For example, adding up member countries' money supplies can yield as many estimates of the aggregate EMS money supply as the number of member countries. The Johnson criterion therefore suggests that estimating an EMS demand for money function for the period since the establishment of

Yonghao Pu and George Zis 61

the system and aggregation involving countries not operating the ERM are inappropriate. In the case of Artis et al. (1993), Kremers and Lane (1990) and Monticelli and Straus-Kahn (1993) the sample period begins in 1979; that is, it covers all 11 intra-EMS exchange rate realignments. Furthermore, somewhat surprisingly Monticelli and Straus-Kahn (1993) included the UK and Spain in the construction of the aggregate variables. In contrast, Cassard et al. (1994), in recognition of the importance of exchange rate ®xity for the applicability of the aggregate approach, con®ne their aggregation to the core ERM countries, de®ned as France, Germany, the Benelux countries and Denmark, as they had consistently operated in the narrow band of the ERM and had not realigned their currencies against the Deutschmark since at least January 1987. These criteria explain the exclusion of Italy but not of Ireland. Furthermore, their sample period begins in the fourth quarter of 1982; that is, it covers the major 1983 realignment. In order to minimise the distortionary effects of exchange rate changes our sample period extends from the second quarter of 1983 to the fourth quarter of 1991, during which time only four of the 11 realignments occurred. In constructing the aggregate variables we excluded the UK, Spain, Portugal and Greece. The rationale for estimating an aggregate ERM money demand function raises the question of how to proxy the opportunity cost of holding money balances. If currency substitution provides the principal motivation, then the German interest rate should feature as the opportunity cost variable in the speci®cation of the ERM money demand function. Currency substitution predicts that economic agents, in anticipation of exchange rate changes, will alter the composition of their portfolios of currencies by substituting the currency expected to appreciate for that which is expected to depreciate but leaving their demand for money balances unchanged. It follows, then, that if, for example, the French rate of interest rises because of an expectation that the franc will be devalued, the demand for francs will fall but the ERM demand for money will not be affected. Since at no time during the 1980s was there an expectation that the Deutschmark would be devalued against any of the other ERM member countries' currencies, the German rate of interest is the appropriate opportunity cost variable for the investigation of the ERM demand for money. Artis et al. (1993), Kremers and Lane (1990) and Monticelli and Straus-Kahn (1993) all use as their opportunity cost variable a weighted average of national interest rates. The weights used are the 1989 shares of the national currencies in the ECU. No justi®cation for this approach is offered. But perhaps more signi®cantly, the representation of the opportunity cost variable by the weighted average of national interest

62 The Demand for Money in the EU

rates is not consistent with the strong emphasis they place on currency substitution in motivating their investigation of the demand for money at the EMS level. In contrast to these studies, we employ the German rate of interest as the appropriate opportunity cost variable in estimating the ERM money demand function. A further difference involves the speci®cation of the ERM money demand function to be estimated. Artis et al. (1993) and Kremers and Lane (1990) include the dollar±ecu exchange rate as an explanatory variable while Monticelli and Straus-Kahn (1993) present estimates of EMS money demand functions with and without the four-quarter rate of change in the dollar±ecu exchange rate. Their preferred speci®cation includes the dollar±ecu exchange rate change. The objective is to capture the potential in¯uence of expected movements in the dollar±ecu exchange rate on the demand for money. The implied expectations formation scheme has economic agents basing their expectations of future changes in the dollar± ecu exchange rate on its current value. However, Lane and Poloz (1992) point out that this is inconsistent with rational expectations as the exchange rate is a non-stationary variable and therefore contains no information about its future movements. They also suggest that if shocks affect both the exchange rate and the demand for money, then the resultant relationship between the two need not re¯ect the in¯uence of expected exchange rate changes. Finally, they draw attention to the implications of monetary authorities using exchange rate changes as a guide to monetary policy. Thus they conclude that it is an `unresolved question . . . whether the presence of exchange rates in money-demand equations really does re¯ect currency substitution' (ibid., p. 28). Indeed Artis (1994, p. 222) describes the inclusion of the dollar±ecu exchange rate as `analytically embarrassing'. We have therefore excluded the dollar±ecu exchange rate and its rate of change when specifying the ERM demand for money function. Finally, the estimation strategy employed to investigate the aggregate ERM demand for money is different from those used in the other empirical studies. Artis et al. (1993), Cassard et al. (1994), Kremers and Lane (1990) and Monticelli and Straus-Kahn (1993) all rely on the Engle± Granger (1987) two-step methodology. Monticelli and Straus-Kahn (1993) use the Johansen (1988) and Johansen and Juselius (1990) multivariate cointegration technique to con®rm the empirical ®ndings derived from the application of the Engle±Granger procedure. Cassard et al. (1994) also present estimates based on the former estimation technique. Reliance on the Engle±Granger methodology is not appropriate for the estimation of money demand functions. The use of ordinary

Yonghao Pu and George Zis 63

least squares to estimate a cointegration relationship for an N dimensional vector does not establish whether the identi®ed cointegrating vector is unique or simply a linear combination of all the distinct cointegrating vectors that exist within the system. Cassard et al. (1994) and Monticelli and Straus-Kahn (1993) are not particularly successful in strengthening their conclusions by complementing their empirical ®ndings based on the Engle±Granger methodology with results derived from the application of the Johansen procedure. Both studies identify more than one cointegrating vector, and therefore their empirical estimates cannot be unambiguously interpreted. As Wickens (1993) has observed, only when there is a single cointegrating vector is the vector equivalent to the coef®cients of a cointegrating regression, and thus yields an estimator of the long-term relationship coef®cients of the I(1) variables in the system. In brief then, this study differs from other investigations of the ERM money demand function in terms of speci®cation, level of aggregation, the sample period covered, the proxying of the opportunity cost variable of money holding and the estimation methodology employed.

4.3

The data

Demand functions for both M1 and M2 were estimated. As we relied for our data on the International Financial Statistics (IFS) our measure of broad money is narrower than those of Cassard et al. (1994) and Monticelli and Straus-Kahn (1993), who utilised national sources. For M1 IFS line 34b and for M2 lines 34 and 35 were used. For the income variable, GDP or GNP data, IFS line 99b, c line, were employed. The German nominal interest rate is given by IFS line 60b (money market rate) and the aggregate ERM price index was proxied by the weighted average of national consumer price indices. For the construction of the aggregate ERM money supply and income variables, we converted national currency values into ecus at the central ecu exchange rates established following the January 1987 intra-EMS exchange rate realignment. In the construction of the ERM price index, national price indices were weighted by countries' shares in the constructed ERM income variable.

4.4

The empirical results

We estimated the simplest theory-consistent money demand function, with income and the interest rate as the only independent variables. Inevitably the speci®cation adopted involved the reduction of the data

64 The Demand for Money in the EU

generation process through the omission of variables relevant to the system under consideration. Whether or not the implied degree of marginalisation is valid is an empirical issue. If the conditions for weak exogeneity are satis®ed, then no loss of information is involved in analysing the marginal density only. In our case, the `variables of interest' are real money balances, real income and the German money market interest rate as we hypothesised that, assuming long-term price homogeneity, the ERM demand for real money balances depends on real income and Germany's money market interest rate. In what follows, m is the log of real money balances, y is the log of real income and srg is Germany's money market interest rate. There exist, therefore, two systems in the three stochastic variables (m, y and srg), one for M1 and one for M2, with a constant and a trend. Before we proceed with the empirical analysis, the treatment of the deterministic variables needs clari®cation. Following Germany's monetary uni®cation, the country's money supply growth rate behaved erratically for a brief time. Thus, Germany's M1 grew at 22 per cent during the ®rst quarter of 1991 and M2 increased by 14 per cent in the second quarter of 1990. These sharp rises in Germany's money supply resulted in the ERM-wide money aggregates exhibiting ephemeral volatility. Thus, we employed a dummy variable, D91Q1, for M1 and one for M2, D90Q2. However, given that Germany's monetary uni®cation only had a short-term impact on the evolution of the ERM money supply growth, these dummy variables were included only in the shortterm dynamics and were excluded from the long-term cointegration space. Second, as M2 is seasonally unadjusted, three seasonal dummies were included in the M2 system. Again, these were included only in the short-term dynamics. Third, following Hendry and Doornik (1994), the intercept appeared only in the short-term dynamics and the trend only in the long-term cointegration space. Implementation of the Johansen methodology required identi®cation of the lag length for the VAR and determination of the orders of integration of the series entering the VAR. For the latter, we employed the Augmented Dickey±Fuller (ADF) and Phillips±Perron (PP) tests. These tests for a unit root, available on request, indicated that all the stochastic variables are I(1). Next, starting from a VAR with four lags on all the stochastic variables we established through simpli®cation tests the adequacy of two lags. This reduction was implemented by using likelihood ratio tests adjusted for degrees of freedom. The relevant statistics are F(18.51) = 0.65 [0.84] and F(18.42) = 1.44 [0.16]. As they are both insigni®cant, the simpli®cation for both M1 and M2 is accepted. The

Yonghao Pu and George Zis 65 Table 4.1

Residual correlations, goodness of ®t and evaluation

^ Far (4, 22) Farch (3, 20) Fhet (14, 11) c2nd (2) Fvar (36, 36) cvhet (84, 39) c2v nd (6)

^ Far (4, 19) Farch (3, 17) Fhet (18, 8) c2nd (2) Fvar (36, 27) cvhet (84, 23) c2v nd (6)

M1

y

srg

0.7% 0.62 [0.65] 0.58 [0.63] 0.41 [0.94] 2.38 [0.30]

0.7% 0.36 [0.84] 0.04 [0.99] 0.43 [0.93] 3.07 [0.21]

0.3% 1.28 [0.31] 0.02 [0.99] 0.36 [0.96] 0.86 [0.65]

M2

y

srg

0.5% 1.73 [0.18] 0.72 [0.56] 0.15 [0.99] 3.07 [0.22]

0.7% 0.3% 0.87 [0.50] 0.08 [0.97] 0.55 [0.84] 3.11 [0.21]

1.22 0.05 0.42 1.60

[0.33] [0.98] [0.93] [0.45]

VAR

0.72 [0.83] 0.28 [1.00] 4.68 [0.58] VAR

0.98 [0.53] 0.24 [1.00] 7.28 [0.30]

Note: The probabilities of each test statistic are reported in square brackets

summary statistics for these two simpli®ed systems are presented in Table 4.1, where ^ denotes the residual standard deviation. F denotes F-tests for no serial correlation (Far , against fourth-order autoregression); no autoregressive conditional heteroscedasticity (Farch , against fourth order and no heteroscedasticity (Fhet ); the chi-square test for normality is denoted by X2nd ; analogous vector tests are denoted by a superscript v. As all these tests are satis®ed, the implication is that the two simpli®ed VAR systems are an adequate characterisation of the data. As a preliminary step towards the cointegration analysis, the break-point Chow test was employed as an informal test of parameter constancy. Figure 4.1 indicates that for none of the individual equations do the test values exceed the 5 per cent signi®cance level, a ®nding that is consistent with parameter constancy. The next step in our empirical analysis was to test for cointegration in the three equation systems. Table 4.2 presents eigenvalues; (1), the associated maximum eigenvalue, T log (1±1), the trace statistics, TS log (1±1), and the estimated cointegrating vectors. For M2, the maximum eigenvalue and trace statistics indicate that there is only one cointegrating vector. For M1, again only one cointegrating vector is identi®able but only

66 The Demand for Money in the EU M2 (5% crit.)

y (5% crit.)

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

1988

1989

1990

1991

1992

srg (5% crit.)

1.0

0

0.8

0.6

0.6

0.4

0.4

0.2

0.2 1988

1989

1990

1991

1992

0

1989

1990

1991

1992

1991

1992

Chow (5% crit.)

1.0

0.8

0

1988

1988

1989

1990

Notes: 12 and 16 observations are reserved to meet requirement of running Chow N-step constancy test for M1 and M2 (including seasonal dummies) respectively.

Figure 4.1

M1 and M2 sequence of break-point Chow test, 1983(2)±91(4)

marginally so. However Cheung and Lai (1993) argue that the trace test is the more powerful. Table 4.2, then, implies that the problems associated with the identi®cation of more than one cointegrating vector are not present, and we may therefore proceed to test for weak exogeneity. The unrestricted cointegration vectors in Table 4.2 were tested for lying in the cointegration space when testing for y and srg being long-term weakly exogenous for the money demand parameters. Weak exogeneity requires that the ®rst column of a has the form (*,0,0) when b is identi®ed. First, the hypothesis that H1 :a1 = (*,0,0) was tested for M1. The LR(2) test statistic is 4.33 [0.11], which suggests that y and srg are long-term weakly exogenous for the money demand parameters. Next, we investigated whether restricting the income elasticity for M1 to 0.5, in line with the Baumol (1952) and Tobin (1956) models of the transactions demand for money, is acceptable. This involved testing the hypothesis that H2 :b1 = (1,-0.5,*). In jointly testing hypotheses H1 and H2 the LR(3) test statistic is

Yonghao Pu and George Zis 67 M1 (5% crit.)

y (5% crit.)

1.0

1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

1.0

1990

0

1995

srg (5% crit.)

0.8

0.6

0.6

0.4

0.4

0.2

0.2 1990

1995

1995

Chow (5% crit.)

1.0

0.8

0

1990

0

1990

1995

Figure 4.1 M1 and M2 sequence of break-point Chow test, 1983(2)±91(4) (continued)

4.54 [0.21]. Thus, the joint hypothesis is accepted. For M2, when testing the H1 hypothesis, the LR(2) test statistic is 2.52 [0.28]. That is, y and srg are long-term weakly exogenous for the money demand parameters. Table 4.3 presents the estimated restricted cointegrating vectors for M1 and M2.1 These vectors may be interpreted as long-term money demand equations.2 The implied coef®cients are theory consistent and, perhaps, more intuitively plausible than those obtained in the other studies of the EMS demand for money function. This is especially the case with the estimated income elasticities. They are signi®cantly lower than those presented in the other studies, particularly for M2, for which Artis et al. (1993) obtained an estimate of 1.99. The next step is to map all the variables to I(0) space by means of differencing and cointegration transformations so that standard inference procedures are applicable. The ECM identities, retaining the differences of the three original variables, are given by: Cm1ot Dm1t

0:5Dyt ‡ 0:97 Dsrgt

0:0072

Cm1t

1

68

Table 4.2 Cointegration analysis of M1 and M2 Eigenvalues and test statistics for (M2)

Eigenvalues and test statistics for (M1) H0 r £2 r £1 r=0

Eigenvalue (1i )

T log (1 ± 1i )

TS log (1 ± 1i )

H0

Eigenvalue (1i )

0.1106 0.3559 0.4329

4.10 15.40 19.85

4.10 19.50 39.35*

r £2 r £1 r=0

0.1311 0.3818 0.6096

1.000 ±0.065 ±15.66

4.92 21.75 54.67**

y

srg

Trend

M2

y

srg

Trend

±0.823 1.000 8.389

1.064 ±0.004 1.000

±0.0047 0.0074 0.0807

1.000 ±0.289 ±1.978

±0.753 1.000 1.351

0.777 ±0.130 1.000

±0.0042 ±0.0051 0.0053

Standardised a eigenvectors

Standardised a eigenvectors M1 y srg

4.92 16.83 32.92**

TS log (1 ± 1i )

Standardised b eigenvectors

Standardised b eigenvectors M1

T log (1 ± 1i )

±0.503 0.087 ±0.154

±0.279 ±0.358 0.281

0.003 ±0.007 ±0.003

M2 y srg

±0.451 0.183 ±0.164

±0.326 ±0.025 ±0.156

*/**/***Test statistics above the 10 per cent/5 per cent/1 per cent critical value, respectively (see Osterwald-Lenum, 1992).

0.003 0.002 ±0.001

Yonghao Pu and George Zis 69 Table 4.3

* a*

* a*

Restricted cointegration vector * and loadings a* M1

y

srg

Trend

1.0000 ±0.5597

±0.5000 0.0000

0.9656 0.0000

±0.0072

M2

y

srg

Trend

1.0000 ±0.5083

±0.5961 0.0000

0.7562 0.0000

±0.0054

Cm2ot Dm2t

0:6 Dyt ‡ 0:76 Dsrgt

0:0054

Cm2t

1

The resultant I(0) VAR systems involve the retention of the stochastic regressors Dmt 1 , Dyt 1 and Dsrgt 1 , while the deterministic part includes a constant, seasonal dummies and the dummy variables already discussed. The implied equations for each I(0) VAR were estimated by full information maximum likelihood (FIML). The results of the estimation are not reported but are available on request. However, it is worth noting that the correlations between the actual and ®tted values for the three stochastic variables are 0.74, 0.30 and 0.54 for the M1 VAR, while the corresponding values for the M2 VAR are 0.98, 0.41 and 0.50. The coef®cients for Cm1t and Cm2t in the three equations in each VAR are reported in Table 4.4. Except for the money equations, these coef®cients are insigni®cant. Table 4.4 also shows the results of testing for overidentifying restrictions, which support the conclusion that the restricted models parsimoniously encompass the VAR. It follows, then, that the weak exogeneity conclusion is con®rmed, enabling the inclusion of contemporaneous observations of the weakly exogenous variables y and srg in the estimation of a single conditional dynamic demand for money equations, instead of the restricted VAR, in which y and srg are determined within the system. The preferred speci®cations of the error correction formulation of the long-term demand functions for M1 and M2 are presented in Table 4.5. For both equations all the diagnostic tests are satis®ed. The estimated coef®cients are statistically signi®cant and theory consistent. The coef®cients on the error correction terms Cm1t 1 and Cm2t 1 imply signi®cantly faster and more plausible rates of adjustment than the ®ndings of single-country money demand equations. In brief, the evidence reported in Table 4.5 suggests that the estimated equations are reasonably data coherent representations.

70

Table 4.4 Signi®cance of ECMs in the dynamic system of M1 and M2 Equation Dm1t Dyt Dsrgt

M1 Cm1t coef®cient

M2 T-test [prob.]

Equation

Cm2t coef®cient

T-test [prob.]

±0.5763 ±0.0979 ±0.1173

±4.235 [0.0002] ±0.688 [0.4972] ±1.531 [0.1366]

Dm2t Dyt Dsrgt

±0.4764 ±0.0669 ±0.0890

±4.893 [0.0000] ±0.429 [0.6717] ±1.004 [0.3251]

Encompassing test: LR(2)

1.6453 [0.4393]

Encompassing test: LR(2) 3.3378 [0.1885]

Table 4.5

Dynamic equations of European money (M1 and M2) demand function Dm2t

Dm1t Variable

Coeff.

s.e.

T-test

Variable

Coeff.

s.e.

T-test

Cons. Dsrgt Cm1t 1 D91Q1

0.0079 ±0.4862 ±0.4780 0.0268

0.001 0.254 0.121 0.007

6.353 1.915 ±3.939 3.903

Cons. Dm2t 1 Dsrgt 1 Cm2t 1 Q1 Q2 Q3 D90Q2

0.3286 0.2627 ±0.4733 ±0.4325 ±0.0430 ±0.0271 ±0.0290 0.0336

0.061 0.097 0.191 0.089 0.004 0.003 0.002 0.005

5.411 2.722 ±2.477 ±4.832 ±11.750 ±9.512 ±13.323 6.898

R2 = 0.73 e(%) = 0.67 AR(4, 27)1 4 = 0.97 [2.73] HET(4, 26) = 0.25 [2.74] ARCH(4, 23) = 0.35 [2.80] NORM(2) = 2.31 [5.99]

R2 = 0.96 e(%) = 0.44 AR(4,22)1 4 = 1.45 [2.82] HET(9, 16) = 0.88 [2.54] ARCH(4,18) = 1.67 [2.93] NORM(2) = 5.79 [5.99]

Notes: AR(4, j)t …i† is an LM test for fourth order of serial correlation; HET(i, j) is an F test for heteroscedasticity; fourth order autoregressive conditional heteroscedasticity test is denoted by ARCH(4, j); non-normality test is denoted by NORM(2).

71

72 The Demand for Money in the EU

Key Dm2 Fitted

0.042 0.035 0.028 0.021 0.014 0.007 0 –0.007 –0.014 –0.021 –0.028

0.040 0.035 0.030 0.025 0.020 0.015 0.010 0.005 0 –0.005 –0.010 –0.015 Figure 4.2 models

1984

1985

1986

1987

1988

1989

1990

1991

1992

Key Dm1 Fitted

1984

1985

1986

1987

1988

1989

1990

1991

1992

Dm1t , Dm2t ®tted and actual values from the conditional dynamic

To check the constancy of the models over the sample period, Figure 4.2 shows plots of the actual outcomes and ®tted values for Dm1 and Dm2. The latter track the former reasonably well, particularly in the case of Dm2. Figure 4.3, based on the recursive FIML estimates, highlights the models' stability in terms of residual sums of squares, one-step residuals with 2SE, the one-step Chow forecasting failure test and the N-step Chow stability test. For both equations the one-step errors lie within the 95 per cent con®dence bands with constant errors, while for each equation the values of the break-point Chow F-tests never exceed the 1 per cent signi®cance level. In summary, then, the models appear to be reasonably constant.

Yonghao Pu and George Zis 73

Dm1

Dm1

Res1Step ± 2*S.E.

RSS 0.0016 0.0014

0.014

0.0012

0.007

0.0010

0

0.0008

0.007

0.0006

0.014

0.0004

1990

1995

0.021

Chow (1% crit.)

0.8

0.6

0.6

0.4

0.4

0.2

0.2 1990 Figure 4.3

4.4

1.0

0.8

0

1995

Dm1

Dm1 1.0

1990

1995

0

Chow (1% crit.)

1990

1995

System stability and constancy tests

Conclusions

The EU demand functions for M1 and M2 were investigated. These differ sharply from those in other studies in terms of the level of aggregation implemented, the speci®cation adopted, the sample period covered, the proxying of the opportunity cost variable of money holding and the estimation methodology employed. The statistical evidence presented is consistent with stable and well-behaved M1 and M2 demand functions. Suggestive though this evidence is, it would be premature to assert that the EU money demand function is, in fact, stable and proceed with con®dence to prescribe policy. The reasons why the empirical study of the EMS money demand yields results that are more theory consistent and

74 The Demand for Money in the EU

intuitively more plausible than those obtained in single-country money demand investigations are yet to be unambiguously established. The theory of currency substitution has motivated aggregation across EMS member countries. The statistical performance of aggregate EMS money demand functions has, then, been interpreted to indicate the empirical signi®cance of currency substitution and the validity of aggregating across countries. But the performance of aggregate money demand functions does not necessarily imply that only currency substitution is at work, or that this phenomenon is the principal reason why aggregate functions outperform single country money demand functions. For example, the behaviour of intra-EMS exchange rates during the 1980s is not consistent with currency substitution being empirically signi®cant. It is therefore necessary directly to test currency substitution. What is at issue is not only the legitimacy of aggregating across countries. If currency substitution were con®rmed as a signi®cant phenomenon, then there necessarily would follow implications for the appropriate level of aggregation to be implemented, the speci®cation of the aggregate money demand function to be estimated, the relevance of potential aggregation bias and the design of monetary policy at the member country and EU level during the transition to the establishment of the European currency union. It is beyond dispute that the dif®culties in interpreting the evidence for the existence of a stable EU money demand function for policy prescription purposes mean that one should be cautious about advancing policy proposals. Nonetheless, this evidence cannot be ignored. When combined with other monetary developments, certain policy judgements may be made. For example, the 1992±93 EMS `crisis' demonstrated that ®nancial integration of the EU has proceeded suf®ciently to erode the ability of the Bundesbank to pursue a unilaterally determined monetary policy aimed at achieving its preferred rate of in¯ation. It is arguable that Germany would be better placed to realise its anti-in¯ation objectives if it Europeanised its monetary policy. From a German perspective, the case for such a shift of focus is reinforced now that the emergence of a two-speed Europe seems inevitable. The potential costs for Germany would be, at worst, negligible. The ®ndings reported in this chapter suggest that a Europeanised German monetary policy would be ef®cient. Notes 1. The other elements of the restricted matrices a and b (not tabulated in Table 4.3) are zeros.

Yonghao Pu and George Zis 75 2. We also applied the same speci®cation and estimation approach to model money demand function for the seven nations individually. Using such a simple speci®cation, we were unable to identify a well-behaved VAR system for France, Italy and Belgium. We also failed to detect a cointegration relationship for the Netherlands and Ireland. Among the rest of the ®ve nations, only Germany and Belgium passed the weak exogeneity test, but the coef®cient on the income variable in the Belgium equation was wrongly signed. That is, only Germany is found to have an empirically robust and theory-consistent money demand function. The full results are available on request.

References Artis, M. J. (1994) `Stage Two: Feasible Transitions to EMU', in D. Cobham (ed.), European Monetary Upheavals (Manchester: Manchester University Press). Artis, M. J., R. C. Bladen-Hovell and W. Zhang (1993) `A European Money Demand Function', in P. R. Masson and M. P. Taylor (ed), Policy Issues in the Operation of Currency Unions (Cambridge: Cambridge University Press). Baumol, W. J. (1952) `The Transactions Demand for Cash: An Inventory Theoretic Approach', Quarterly Journal of Economics, vol. 66, pp. 545±56. Cassard, M., T. Lane and P. R. Masson (1994) `ERM Money Supplies and the Transition to EMU', IMF Working Paper, WP/94/1 (Washington, DC: IMF). Cheung, Y. W. and K. S. Lai (1993) `Finite-Sample Size of Johansen's Likelihood Ratio for Cointegration', Oxford Bulletin of Economics and Statistics, vol. 55, no. 3, pp. 313±28. Engle, R. and C. Granger (1987) `Cointegration and Error Correction: Representation, Estimation and Testing', Econometrica, vol. 55, pp. 251±76. Hendry, D. F. and J. A. Doornik (1994) `Modelling Linear Dynamic Econometric Systems', Scottish Journal of Political Economy, vol. 41, pp. 1±33. Johansen, S. (1988) `Statistical Analysis of Cointegration Vectors', Journal of Economic Dynamics and Control, vol. 12, pp. 231±54. Johansen, S. and S. Juselius (1990) `Maximum Likelihood Estimation and Inferences on Cointegration with Applications to the Demand for Money', Oxford Bulletin of Economics and Statistics, vol. 52, no. 2, pp. 169±210. Johnson, H. G. (1972a) `In¯ation: A Monetarist View', in Further Essays in Monetary Economics (London: Allen & Unwin). Johnson, H. G. (1972b) In¯ation and the Monetarist Controversy (Amsterdam: NorthHolland). Kremers. J. and T. D. Lane (1990) `Economic and Monetary Integration and the Aggregate Demand for Money in the EMS', IMF Staff Papers, vol. 37 (Washington, DC: IMF), pp. 777±805. Kremers, J. and T. D. Lane (1992) `The Implications of Cross-Border Monetary Aggregation', IMF Working Paper WP/92/71 (Washington, DC: IMF). Lane, T. D. and S. S. Poloz (1992) `Currency Substitution and Cross-Border Monetary Aggregation: Evidence from the G±7', IMF Working Paper WP/92/81 (Washington, DC: IMF). Monticelli, C. and M. Straus-Kahn (1993) `European Integration and the Demand for Broad Money', Manchester School, vol. L X I , pp. 345±66. Osterwald-Lenum, M. (1992) `A Note with Fractiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics: Four Cases', Oxford Bulletin of Economics and Statistics, vol. 54, no. 4, pp. 461±72.

76 The Demand for Money in the EU Tobin, J. (1956) `The Interest Elasticity of Transactions Demand for Cash', Review of Economics and Statistics, vol. 38, pp. 241±47. Wickens, M. (1993) `Interpreting Cointegrating Vectors and Common Stochastic Trends', DP no. 14±93 (London: Centre for Economic Forecasting, London Business School).

5

Credibility and In¯ation Expectations in the European Monetary System Nicholas Sarantis

5.1

Introduction

Although the original emphasis of the European Exchange Rate Mechanism (ERM) was on exchange rate stability, recent experience suggests that the focus shifted during the 1980s towards the achievement of lower and stable in¯ation rates in the European Community. The counterin¯ationary bene®ts of ERM membership are understood to arise primarily from a credibility effect on the formation of domestic in¯ation expectations. According to the credibility hypothesis (see, for example, Collins, 1988; Giavazzi and Giovannini, 1988; Giavazzi and Pagano, 1988; Melitz, 1988), the Deutschmark provides the anchor for the ERM system due the Bundesbank's strong anti-in¯ationary reputation.1 This reputation has two fundamental bases: Germany's experience with hyperin¯ation in the 1920s and immediately after the Second World War, which left a deeprooted fear among the German population; and the Bundesbank's constitution, which states that the defence of the Deutschmark's real value is the central bank's primary objective. Consequently, Germany has experienced low equilibrium in¯ation rates throughout the postwar period. Membership of the ERM obliges high-in¯ation countries to use monetary policy to keep their currency within the exchange rate band, and hence reduce domestic in¯ation to the German level. Although governments may still opt to devalue their currencies, the political cost of doing so acts as a restraint.2 Private agents are fully aware of this commitment and therefore revise their expectations downwards. The result is a reduction of domestic in¯ation at a lower cost in terms of output and employment losses. In effect, ERM membership enables high-

77

Per cent

78 Credibility and In¯ation

0

1951 1953 1955 1957 1959 1961 1963 1965 1967 1969 1971 1973 1975 1977

s Figure 5.1 Domestic in¯ation rates relative to Germany's in¯ation, pre-ERM

in¯ation countries to increase the credibility of their own counterin¯ationary policy announcements by borrowing the Bundesbank's reputation. Casual observations suggest that there has been some convergence of the in¯ation rates of the ERM member countries. Figures 5.1±5.4 show the differences between domestic in¯ation and German in¯ation rates for seven ERM countries during the pre-ERM period (1951±78) and the ERM period (1979±92) respectively.3 There is no clear pattern of in¯ation convergence during the pre-ERM period, though all in¯ation differentials were relatively stable during the 1950s and 1960s. That was the Bretton Woods period, which was characterised by low global in¯ation rates. The 1970s marked the beginning of high and volatile in¯ation rates. This continued into the early 1980s, when the ERM experienced frequent realignments. A consistent pattern of in¯ation convergence emerged during 1982±83, the beginning of the second and more stable phase of the ERM, and the European in¯ation rates gradually converged towards the low German rates by the late 1980s.4 Whether this in¯ation convergence was due to the credibility hypothesis is an open question. Egebo and Englander (1992) and Goodhart

Per cent

Nicholas Sarantis 79

1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978

Denmark

Ireland

Italy

UK

Per cent

Figure 5.2 Domestic in¯ation rates relative to Germany's in¯ation, pre-ERM

0

Denmark

Ireland

Italy

UK

Figure 5.3 Domestic in¯ation rates relative to Germany's in¯ation, ERM

Per cent

80 Credibility and In¯ation

0

Belgium

France

Netherland

Figure 5.4 Domestic in¯ation rates relative to Germany's in¯ation, ERM

(1990) question the modelling of the ERM as a credibility±reputation game. They argue that the assumptions of forward-looking behaviour, perfect competition and continuous market clearing contrast sharply with the experience of practitioners in the ERM countries. The crux of the matter lies in the ¯exibility of labour markets and the reaction of these markets to monetary regime shifts. Any serious investigation of the credibility theory of the ERM needs to consider the in¯uence of German in¯ation on the formation of domestic in¯ation expectations. Surprisingly, there has been little work that seeks to investigate empirically this transmission mechanism. Giavazzi and Giovannini (1988) estimated an `atheoretic' VAR system of price in¯ation, wage in¯ation and output growth, and then tested for parameter shifts after the ERM formation. But there was no testing of the credibility hypothesis. Anderton and Barrell (1995) and Egebo and Eglander (1992) tested for structural changes in the labour market, but they did not examine the credibility hypothesis. Artis and Nachane (1990) and Artis and Ormerod (1991) tested the in¯uence of German in¯ation on domestic in¯ation expectations, though the authors employed just an autoregressive scheme rather than a complete model for in¯ation expectations. Only Bleaney and Mizen (1995) examined the credibility hypothesis within an in¯ation prediction model that contained economic variables and German in¯ation.

Nicholas Sarantis 81

In this chapter we investigate the credibility theory of the European Monetary System for eight European countries, using data for the postSecond World War period. The chapter is organised as follows. Section 5.2 sets out the expectations-formation model used for testing the credibility hypothesis. Section 5.3 examines the measurement of variables. Section 5.4 analyses the empirical results for in¯ation expectations. Finally, Section 5.5 summarises the main empirical ®ndings.

5.2

Credibility and in¯ation expectations

The credibility theory of the ERM implies the following model for the expected price in¯ation rate:5 pet‡1 ˆ E…pt‡1 = t

…5:1†

1

According to this model, rational economic agents formulate unbiased and ef®cient expectations for the next period's in¯ation rate, pet‡1 , based on the latest set of information available, t 1 . This set of information includes domestic determinants of expectations plus the German rate of in¯ation, pg . The important question is what factors should enter the information set. Artis and Nachane (1990) and Artis and Ormerod (1991) used a purely autoregressive process, so price in¯ation was made a function of past in¯ation rates and the German in¯ation rate. But this is inconsistent with the rational expectations hypothesis (REH). The main argument put forward by proponents of the REH is that economic agents take into account not only past in¯ation but also changes in all economic variables relevant to price setting. What is required is a structural price determination model that would indicate the types of variable entering the information set. Bleaney and Mizen (1995) added the growth rate of output on the assumption of the Lucas supply curve. It is widely accepted that prices in industrial countries are determined in accordance with the mark-up model. Here we use an extended version of the price model employed by Sarantis (1991). Prices are determined by a mark-up on unit costs, the latter consisting of unit labour costs (ulc), the price of imported oil (po ) and commodities (pc ), exchange rates (e), interest rates (r), and indirect taxes (ti ), while the mark-up depends on the growth of economic activity, d (see Sarantis, 1994). The information set becomes:

t

1

ˆ …dt

1;

u1ct

1;

pot 1 ; pct 1 ; et

i 1 ; r t 1 ; tt 1 ;

pt

1;

pot 1 †

…5:2†

where all variables are measured in rates of change (see Appendix 5.1), apart from the interest rate, which is already in percentages.

82 Credibility and In¯ation

To generate the one-period-ahead forecast for the rate of price in¯ation, we estimate the following equation: pt ˆ t

2

‡ ut

…5:3†

and then use the estimated parameters, , to produce one-step-ahead forecasts: pet‡1 ˆ t

1

…5:4†

This method implies that only information available at time t±1 is used to generate in¯ation forecasts, which accords with the rational expectations hypothesis (see Artis and Ormerod, 1991; Hall, 1993). One could argue that some information might be available at time t, but as Hall (1993) points out, this not only makes it dif®cult to decide what information is available in the current period when forecasts are made, but also raises problems of multicollinearity. Furthermore, it should be pointed out that this is a forecast equation and not a structural price model where knowledge of the expected in¯ation rate is presupposed. Given the long sample period used in the econometric investigation, one could argue that the parameters i in Equation 5.3 may vary over time. To investigate this possibility we also consider a time-varying parameters version of Equation 5.3: pt ˆ t

2

t ˆ t

‡ ht

1

‡ ut

…5:5† …5:6†

where 5.5 is the measurement equation, 5.6 is the transition equation, the state vector t is the vector of time-varying parameters t , and t is a white noise error term with zero mean and is uncorrelated with ut . Following Cuthbertson et al. (1992) and Hall (1993), we estimate Equations 5.5 and 5.6 with the Kalman ®lter conditional on the variance of the error term, ut and the covariance of the error term, ht which is assumed to be diagonal.6

5.3

Data and measurement of variables

In this study we use annual data for the period 1952±92 for eight ERM countries: Belgium, Denmark, France, Germany, Ireland, Italy, the Netherlands and the United Kingdom. One advantage of using annual data is that it can help shed light on European price and wage equations for the entire post-war period, whereas studies using quarterly data tend to cover a much shorter period. The de®nition of variables and data sources

Nicholas Sarantis 83

are outlined in Appendix 5.1. However, the measurement of demand growth requires further explanation. According to the post-Keynesian view (for example, Sarantis, 1991; Sawyer, 1982) this variable should be measured by growth in the level of demand and not by excess demand. We therefore use two alternative measures: the growth rate of GDP, and growth in capacity utilisation. To examine the potential in¯uence of excess demand rather than of the level of demand, we use deviations of GDP from its trend as a proxy for excess demand.

5.4

Prediction equations and in¯ation expectations

To test the credibility hypothesis, we introduce German in¯ation during the period 1979 (when the ERM was founded) to 1992 into the price equation (5.3) for all non-German members of the ERM. Some authors (for example, Anderton and Barrell, 1995; Artis and Ormerod, 1991) argue that the ERM effects became more established after 1982, when a harder ERM is supposed to have been formed. Egebo and Englander (1992) identify 1987 as the crucial year, on the ground that nominal exchange rates became largely ®xed among ERM countries after that year. In view of the above, if German in¯ation over the period 1979±92 was insigni®cant for a particular country, we instead use German in¯ation for the 1982±92 period. If that is insigni®cant as well, we use the shorter period 1987±92. To separate the ERM-related credibility effect from the potential comovement of German and non-German in¯ation rates arising from stronger trade links during the post-war period, we also estimate the price equations using German in¯ation for the entire sample period. 5.4.1

Fixed parameter estimates

When applying Equation 5.3 to each country, we experimented with the three alternative proxies of demand and the reported results are based on the measure that produces the best estimates (in terms of econometric criteria and signi®cance of coef®cients). Changes in capacity utilisation yield the best estimates for four countries and output growth for the other four countries (Table 5.1). The initial regressions tend to fail either the normality or the functional tests due to one or two outliers (usually for 1974, when the ®rst oil crisis occurred). The use of dummies for such outliers removes these problems.7 The prediction equations pass all the normal econometric criteria of autocorrelation, normality, functional form and heteroscedasticity. The empirical results suggest that the credibility hypothesis is supported for two countries only: France and Belgium. The in¯uence of

Table 5.1 Forecasting equations for price in¯ation (®xed-parameters model) Variables

ulct dt

2

et

2

pct

2

2

pot 2 pot

3

tit 2 pt

2

pgt 2 pgw t 2 R2 S LMA(1) LMF(1) LMN(2) LMH(1)

Den. (1957±92)

France (1957±92)

Ger. (1955±92)

Ire. (1957±92)

Italy (1957±92)

Neth. (1957±92)

UK (1955±92)

0.019 (4.9) 0.312 (4.8) 0.090 (1.3) ± ± 0.102 (4.4) ± ± 0.016 (1.7) ± ± ± ± 0.535 (4.1) ± ± 0.783 0.013 0.947 2.198 0.415 0.587

0.020 (2.3) 0.115 (2.3) 0.238 (2.5) ± ± ± ± 0.028 (1.8) 0.030 (2.1) 1.460 (2.2) 0.381 (2.5) ± ± ± ± 0.617 0.020 3.471 0.214 0.675 1.508

±0.011 (0.7) 0.576 (6.5) 0.627 (2.4) 0.090 (2.4) 0.043 (1.4) ± ± ± ± ± ± ± ± 0.605 (2.5) ± ± 0.714 0.020 2.334 0.371 5.188 0.002

0.022 (8.3) 0.147 (2.1) 0.158 (3.1) 0.027 (1.4) ± ± 0.019 (3.1) 0.017 (2.7) ± ± ± ± ± ± ± ± 0.651 0.010 1.896 1.222 1.356 0.004

0.006 (0.7) 0.567 (8.2) 0.605 (3.7) ± ± 0.091 (2.6) 0.045 (3.2) ± ± ± ± ± ± ± ± ± ± 0.832 0.022 0.333 2.193 0.652 0.397

0.028 (4.1) 0.490 (6.9) 0.328 (2.7) ± ± ± ± 0.045 (3.0) 0.030 (1.7) ± ± ± ± ± ± ± ± 0.808 0.025 0.106 2.710 1.417 0.908

±0.004 (0.5) 0.173 (2.0) 0.411 (3.4) ± ± 0.065 (2.4) ± ± 0.015 (1.3) ± ± ± ± ± ± 0.685 (3.1) 0.625 0.016 3.691 0.603 1.346 0.669

±0.005 (0.3) 0.521 (5.2) 0.566 (2.0) ± ± 0.114 (3.4) ± ± ± ± ± ± ± ± ± ± 0.478 (2.1) 0.803 0.021 1.752 0.582 1.668 0.470

Notes: Growth of demand (d) is measured by changes in capacity utilisation [
ln(CU)t1 ] in Belgium, Denmark, Germany and Italy, and by output growth (y) in France, Ireland, the Netherlands and the UK. The variable pgw is the German price in¯ation rate for the whole sample period. R2 is the coef®cient of determination adjusted for degrees of freedom; S is the standard regression error; LMA is the Lagrange multiplier for autocorrelation; LMF is the Lagrange multiplier for functional form; LMN is the Lagrange multiplier for normality; LMH is the Lagrange multiplier for heteroscedasticity. All the LM statistics are distributed as 2 (v) with v degrees of freedom.

84

CON

Belg. (1956±92)

Nicholas Sarantis 85

German in¯ation for the shorter periods 1982±92 and 1987±92 is entirely insigni®cant (often with a t-value below unity) for all countries. Using German in¯ation for the entire sample period is equally unsatisfactory; it is insigni®cant for all countries except the Netherlands and the UK. A similar ®nding for the Netherlands is reported in Artis and Ormerod (1991) and re¯ects the close links between the German and the Dutch economies (in particular between their monetary systems). The ®nding for the UK is more dif®cult to interpret.8 Our negative results for the credibility hypothesis contrasts with the positive support reported in Artis and Ormerod (1991) and Bleaney and Mizen (1995). One the other hand, our results support the negative ®ndings by De Grauwe (1989), Fratianni and Von Hagen (1990) and Weber (1992). When comparing these studies, one needs to exercise caution, since they employ diverse methodologies and sample periods. Another important observation is that past in¯ation rates are signi®cant only in the case of Denmark. This raises doubts about the use of autoregressive schemes for predicting in¯ation rates, as in Artis and Ormerod (1991). In¯ation expectations for all countries are highly in¯uenced by movements in unit labour costs and excess demand, and by world commodity and oil prices.9 The parameter estimates shown in Table 5.1 are used to predict the next period's in¯ation rate, pet‡1 , in accordance with Equation 5.4. To test the unbiasedness and ef®ciency of these predictions, we estimate the following equations:10 pt‡1 ˆ 2 pet‡1 …pt‡1

pet‡1 † ˆ 1

pt‡1 ˆ 1 ‡

2 pet‡1

…5:7† …5:8† …5:9†

Equations 5.7 and 5.8 test for unbiasedness; this requires that 2 = 1 and 1 = 0. The joint test 1 = 0 \ 2 = 1 in Equation 5.9 tests for both unbiasedness and ef®ciency (see Hall, 1993). The statistics are reported in Table 5.2. The tests 2 = 1 and 1 = 0 are easily accepted at the 5 per cent signi®cance level for all countries except with regard to 2 = 1 test for Ireland (though it passes the 1 per cent signi®cance level), thus strongly supporting the unbiasedness hypothesis. The Wald statistics for the joint hypothesis 1 = 0 \ 2 = 1 pass a 5 per cent signi®cance level test for all countries except Denmark, Ireland and Italy (though it passes the 1 per cent signi®cance level). Hence, the in¯ation expectations series generated by our model satisfy the crucial assumptions of unbiasedness and ef®ciency for almost all countries.

86 Credibility and In¯ation Table 5.2 Tests for unbiasedness and ef®ciency (®xed-parameters model) Tests

Belg.

Den.

Fr.

Ger.

Ire.

Italy

Neth.

UK

1 2 1 2

0.080 1.023

0.251 0.794

0.094 0.630

0.953 0.061

0.656 2.118

0.011 1.340

0.130 0.562

0.332 0.426

4.968

6.786

2.387

5.310

8.460

6.664

2.187

1.904

=0 =1 = 0\ =1

Notes: The statistics for 2 = 1 and 1 = 0 are t-values; the 5 per cent and 1 per cent critical values are 2.044 and 2.75 respectively. The statistic for the joint test 1 = 0 \ 2 = 1 is the Wald test, which is distributed as 2 (2) on the null; the 5 per cent and 1 per cent critical values are 5.99 and 9.21 respectively.

5.4.2

Time-varying parameters model

An important aspect in the speci®cation of the time-varying parameters model is the behaviour of the residuals in the measurement Equation 5.6, estimated by the Kalman ®lter technique.11 Table 5.3 shows the diagnostic statistics for normality and serial correlation. The Jarque±Berra statistics suggest that the residuals are normally distributed in all countries. Both the Ljung±Box and Box±Pierce statistics strongly reject the null hypothesis of serial correlation (for varying lags) in the error term process for all countries. In addition to the absence of serial correlation, the expectations series generated by the model should not be consistently biased for the learning model 5.5±5.6 to be weakly rational (see, Hall, 1993). Following the same procedure as for the ®xed-parameters model, we generate one-periodTable 5.3 Diagnostic statistics for the time-varying parameters model Bel. Den. Fr. Ger. Ire. Italy Neth. UK (1956±92) (1957±92) (1957±92) (1955±92) (1957±92) (1957±92) (1957±92) (1955±92) Log L 106.01 Normal (J±B) 3.483 LB(1) 0.21 LB(2) 0.60 LB(4) 7.99 BP(1) 0.19 BP(2) 0.52 BP(4) 6.60

96.78 1.142 0.11 0.20 1.52 0.10 0.18 1.26

111.43 3.343 0.32 3.64 6.70 0.29 3.16 5.69

117.19 5.190 1.58 1.77 3.27 1.42 1.58 2.80

80.95 1.466 0.61 3.26 3.41 0.54 2.78 2.90

75.69 4.329 0.03 0.28 2.11 0.02 0.24 1.70

101.80 2.378 3.50 3.51 4.05 3.14 3.15 3.58

104.62 3.361 0.23 0.87 5.12 0.21 0.77 4.35

Notes: Log L is the maximum of the likelihood function. J±B is the Jarque±Berra test for normality, which is distributed as 2 (2) on the null. LB(v) and BP(v) are the Ljung±Box and Box±Pierce statistics, respectively, for vth order autocorrelation, which are distributed as 2 (v) on the null. The 5 per cent critical values for v = 1, v = 2 and v = 4 are 3.84, 5.99 and 9.49 respectively.

Nicholas Sarantis 87 Table 5.4 Tests 1 2 1 2

= = = =

0 1 0 1

Tests for unbiasedness and ef®ciency (time-varying parameters model) Bel.

Den.

Fr.

Ger.

Ire.

Italy

Neth.

UK

0.129 4.302

0.582 3.641

1.378 2.935

0.008 2.088

0.995 3.000

0.338 2.868

0.153 2.181

0.718 1.481

68.37

53.70

14.98

25.62

29.84

37.87

18.17

15.41

Notes: The statistics for 2 = 1 and 1 = 0 are t-values; the 5 per cent and 1 per cent critical values are 2.044 and 2.75 respectively. The statistic for the joint test 1 = 0 \ 2 = 1 is the Wald test which is distributed as 2 (2) on the null; the 5 per cent and 1 per cent critical values are 5.99 and 9.21 respectively.

ahead in¯ation forecasts for each country, and then test for unbiasedness and ef®ciency by applying the tests 5.7±5.9. The results are reported in Table 5.4. The statistics in Table 5.4 indicate that the test 1 = 0 cannot be rejected for any country. However, this test imposes the restriction 2 = 1, so this should be tested before drawing any inferences about unbiasedness. The results suggest that the test 2 = 1 cannot be accepted at the 5 per cent signi®cance level for any of the original members of the ERM, though the statistic for Germany is close to its critical level. The only country to pass this test is the UK. The Wald statistics for the joint hypothesis of unbiasedness and ef®ciency, 1 = 0 \ 2 = 1, is strongly rejected at every signi®cance level for all countries. These ®ndings indicate that the in¯ation expectations produced by the time-varying parameters model are both biased and inef®cient.

5.5

Conclusions

This chapter has investigated price in¯ation expectations in the ERM member countries during the post-Second World War period with the aim of assessing the importance of the credibility theory of the ERM. We have employed a ®xed parameter and a time-varying parameter model for the in¯ation-prediction equations. The latter model is both biased and inef®cient. The in¯ation expectations generated by the ®xed-parameter model, on the other hand, are both unbiased and ef®cient for almost all countries, so this model provides a satisfactory expectations-generating mechanism. We have been unable to ®nd any strong support for the credibility theory of the ERM. Only the Belgian and French in¯ation expectations have been in¯uenced by German in¯ation since the formation of the ERM in 1979. Tests for the shorter periods of 1982±92 and 1987±92, which

88 Credibility and In¯ation

re¯ect the formation of a harder ERM, failed to identify credibility gains emanating from importing the Bundesbank's reputation, thus undermining the credibility hypothesis. This evidence questions the tendency to model the ERM as a credibility± reputation game. As Egebo and Englander (1992) and Goodhart (1990) argue, credibility might be of secondary importance to labour market imperfections and rigidities, backward-looking behaviour and the reaction of labour markets to monetary regime shifts. This does not imply that credibility is not important, but any ERM-induced bene®ts for domestic in¯ation are more likely to arise from gradual institutional changes in product and labour markets and the monetary sector rather from borrowing the Bundesbank's reputation. The pricing forecasting equations indicate that in¯ation expectations for all countries are strongly in¯uenced by movements in unit labour costs and demand, and by world commodity and oil prices. Past in¯ation is signi®cant only in Denmark. Combined with the unbiasedness of the generated expectations series, this evidence seriously undermines the use of autoregressive schemes for price in¯ation in previous studies.

Appendix 5.1 p w y (CU)

d Q

q (ulc) Ti ti

Data employed

Price in¯ation rate (log difference of the consumer price index, 1985 = 100) (IMF international ®nancial statistics Wage in¯ation rate (log-difference of the wage index, 1985 = 100) (IMF international ®nancial statistics) Real GDP growth rate (log-difference), at 1985 prices (OECD national accounts) Index of capacity utilisation computed from (CU)t = [IIP/ (IIP)*]t *100, where (IIP) is the index of industrial production (1985 = 100) (IMF international ®nancial statistics) and (IIP)* is the maximum attainable or full capacity output, which was estimated with the Warton School method for each country Growth of aggregate demand, measured by y and Dln(CU) Productivity (real GDP/total employment), converted to an index form (1985 = 100). Data for employment were obtained from the ILO Yearbook of Labour Statistics Productivity growth (log difference of Q) Growth of unit labour costs (w ± q) Indirect tax rate (indirect taxes/GDP at current prices)*100 (OECD national accounts Log difference of T i

Nicholas Sarantis 89

e po pe

(Log difference) Exchange rate index (1985 = 100) (IMF international ®nancial statistics) Growth (log-difference) of Saudi Arabian petrol prices (1985 = 100) (IMF international ®nancial statistics) Growth (log difference) of world commodity prices, excluding fuel (1985 = 100) (IMF international ®nancial statistics)

Notes 1. This analysis builds on the Barro±Gordon model, applied to open economies. 2. This view is a variant of the `disciplinary' argument against ¯oating exchange rates. 3. France, Italy, the Netherlands, Belgium, Denmark and Ireland were among the original ERM members. The United Kingdom joined the ERM only in 1990 (though it had been a member of the EMS since 1979), but from 1985 the UK government pursued a monetary policy whereby the pound shadowed the Deutschmark. 4. Those sceptical of the ERM effect on in¯ation point out that Japan and the United States also reduced their in¯ation rates to low levels during the 1980s, though they did not ®x their exchange rates. 5. See Bleaney and Mizen (1995). 6. For a detailed analysis of the econometrics of the Kalman ®lter and its application to time-varying parameter models, see Cuthbertson et al. (1992, ch. 7). 7. Note that Artis and Ormerod (1991) followed a similar procedure. 8. One possibility is that German in¯ation might act as a proxy for imported in¯ation. We also experimented with the use of the price de¯ator for aggregate imports of goods and services, but this further weakened the in¯uence of German in¯ation, even for France and Belgium, and the overall results were less satisfactory. 9. It is interesting to note that interest rates were insigni®cant in all countries. 10. Note that the Kalman ®lter process always takes a few periods to settle down and track the underlying learning mechanism, so the initial predictions (for the period 1956±60) of the time-varying parameter model are disregarded. Therefore, to provide a fair comparison of the two models, we have estimated Equations 5.7±5.9 over the period 1961±92. 11. Note that the parameter values and their respective standard errors reported by the Kalman ®lter algorithm represent deviations from the ®xed parameters and are therefore meaningless. That is why we do not report them. The Kalman ®lter algorithm does, of course, generate graphs of the individual time-varying parameters, but given the large number of coef®cients and countries it is impossible to report them due to space constraints.

References Anderton, R. and R. Barrell (1995) `The ERM and Structural Change in European Labour Markets: A Study of 10 Countries', Weltwirtschaftliches Archiv, Band 131, pp. 49±66.

90 Credibility and In¯ation Artis, M. and D. Nachane (1990) `Wages and Prices in Europe: A test of the German Leadership Thesis', Weltwirtschaftliches Archiv, Band 131, pp. 59±77. Artis, M. and P. Ormerod (1991) `Is There an ``EMS'' Effect in European Labour Markets?', CEPR Discussion Paper, no. 598 (London: CEPR, December). Bleaney, M. and P. Mizen (1995) `Credibility and Disin¯ation in the European Monetary System', Paper presented at the 10th Annual Congress of the European Economic Association, Prague. Collins, S. M. (1988) `In¯ation and the European Monetary System', in F. Giavazzi et al. (eds), The European Monetary System (Cambridge: Cambridge University Press). Cuthbertson, K., S. Hall and M. Taylor (1992) Applied Econometric Techniques (Brighton: Harvester Wheatsheaf). De Grauwe, P. (1989) `Disin¯ation in the EMS and in the Non-EMS Countries: What Have we Learned?', Empirical-Austrian Economic Papers, vol. 16, pp. 161±76. De Grauwe, P. (1996) `In¯ation Targeting to Achieve In¯ation Convergence in the Transition Towards EMU', CEPR Discussion Paper, no.1457 (London: CEPR). Egebo, T. and S. Englander (1992) `Institutional Commitments and Policy Credibility: A Critical Survey and Empirical Evidence from the ERM', OECD Economic Studies, no. 18 (Paris: OECD) pp. 45±84. Franz, W. and R. J. Gordon (1993) `German and American Wage and Price Dynamics: Differences and Common Themes', European Economic Review, vol. 37, pp. 719±62. Fratianni, M. and J. Von Hagen (1990) `Credibility and Asymmetries in the EMS', in V. Argy and P. De Grauwe (eds), Choosing an Exchange Rate Regime ± The Challenge for Smaller Industrial Countries (Washington, DC: IMF). Giavazzi, F. and A. Giovannini (1988) `The Role of the Exchange-Rate Regime in a Disin¯ation: Empirical Evidence on the European Monetary System', in F. Giavazzi et al. (eds), The European Monetary System (Cambridge: Cambridge University Press). Giavazzi, F. and M. Pagano (1988) `The Advantage of Tying One's Hands ± EMS Discipline and Central Bank Credibility', European Economic Review, vol. 32, pp. 1055±82. Goodhart, C. (1990) `Economists' Perspectives on the EMS ± A Critical Essay', Journal of Monetary Economics, vol. 26, pp. 471±87. Hall. S. (1993) `Modelling the Sterling Effective Exchange Rate Using Expectations and Learning', The Manchester School, vol. L X I , pp. 270±86. Melitz, J. (1988) `Monetary Discipline, Germany and the European Monetary System: A Synthesis', in F. Giavazzi et al. (eds), The European Monetary System (Cambridge: Cambridge University Press). Sarantis, N. (1991) `Con¯ict and In¯ation in Industrial Countries', International Review of Applied Economics, vol. 5, pp. 155±69. Sarantis, N. (1994) `In¯ation', in P. Arestis and M. Sawyer (eds), The Elgar Companion to Radical Political Economy (Aldershot: Edward Elgar). Sawyer, M. (1982) `Collective Bargaining, Oligopoly and Macroeconomics', Oxford Economic Papers, vol. 34, pp. 428±48. Weber, A. (1992) `The Role of Policymakers' Reputation in EMS Disin¯ation: An Empirical Evaluation', European Economic Review, vol. 36, pp. 1473±92.

6

Exchange-Rate Regimes, In¯ation and Credibility: Evidence from Greece1 George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos

6.1

Introduction

Recent theories of in¯ation stress how the limited ability of policy makers to commit themselves to price stability results in high in¯ation without any employment bene®ts. This has been shown by Barro and Gordon (1983) in a sequential move game between wage setters and a central bank along the Phillips curve. An important extension of this model is the exchange rate regime model of Giavazzi and Giovannini (1987) and Giavazzi and Pagano (1988). These authors have shown that participation in a regime of exchange rate commitments can give an institutional solution to the inef®cient Barro±Gordon outcome.2 In particular, participation in a regime of ®xed exchange rates, in which monetary policy is determined by an in¯ation-averse core country, ties the hands of domestic policy makers. This is re¯ected in wage-setters' expectations, and so the domestic economy ends up with the same average in¯ation as the core country.3 This chapter presents a version of the exchange rate regime model of in¯ation and investigates its applicability to the post-war Greek economy. Greece is particularly suitable for such an investigation because (1) governments have been responsible, under the terms of the 1975 constitution, for maintaining full employment (this is in line with the Barro±Gordon model of in¯ation), and (2) it has had long experience of both ®xed and independently managed ¯oating exchange rates (this is in line with the exchange rate regime model of in¯ation). Therefore, this chapter investigates the joint determination of wage in¯ation, price in¯ation and unemployment by imposing the cross-equation restrictions of the Barro±Gordon model and the exchange rate regime model. By

91

92 Exchange-Rate Regimes, In¯ation and Credibility

doing so we respect the Lucas critique,4 and also obtain estimates of the structural parameters of the model.5 The theoretical model is as follows. Under ®xed exchange rates there is no monetary policy independence, hence in¯ation is exogenously determined. In contrast, under managed ¯oating exchange rates there is room for policy independence, hence in¯ation is endogenously determined via a Barro±Gordon game between policy makers and wage setters. The model leads to structural equations for wage in¯ation, price in¯ation and unemployment. We test and estimate the model for Greece during 1960±94. We take a two-stage approach to estimation. We ®rst investigate the integration and cointegration properties of the data. Indeed, given these properties, we transform the theoretical model so as to achieve stationarity, and then estimate it. The cross-equation theoretical restrictions are not rejected by the data, and all crucial coef®cients are signi®cant and have plausible values. The main empirical result is that the exchange-rate regime matters for wage and price in¯ation. In particular, wage and price in¯ation have been higher, and more persistent, during periods of managed ¯oating than during the Bretton Woods period. After the fall of the Bretton Woods system of ®xed exchange rates in 1972, there was a signi®cant Barro± Gordon type of in¯ation bias due to the unwillingness of policy makers to commit themselves to low in¯ation. The policy implication is that membership of an exchange rate mechanism can make the disin¯ation process easier. However, as we discuss below, it is important to correct the economic fundamentals before moving to such a mechanism (see also Dornbusch, 1991). The rest of the chapter is organised as follows. Section 6.2 presents the theoretical model, Section 6.3 presents the econometric model and results, and Section 6.4 discusses policy implications and possible extensions.

6.2

The theoretical model

In each time-period t, wage setters sign one-period nominal wage contracts before they observe the current price level, or equivalent price in¯ation. The way price in¯ation is determined depends on the nominal exchange rate regime. The institutional constraints of the Bretton Woods system of ®xed exchange rates reduced the ability of member countries (except the United States) to follow an independent monetary policy. Since Greece participated fully in the Bretton Woods system, we assume that, until its collapse in 1972, price in¯ation was exogenously

George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos 93

determined. In contrast, in the post-1972 period we assume that policy makers were free to choose price in¯ation by playing a Barro±Gordon game with wage setters (see, McKinnon, 1993, for the international ®nancial system). For simplicity, we will use a log-linear and deterministic model. 6.2.1

The labour market

Output yt is produced by using labour `t via a Cobb-Douglas function. So yt = `t ‡ t , where 0 <  < 1. Productivity t follows the exogenous deterministic process t ˆ g ‡ t 1 , where g is a constant. Pro®t maximisation by competitive ®rms leads to demand for labour: `dt ˆ

…wt

pt

t †

…6:1†

where  [1/(1 ± )] > 0, wt is the nominal wage rate and pt is the price level at time t. Nominal wages are set by a group of insiders n t , where n t follows an exogenous deterministic process (de®ned below). On the other hand, policy makers care about the whole labour force, n, where n is exogenous and constant and n > n t as in Barro and Gordon (1983). Insiders n t solve: min Et …`dt wt

nt †2 subject to …6:1†

…6:2†

where Et is the rational expectations operator. Since the model is deterministic, and due to the sequential-move structure of the game, only pt is not observable by wage setters when they choose wt .6 Using the ®rst-order condition of (6.2) into (6.1), we get for unemployment ut  …n `t ), and wage in¯ation
wt  …wt wt 1 ): ut ˆ ut

…
pt

wt ˆ g ‡ Et
pt

Et
pt † 1 …ut

1

…6:3† ut †

…6:4†

where u t  …n nt ) is natural unemployment,
pt  pt pt 1 ) is price in¯ation and Et
pt  (Et pt pt 1 ) is expected price in¯ation. Equation (6.3) is a surprise supply function and Equation (6.4) is a Phillips-curve type expression. In what follows, we set = 1 for notational simplicity.7 6.2.2

Solution under ®xed exchange rates

Under ®xed exchange rates (denoted by a superscript f ), price in¯ation is exogenously determined. In particular we assume that price in¯ation is exogenous and constant at a rate f > 0:8 Therefore, by using

94 Exchange-Rate Regimes, In¯ation and Credibility

Equations 6.3 and 6.4, we have the following, respectively, for equilibrium price in¯ation, wage in¯ation and unemployment:9 f


pt ˆ f f
wt f ut

6.2.3

…6:5a†

ˆg‡

f

…ut

1

ut †

…6:5b†

ˆ ut Solution under managed ¯oating exchange rates

Under managed ¯oating exchange rates (denoted by a superscript m), in¯ation is endogenously determined via a Barro±Gordon game between policy makers and wage setters. Policy makers solve:   2 m 2 min …
pm m subject to …6:3† …6:6† t t † ‡ …ut † m
t

where m t is the policy makers' exogenous target rate of price in¯ation at t under managed ¯oating (de®ned below). We make the (testable) f assumption that m t , >  (that is, policy makers' target in¯ation under managed ¯oating is higher than in¯ation under ®xed rates). The parameter  > 0 is the weight given to unemployment relative to in¯ation. The equilibrium of this one-shot game is the Nash equilibrium. In this equilibrium we have the following, respectively, for price in¯ation, wage in¯ation and unemployment: m
pm t ˆ t ‡ ut


wtm um t ˆ

ˆg‡

…m t

‡ ut †

…6:7a† …ut

1

ut †

ut

…6:7b† …6:7c†

f Since m t >  ; a comparison of Equations 6.5 and 6.7 implies that wage and price in¯ation are higher under managed ¯oating than under ®xed exchange rates, and that (in the absence of uncertainty) unemployment equals its natural rate under both exchange rate regimes. This completes the theoretical model.

6.3

The econometric model: Greece 1960±94

This section presents the econometric speci®cation of the model (6.5 and 6.7), and discusses estimation and testing using Greek annual data for the period 1960±94.10 We proceed in three steps. First, in order to derive an estimable model we specify the exogenous targets and introduce dummy variables to capture policy-induced parameter changes. Second, we

George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos 95

investigate the integrating and cointegrating properties of the data. Third, we test and estimate the econometric model. 6.3.1

Exogenous targets and dummy variables

The model (6.5 and 6.7) cannot be estimated as it stands before we specify the exogenous targets, u t and m t , and introduce appropriate dummy variables to capture changes in the exchange rate regime (®xed and managed ¯oating). To model the employment target of wage setters, n t , we assume that n t is a weighted average of those who were employed in the previous period, `t 1 , and the exogenous labour force, n. Thus n t = `t 1 ‡ (1 ± )n, where 0    1. Then, by using ut  …n `t ) and u t  (n ± n t ) we get: ut ˆ ut

…6:8†

1

where 0    1. m To model the in¯ation target of policy makers, m t , we assume that t is a linear function of lagged-once in¯ation (that is, the higher the inherited in¯ation rate, the more costly the policy adjustment required to reduce it to an arbitrary rate). Thus we assume that: m m t ˆ  ‡ 
pt

…6:9†

1

where we make the (testable) assumption that m > f . Note that while Equation 6.9 accounts (via, 
pt 1 , where   0) for persistent in¯ation under managed ¯oating exchange rates, this is not the case under ®xed exchange rates (see Equation 6.5a above). This is deliberate in order to simplify notation; there is no empirical evidence of persistent in¯ation during the Bretton Woods period. f We continue with dummy variables. De®ne dt (respectively dtm ) to be a dummy for ®xed (respectively managed ¯oating) exchange rates that takes the value of 1 during the pre-1972 (respectively post-1972) period and zero otherwise. By using Equations 6.8 and 6.9 and the exchange rate dummies, the wage equations (6.5b and 6.7b), the price equations (6.5a and 6.7a) and the unemployment equations (6.5c and 6.7c) can be summarised by: f


wt ˆ gw1 dt ‡ gw2 dtm ‡ gw3 dtm
pt
pt ˆ

f gp1 dt

ut ˆ gut ut

1

‡

gp2 dfm

‡

gp3 dtm
pt 1

1

‡ gw4 ut

‡

1

‡ gw5 dtm ut

gp4 dtm ut 1

1

…6:10a† …6:10b† …6:10c†

96 Exchange-Rate Regimes, In¯ation and Credibility

where gw1 ˆ …g ‡ f †; gw2 ˆ …g ‡ m †; gw3 ˆ ; gw4 ˆ

…1

†; gw5 ˆ ;

gp1 ˆ f ; gp2 ˆ m ; gp3 ˆ ; gp4 ˆ ; gu1 ˆ 

…6:11†

[(6.10a)±(6.10b)±(6.10c) is a system of reduced-form equations that displays the theoretical cross-equation restrictions (6.11). All variables are observable, so this is an estimable equation system. 6.3.2

Integration and cointegration properties of the data

Univariate analysis indicates that the series
wt ,
pt and ut are each individually I(1) with constant. In particular, Dickey±Fuller (DF) t-statistics (allowing for the possibility of a constant, and with zero-lagged differenced terms)11 testing for unit roots in
wt and
pt and ut have values of ±2.55, ±1.75 and 0.04 respectively, which are all higher than the 5 per cent Mackinnon critical value of ±2.95. Thus, the null of a unit root in the series
wt ,
pt and ut is accepted. In turn, we test for second-order integration. The null of a unit root is easily rejected when
wt and
pt are ®rstdifferenced, while the null of a unit root in
ut can be rejected at the 10 per cent level (but not the 5 per cent level).12 Taken together, these results suggest that the hypothesis that
wt ,
pt and ut are I(1) with constant provides a good description of the statistical properties of the data. Next we look for cointegrating relationships, and if necessary transform the model so as to achieve stationarity (see also, Banerjee et al., 1993; King et al., 1991). The ®rst regression (6.10a) contains the series
wt ,
pt and ut . The Johansen maximum likelihood test does not reject the null that (
wt ,
pt , ut ) are cointegrated. In particular, when we test the hypothesis of one cointegrating equation against the hypothesis of no cointegrating equations, the likelihood ratio is 10.34, which is less than the 5 per cent critical value of 19.96. Hence, the null that there is one cointegrating equation is accepted.13 Next consider the second regression (6.10b). This is an unbalanced regression because it explains an I(0) variable (that is, differenced price in¯ation (
pt
pt 1 ), by an I(1) variable (that is unemployment, ut ). To avoid possible spurious regression problems we shall therefore use the ®rst-difference
ut 1 , rather than the level ut 1 , when we estimate the model (our estimation results ± available upon request ± do not depend on whether we use ut 1 or
ut 1 as a regressor in Equation 6.10b). Finally, consider the third regression (6.10c). This regression can also be spurious. We shall therefore ®rst-difference it when we estimate the model (again our estimation results ± available upon

George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos 97

request ± do not depend on whether we use the levels regression, 6.10c, or its ®rst-differenced version).14 6.3.3

Tests and estimation results

The ®nal econometric model is: f


wt ˆ gw1 dt ‡ gw2 dtm ‡ gw3 dtm
pt
pt ˆ

f gp1 dt

‡


ut ˆ gu1
ut

gp2 dtm

‡

gp3 dtm
pt 1

1

‡ gw4 ut

‡

gp4 dtm

1

‡ gw5 dtm ut

‡
ut

1

1

…6:12a† …6:12b† …6:12c†

1

subject to the cross-equation restrictions (6.11). Given the integration and cointegration properties of the data, this transformed model does not suffer from stationarity problems.15 We emphasize that our estimation results are robust to transformations being made to achieve stationarity. By estimating the equation system (6.12a)±(6.12b)±(6.12c) subject to (6.11), we respect the Lucas critique and obtain estimates of the structural parameters (g, f , m , , , ). The data do not reject the overidentifying restrictions (6.11). In particular, restricted and unrestricted SUR estimates of the reduced-form equation system (6.12a)±(6.12b)±(6.12c) give an Fstatistic of 0.464, while the relevant 2 statistic is equal to 2.787. These calculated values are much lower than the critical values at any conventional level and the appropriate degrees of freedom. Thus, we cannot reject the cross-equation restrictions (6.11). Restricted SUR estimates are reported in Table 6.1. All estimated coef®cients (except ) have signs and magnitudes that are consistent with the theory and are statistically signi®cant. Note that m is signi®cantly higher by 6 per cent than f . That is, price and wage in¯ation are higher during managed ¯oating than during the Bretton Woods Table 6.1 Restricted SUR estimates of (6.12a)±(6.12b)±(6.12c) subject to (6.11), Greece, 1960±94 Parameter G f m   A

Estimate

t-statistic

0.077 0.020 0.080 0.511 0.692 ±1.132

8.57 2.36 4.40 4.54 6.31 4.27

Note: The adjusted R2 , the standard error of regression () and the Durbin±Watson statistic (DW) for (6.12a), (6.12b) and (6.12c) are respectively: R2 ± 0.683, 0.798, 0.393;  ± 0.038, 0.034, 0.006; DW ± 1.928, 1.889, 1.594.

98 Exchange-Rate Regimes, In¯ation and Credibility

period. Also, price and wage in¯ation persist over time during managed ¯oating by  = 0.511, while in¯ation persistence is insigni®cant during Bretton Woods. These results are consistent with the predictions of the exchange rate regime model. The parameter  (that is, the relative weight of insiders on wage-setting) is estimated to be 0.692, which is a very plausible value. Only the estimate of  (that is, the relative weight given to unemployment relative to in¯ation) does not have the right sign. We believe that this is due to the negative Phillips-curve type effect of unemployment on in¯ation, which more than outweighs the positive Barro±Gordon type effect. Finally, we report that our results are robust to a number of changes in the model; for example, the introduction of a dummy variable for incomes policy in 1986±87 and the introduction of a dummy variable for oil price shocks in the 1970s.

6.4

Conclusions, policy implications and extensions

Our results show that after the fall of the Bretton Woods system in 1972, there was an in¯ation bias due to the inability of policy makers in Greece to commit themselves to low in¯ation. This suggests strong support for the exchange rate regime model of in¯ation. Although these results are consistent with the widely accepted belief that exchange rates affect the in¯ation process, we should emphasise that they do not imply that low in¯ation can be attributed only to exchange rate policy. For instance, we know from the recent European experience that there is a common in¯ation performance independent of EMS membership (see Dornbusch, 1991). That is, a country can achieve disin¯ation on its own. However, the interesting question is not whether a country can disin¯ate on its own, but whether participation in an exchange rate mechanism (for example, ®xed exchange rates or target zones for the exchange rate) can make the disin¯ation process easier. If such participation allows a form of incomes policy and breaks the inertia of in¯ation, it cannot make disin¯ation harder. If, however, nominal wages are driven without regard for in¯ation and exchange rate targets, any use of a nominal exchange rate commitment to control in¯ation will sooner or later result in overvaluation and stag¯ation, and eventually in speculative attacks on the currency. Hence, it is important to work further on disin¯ation before moving to an exchange rate mechanism. Disin¯ation therefore requires the correction of economic fundamentals. Only when these fundamentals have been corrected can commitment to an exchange rate mechanism function as an effective preventive

George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos 99

device against the repetition of high-in¯ation episodes (see also, Cukierman, 1994; Dornbusch, 1991). We close with some extensions. First, it would be interesting to introduce traded and non-traded goods in order to have different effects on prices and exchange rates (see also, Rogoff, 1985b). Second, it would be interesting to introduce exchange rate uncertainty (as in Engel and Hamilton, 1990, and Kaminsky, 1993) in order to estimate how wage setters' beliefs about switches in exchange rate policy, and hence possible `peso' problems, affect in¯ation and unemployment. Third, it would be interesting to introduce a public sector in order to examine how employment and wages in the public sector affect in¯ation and employment in the private sector. Finally, it would be interesting to use a model where the nominal exchange rate regime matters for real variables, in order to study the joint determination of in¯ation and competitiveness across regimes. We leave these extensions for future work. Notes 1. We thank M. Ercolani, C. Li and E. Tzavalis for comments and suggestions. All errors are ours. 2. For other solutions and a recent survey of the literature, see, for example, Svensson (1997). 3. Coles and Philippopoulos (1997) extended this result to a target zone model. The crucial assumption of all the literature on imported credibility is that exchange rate commitments (for example, ®xed exchange rates and target zones) are more credible that domestic anti-in¯ationary policy announcements. On the other hand, it is known from the literature on speculative attacks that exchange rate commitments can collapse, and that intervention (for example, capital controls, borrowing of foreign reserves, increases in interest rates) can only delay the collapse. While such a credibility problem has characterised most of the ERM/EMS period and the last years of Bretton Woods, it was not a serious problem in most of the Bretton Woods era (see, McKinnon, 1993). Besides, capital controls enhanced the credibility of Bretton Woods; this was particularly true for a country such as Greece where regulation was the economic dogma. Also recall that Greece is not yet a member of the ERM/EMS. 4. That is, by imposing the theoretical cross-equation restrictions we can identify the `correct' model and also deal with policy-induced parameter changes. 5. Our methodology is therefore similar to that of Alesina and Sachs (1988) and Alogoskou®s et al. (1997), speci®cally, Alesina and Sachs (1988) test and estimate, and Alesina's (1987) rational partisan model for the United States. Alogoskou®s et al. (1997) combine the exchange rate regime model with the rational partisan model, and then test and estimate it for Greece. So the present chapter is a modi®ed version of Alogoskou®s et al. (1997); however, here we focus on exchange rate policy. Finally, note that our methodology differs from that of Alogoskou®s et al. (ibid., 1992) and Alogoskou®s and

100 Exchange-Rate Regimes, In¯ation and Credibility

6. 7.

8.

9.

10. 11.

12.

13. 14.

Philippopoulos (1992), whose results for the UK and Greece respectively are based only on unrestricted single-equation estimates. Therefore we do not consider the issue of stabilisation in the presence of unexpected shocks. See Rogoff (1985a) for the trade-off between ¯exibility for stabilisation and credibility. Although here we do not model exchange rates (actual or expected), they can be easily introduced. For instance, we can use a small open economy model with purchasing power parity. Then our solution remains unchanged, if we replace domestic prices with exchange rates (see, for example, Horn and Persson, 1988). This is especially true for a country such as Greece, in which the traded sector is particularly important, and hence there is a strong link between price in¯ation and exchange rate depreciation. Alternatively we can use a more general model with traded and non-traded goods. Again this would not change our qualitative results (see, for example, Rogoff, 1985b). Under Bretton Woods, it was only the United States that could follow an independent monetary policy. Therefore, assuming that f is exogenous is not restrictive since it was unlikely that the Federal Bank took Greek economic fundamentals into account when determining its monetary policy. We assume that exchange rate regimes are fully credible and that wage setters fully predict regime switches. Such deterministic modelling has been often used in the literature on exchange rates and speculative attacks. Alternatively, we could use probabilistic uncertainty about the exchange rate regime (see, for example, Engel and Hamilton, 1990; Kaminsky 1993). Here we use a deterministic model because the period of uncertainty about the viability of Bretton Woods did not last long, and because Greece has been following some form of managed ¯oating exchange rates since 1973 (Greece is not yet a member of the Exchange Rate Mechanism of the European Monetary System). Wage series (hourly earning in manufacturing, 1990 = 1), price series (retail price index, 1990 = 1) and the unemployment rate are from the OECD's main economic indicators. There is nothing in our theoretical model (6.10a)±(6.10c) to suggest the presence of time trends and/or lagged differenced terms. However, we have allowed for the possibility of a trend and have used augmented DF tests with lagged differenced terms. These tests reveal the same conclusions. Here we report DF tests with constant and zero lagged differenced terms. In particular the DF t-statistic test for a unit autoregressive root in
ut has a value of -2.61, which is higher than the 5 per cent Mackinnon critical value of -2.95 (hence, the null of I(2) is accepted at 5 per cent), but equal to the 10 per cent Mackinnon critical value. Since the introduction of exchange rate regime dummies reduces even further the estimated DF t-statistic, it seems reasonable to reject the null of I(2). The null of two cointegrating equations cannot be rejected either. However, only one of them satis®es the theoretical sign restrictions. Alternatively we could impose  = 1 in Equation 6.10c so that unemployment would follow a random walk. However, in this case the restriction  = 1 should be also imposed in Equations 6.10a and 6.10b. The null that  = 1 in (6.10a)± (6.10b)±(6.10c)±(6.11) is easily rejected by the data. Therefore, we prefer to ®rst-difference equation 6.10c. We report that the nature of our estimation results is robust to the transformation we use.

George S. Alogoskou®s, Apostolis Philippopoulos and Vanghelis Vassilatos 101 15. When we test for unit roots in the residuals from the ®nal restricted regressions (that is, (6.12a)±(6.12b)±(6.12c) subject to (6.11)), we can reject the null. That is, the residuals follow a stationary pattern. Hence, there is no inconsistency between the statistical properties of the unrestricted data and the residuals from the econometric model.

References Alesina, A. (1987) `Macroeconomic policy in a two-party game with rational voters', Quarterly Journal of Economics, vol. 102, pp. 651±78. Alesina, A. and J. Sachs (1988) `Political parties and the business cycle in the United States', Journal of Money, Credit and Banking, vol. 20, pp. 63±82. Alogoskou®s, G. (1995) `The two faces of Janus: Institutions, policy regimes and macroeconomic performance in Greece', Economic Policy, vol. 20, pp. 149±92. Alogoskou®s G., D. H. Lee and A. Philippopoulos (1997) `Exchange-rate regimes, political parties and the in¯ation-unemployment tradeoff: Evidence from Greece', Open Economies Review. Alogoskou®s, G., B. Lockwood and A. Philippopoulos (1992) `Wage in¯ation, electoral uncertainty and the exchange rate regime: Theory and the UK evidence', Economic Journal, vol. 102, pp. 1370±94. Alogoskou®s G. and A. Philippopoulos (1992) `In¯ationary expectations, political parties and the exchange rate regime: Greece 1958±1989', European Journal of Political Economy, vol. 8, pp. 375±99. Banerjee A., J. Dolado, J. Galbraith and D. Hendry (1993) Cointegration, ErrorCorrection, and the Econometric Analysis of Nonstationary Data (Oxford: Oxford University Press). Barro, R. and D. Gordon (1983) `A positive theory of monetary policy in a natural rate model', Journal of Political Economy, vol. 91, pp. 589±610. Coles, M. and A. Philippopoulos (1997) `Are exchange rate bands better than ®xed exchange rates? The imported credibility approach', Journal of International Economics, vol. 43, pp. 133±53. Cukierman, A. (1994) `Central bank independence and monetary control', Economic Journal, vol. 104, pp. 1437±48. Dornbusch, R. (1991) `Problems of European monetary integration', in A. Giovannini and C. Mayer (eds), European Financial Integration (Cambridge: Cambridge University Press and CEPR). Giavazzi F. and A. Giovannini (1987) `Models of the EMS: Is Europe a greater DM area?', in R. Bryant and R. Portes (eds), Global Macroeconomics: Policy Con¯ict and Cooperation, (London: Macmillan ± now Palgrave). Giavazzi, F. and M. Pagano (1988) `The advantage of tying one's hands: EMS discipline and central bank independence', European Economic Review, vol. 32, pp. 1055±75. Hamilton, J. (1994) Time Series Analysis (Princeton NJ: Princeton University Press). Horn, H. and T. Persson (1988) `Exchange rate policy, wage formation and credibility', European Economic Review, vol. 32, pp. 1621±36. Kaminsky, G. (1993) `Is there a peso problem? Evidence from the dollar/pound exchange rate, 1976±1987', American Economic Review, vol. 83, pp. 450±72. King, R., C. Plosser, J. Stock and M. Watson (1991) `Stochastic trends and economic ¯uctuations', American Economic Review, vol. 81, pp. 891±940.

102 Exchange-Rate Regimes, In¯ation and Credibility McKinnon, R. (1993) `International money in historical perspective', Journal of Economic Literature, vol. XXXI, pp. 1±44. Rogoff, K. (1985a) `The optimal degree of commitment to intermediate monetary target', Quarterly Journal of Economics, vol. 100, pp. 1169±90. Rogoff, K. (1985b) `Can international monetary policy cooperation be counterproductive?', Journal of International Economics, vol.18, pp. 199±217. Svensson, L. (1997) `Optimal in¯ation targets, ``conservative'' central banks, and linear in¯ation contracts', American Economic Review, vol. 87, pp. 98±114.

Part 2 Microeconomic Issues of Economic Integration

This page intentionally left blank

7

The Effects of European Integration on Greek Manufacture Yannis Katsoulakos and Nicholas Tsounis

7.1

Introduction

In 1985 the European Commission, with its white paper Completing the Internal Market, set the timetable for unifying the European market. The Single Market Programme (SMP) proposed that all member states abolish all barriers to trade and factor ¯ows in order to increase economic growth and convergence. A single market would be created by 1992, permitting the free circulation of goods, services, persons and capital. The SMP listed almost 300 measures to be implemented in order to establish a single market. The programme was rati®ed in January 1987 and became the Single European Act, with a mandate to abolish all national barriers to trade and factor ¯ows. The Single European Market (SEM) came into effect in January 1993. The policy changes required for the implementation of the SMP were as follows:

. Trade liberalisation of goods and services: ± ± ±

removal of customs formalities; liberalisation of public procurement; establishing common technical standards in production and distribution. . Liberalisation of factor ¯ows: ± unconstrained labour mobility; ± opening up the capital markets; ± mutual recognition of approvals by national regulatory agencies. The SMP has raised many theoretical and empirical (policy-related) questions. An important set of questions relates to the impact of the SEM on trade and factor ¯ows from and into the countries on the periphery of

105

106 Effects of European Integration on Greek Manufacture

the European Union (EU). The free movement of ®nal products and factors of production was expected to have positive static and dynamic effects on economic welfare and to provide a stimulus to economic growth, especially with regard to the less developed EU countries. Economic theory suggests that we can categorise the effects of the SEM in terms of allocation, accumulation and location effects (see Baldwin and Venables, 1995). There could be also an important effect on technological change and thus growth, as described in particular by Grossman and Helpman (1992). Various authors have estimated the effects of the SMP on the structure of trade and production for the large EU countries and the EU as a whole (a short list includes Amiti, 1996; Buigues and Sheehy, 1995; Italianer, 1994; Jacquemin, 1990; Jacquemin and Sapir, 1988; Neven and Roller, 1991; Pelkmans, 1993; Sapir, 1990, 1996), but little has been said in the literature about the effects of the SMP on the structure of trade and production on the less developed countries on the periphery of the EU. The purpose of this chapter is to examine the extent of the structural changes that have occurred in the Greek economy since the launch of the SMP. In Section 7.2 the theoretical predictions of the impact of the SMP are surveyed. In Section 7.3 the methodology used to estimate the SMP effects on Greek manufacture is presented, while Section 7.4 examines the changes that have occurred in the structure of specialisation, and analyses whether Greece has changed its sources of supply from domestic products to imports from the EU and non-EU countries. Section 7.5 concludes.

7.2

Theoretical predictions of the impact of the SMP

As already noted, the impact of economic integration can be analysed in terms of allocation, location and accumulation effects. In addition, as has been shown by Grossman and Helpman (1992), economic integration could have a powerful effect on technical change and innovation. This section discusses these four effects. 7.2.1

Allocation effects

These (static) effects refer to an increase in real income (welfare) due to the more ef®cient allocation of factors of production that results from reducing trade barriers and eliminating price distortions in production and consumption (de la Fuente, 1995). However, the distribution of welfare gains may be uneven, with some larger countries experiencing net welfare losses (the optimal tariff argument).

Yannis Katsoulakos and Nicholas Tsounis 107

Another possible source of welfare loss after the removal of trade barriers is trade diversion, as opposed to trade creation. Trade diversion occurs when there is a switch in trade from ef®cient external suppliers to less ef®cient internal suppliers. The original literature on trade diversion versus trade creation was developed by Lipsey (1960) Meade (1955), Viner (1950), and Michaelly (1963) and was surveyed by Krauss (1972). However, Kemp and Wan (1976) and Mundell (1964) developed theoretical frameworks that show how member countries can bene®t from economic union even if there is trade diversion. Furthermore, empirical studies have reinforced the view that economic integration creates rather than diverts trade. The above assume perfectly competitive markets. The allocation effects on income and welfare would probably be larger if we allowed for economies of scale (increasing returns) and imperfect competition. (See, Krugman and Venables, 1994, for an introduction to this literature.) The theoretical literature has not, however, reached unambiguous conclusions. The welfare impact on individual countries is dif®cult to evaluate ex ante and its magnitude depends very much on the assumptions made by the speci®c model about the relevant importance of supply and demand elasticities, economies of scale, market size, the concentration of industry and other trade distortions. Nevertheless, empirical models show that welfare gains are larger under imperfect competition and economies of scale than they are under perfect competition (de la Fuente, 1995). Another source of ef®ciency gains is the reduction of internal organisational slack, termed X-inef®ciency by Leibenstein (1966). Economic integration increases competition, forcing a better (more ef®cient) allocation of ®rms' managerial resources. The allocation effects discussed above are static in the sense that they do not take into consideration the impact of SMP measures on factor accumulation. 7.2.2

Accumulation effects

These (dynamic) effects of the SMP can be analysed in terms of factor accumulation through changes in savings and investment rates on the one hand and technological progress on the other. Both affect output growth. The static effects (allocation effects) discussed above concern the SMP's impact on output, assuming there is no change in the stock of factors of production. The ultimate effect of the (static) increase in output due to SMP will be an increase in savings, thus raising the steady-state capital to labour ratio. This is the so-called induced capital formation argument, or

108 Effects of European Integration on Greek Manufacture

the long-term impact of economic integration on growth and output (Baldwin, 1989, 1992; de la Fuente, 1995, pp. 39±43). This analysis of the effect of trade liberalisation measures on factor accumulation and growth can be expanded to allow for permanent productivity-enhancing factor-accumulation effects (endogenous growth). The new growth literature initiated by Romer (1986) has made the accumulation of factors of production a ceaseless endogenous process of the economic system. This literature emphasises the micro foundations of factor accumulation, specifying the private costs and gains in new investment in human capital, and technical progress. The conclusion reached by this `new' literature is that continuous output growth can be achieved by sustained productivity growth, generated, for example, by a continuous process of R&D investment and innovations. 7.2.3

Location effects

The elimination of trade barriers will affect the geographical concentration of economic (industrial) activity, as stressed in particular in recent work by Krugman (1991a, 1991b) and Krugman and Venables (1990, 1993, 1994). Two factors emphasised in this new literature are: (1) increasing returns to scale in production that are internal to the ®rm, and (2) trade costs such as transport costs, marketing costs and communication costs due to language and/or cultural differences. The distribution of economic activity across regions cannot be a priori determined. It is true that for those industries that experience increasing returns (due to large ®xed costs), the elimination of trade barriers makes it pro®table to concentrate production in speci®c regions. On the other hand, if the economies of scale are not large enough relative to regional demand and the trade costs are high, then economic activity may spread in many regions. This effect, which may be called the location effect of the SMP, may be reinforced by wage differentials that are due to labour immobility. Finally, as already suggested, the SMP is expected to affect the rates of return on savings and factors of production. In particular, returns on investment and the rate of capital accumulation are expected to be positively affected. These effects will be reinforced by technological progress through an increase in investment and diffusion. 7.2.4

Technological progress effects

Economic integration has a technological progress effect through its impact on the accumulation of technological knowledge. Grossman and

Yannis Katsoulakos and Nicholas Tsounis 109

Helpman (1992) have identi®ed four mechanisms by which economic integration might affect the accumulation of technological knowledge. First, economic integration will facilitate the communication of technical information. Second, competition ± which is the expected result of economic integration ± forces private agents to implement new ideas and technologies. Third, economic integration increases the size of the market, creating more pro®t opportunities. This can have a positive effect on the innovation process even though increasing competition may have a negative effect on innovation. Fourth, innovation may be encouraged through the specialisation fostered by economic integration. Baldwin (1992) has developed a theoretical argument that links the accumulation of human capital (knowledge) to the removal of trade barriers and economic growth. Nevertheless, it is quite possible that the SEM might negatively affect the incentive to invest in technological innovations and human capital accumulation, and this may be particularly so for the relatively less-developed countries. Grossman and Helpman (1992) give four reasons why this might be the case. First, increased trade implies increased competition and national ®rms might ®nd that this reduces the anticipated pro®tability of their investment in knowledge. Second, opening up trade with a technologically advanced country may force the less advanced country to reduce its investment in innovation. This might lead to a concentration of technological progress in a few regions that had an advantage in innovation production before economic integration. Third, countries with an unskilled (manual) labour endowment may be forced by economic integration to specialise in commodities that are low in technological content. Fourth, countries that invested relatively more in human capital before economic integration will experience a higher reward after economic integration, reducing the incentive to invest in research and development.

7.3 Methodology for the estimation of SMP effects on trade and specialisation There are two fundamental reasons why countries specialise and trade. First, countries differ in terms of resource endowments or technology, or both, and specialise in producing the goods and services in which they have a comparative advantage. Second, the existence of economies of scale (or increasing returns) make it advantageous for individual countries to specialise in the production of only a limited range of goods and services. This, combined with the fact that consumers `love variety' (or

110 Effects of European Integration on Greek Manufacture

have convex preferences), explains why countries specialise and trade in the products of the same industry (intra-industry trade). As noted in the previous section, economic integration affects both these sources of specialisation and trade, and this has important implications for a country's trade ¯ows. We shall distinguish between three fundamental aspects of the relationship between the effects of integration (discussed in the previous subsection) and the nature and extent of a country's trade ¯ows: (1) trade creation and diversion, (2) intraindustry trade (re¯ecting the relative signi®cance of scale effects) and (3) relative competitiveness/specialisation within the EU. 7.3.1

Trade creation and diversion

When examining the effects of integration on trade creation and trade diversion the shares of apparent consumption are usually used, de®ned as gross domestic production minus total exports plus total imports (from the EU and from non-EU countries) in the three sources of supply: domestic production for the domestic market, EU imports and non-EU imports (these indicators were ®rst used by Truman, 1969, p. 205, 1975, p. 5). Changes in the share of consumption in sources of supply can be related to different effects of economic integration. A decrease in the share of consumption in domestic production indicates trade creation, which can be internal and/or external if the share of EU imports and/or non-EU imports has increased. An increase in the share of consumption in domestic production indicates trade diversion (or trade erosion in Truman's terminology), which again may be internal and/or external if the share of EU and/or non-EU imports has decreased. According to Truman (1975) integration generates both internal and external trade creation in manufacturing sectors. A similar conclusion has been reached by Jacquemin and Sapir (1988) for the period 1973±84, Neven and Roller (1991) for the period 1975±85 and Tsounis (1992) for the period 1976±87 for the Greek economy. The merit of the shares of apparent consumption method is that the approach abstracts from the effects of the growth of the economy. However, the assumption is that growth has a neutral effect with respect to the three shares; that is, the elasticity of each share with respect to increases in income is equal to zero. Furthermore, the method shows trade creation: the substitutions that economic integration causes in production and consumption are studied jointly. The de®ciencies have to do with the usual problem about the period chosen for the construction of the anti-monde. Also, it has been argued that systematic shifts in the domestic share, caused by domestic demand pressure and structural

Yannis Katsoulakos and Nicholas Tsounis 111

change, apart from that caused by integration, tend to render unreliable the conclusions about trade creation (Verdoorn and Van Bochove, 1972, p. 346). To overcome this problem, in the present study the stability of domestic demand pressure before and after the SMP period has been examined and is described in Section 7.4. It should be noted, however: that the logical validity of connection between welfare gains and the ®nding for trade creation and diversion . . . hold up only under a long list of unreasonable assumptions: only tariff cuts . . . or other DRC (domestic rent creating) barriers are considered, tariff rates are uniform across all products and all trading partners initially, and . . . scale, pure pro®t, variety and accumulation effects are unimportant. Subsequent research has showed that these effects are important, so such an approach presents a very limited interest from the point of view of aggregate welfare analysis. Despite these quali®cations, it may be worth undertaking this sort of study simply because the notions of trade creation and diversion are so ®rmly ®xed in the minds of many. Moreover, it is easy to do and even if welfare conclusions cannot be made, bilateral trade ¯ows are intermediate variables that are of interest for positive as well as normative reasons (Baldwin and Venables, 1995, pp. 3±5). An additional indirect bene®t of measuring the extent of trade creation and diversion is that it can provide useful information on the impact of integration on a country's relative competitiveness and specialisation. 7.3.2

Intra-industry trade

As hinted at above, a dominant distinction in modern trade theory is that between interindustry and intra-industry trade. Interindustry trade re¯ects comparative advantage. However, even if countries have the same overall factor proportions, their ®rms will continue to produce and trade differentiated products. It is economies of scale that prevent individual countries from producing the full range of products for themselves; thus, economies of scale coupled with convexity of preferences can be an independent source of international trade. The relative importance of interindustry and intra-industry trade depends on how similar countries are. If their factor proportions are similar, then there will be little interindustry trade, and intra-industry trade, based ultimately on economies of scale, will be dominant. On the other hand, if their factor proportions are very different, all trade will be based on comparative advantage (there may be little intra-industry trade based on

112 Effects of European Integration on Greek Manufacture

economies of scale). With regard to the empirical evidence, Greenaway and Milner (1983) note that: although there are considerable dif®culties associated with the computation of a meaningful summary statistic of the importance of intra-industry trade, relative to total trade, a substantial amount of evidence now exists to suggest that intra-industry trade has increased in importance to the point where it may account for over 60% of total trade in most developed market economies. Economic integration can be expected to have a powerful impact on intraindustry trade by positively affecting a number of factors that are associated with extensive intra-industry trade (IIT). We can distinguish between the short- to medium-term and medium- to long-term effects of integration on IIT. In the short to medium term, integration can be expected to increase for the following reasons. First, product differentiation: the greater the potential for product differentiation, the greater that IIT will tend to be. That is, by increasing ®rms' potential market size, integration increases the potential for product differentiation and this will tend to increase IIT. Second, scale economics: IIT will tend to be greater for commodities with scope for scale economies. Integration ± again by increasing market size ± can, all other things being equal, increase the scope for scale economies and hence IIT. Third, trade barriers (openness): empirical evidence suggests that IIT tends to be greater when trade barriers are low (see, Greenaway and Milner, 1984). This suggests a direct link between the elimination of trade barriers and the size of IIT. However, the effect may be due to longer-term considerations: openness may eventually result in similarities in tastes and/or the convergence of per capita income; both of these will lead to an increase in IIT. The above need to be quali®ed for individual countries by taking distance into account: IIT tends to be less as the average distance between trading partners increases. Thus, the factors outlined above are unlikely, all other things being equal, to have the same effect on IIT between Greece and the rest of the EU and on IIT between, say, Ireland and the rest of the EU. As regards the long-term effects of IIT, it is common to measure similarity according to `taste overlap' (see, for example, Greenaway and Milner, 1984) and/or similarity in per capita income, and tests reveal that IIT is positively correlated with these indicators. Integration is expected to increase `similarity' in these senses, so in the long term this too will have a positive effect on IIT.

Yannis Katsoulakos and Nicholas Tsounis 113

There are a variety of measures of intra-industry trade (see, Greenaway and Milner, 1983), perhaps the best known of which is that proposed by Grubel and Lloyd (1975), which can be written (without a time subscript) as: " # jXEj MfE j Bj ˆ 1 :100 …XEj ‡ MjE † where XEj = exports of industry j's products to the EU, MEj = imports of industry j's products to the EU, j = 1 . . . n, and, 0  Bj  100. The closer the index to its upper bound of 100 (occurring when XEj = MEj ) the greater the proportion of intra-industry trade in the total of sector j; whilst the closer it lies to its lower bound of zero (occurring when XEj or MEj equal zero) the greater the extent to which interindustry trade dominates. The principal problem with any index of intra-industry trade is the dif®culty of unambiguously de®ning the term `industry' for empirical purposes. The problem is not just one of de®ning the concept, but of actually operationalising it. Thus, even if one could arrive at some critical degree of similarity in input requirements that would serve to distinguish activities within an industry from activities outside that industry, one would still face the dif®culty of obtaining suitably classi®ed data for empirical analysis. Regrouping data from established classi®cations such as the Standard International Trade Classi®cation (SITC) or the UK Standard Industrial Classi®cation (SIC) is in principle feasible but in practice dif®cult. The response of most researchers is to adopt a particular digit of the classi®cation with which they are working and to assume that the categories classi®ed as industries can reasonably be expected to contain activities with similar input requirements. A great many investigators have accepted the third digit of the SITC as a suitable level of statistical dissaggregation to approximate the concept `industry'. There is, however, no a priori reason why the third digit of either of these classi®cations should be more suitable than, say, the fourth digit. Indeed, it is likely that for some activities the third digit will be the appropriate level whilst for others it may be the second, fourth or ®fth. Unless one regroups, which would probably necessitate breaking up the classi®cation, one cannot be entirely certain. Some evidence for both the SITC (Finger, 1975) and the UK SIC exists to suggest that the variability in input requirements within the third digit of each of these classi®cations could be greater than the variability between three-digit categories. If this is the case, then there are grounds for arguing that at least some recorded intra-industry trade is

114 Effects of European Integration on Greek Manufacture

simply the outcome of categorical aggregation; that is, the grouping together into statistical categories of activities from different industries (this problem is discussed in detail in Greenaway and Milner, 1983). The above remarks concern the measurement of IIT as practised in most of the empirical literature on testing the determinants of IIT. In this literature the indicator used is Bi ; that is, the share of IIT in each industry's gross trade. 7.3.3

Relative international competitiveness/specialisation

One of the most powerful anticipated effects of economic integration is ± through the scale, variety and other allocation and location effects identi®ed in Section 7.2 ± a change in the relative international competitiveness and specialisation of the countries concerned. This will occur as ®rms in different sectors adjust to more ef®cient capacities and to new optimal product lines. With regard to the location effects of integration, Krugman (1991a), in an examination of specialisation patterns in the United States in 1977 and the EU in 1985, found that the degree of economic specialisation in the United States was higher than that of the EU and ascribed this to the existence of trade barriers inside Europe, arguing that the SMP is likely to increase specialisation in the EU, particularly in sectors operating under increasing returns to scale. Furthermore, a priori increased specialisation due to the SMP is expected to result from the liberalisation of trade and the abolition of trade impediments, so that the full extent of comparative advantages and disadvantages will be revealed. Other factors affecting the pattern of a country's specialisation during the process of integration are the effect of integration on the strategic ability of ®rms to adjust to the increased competition that is implied by integration and to exploit the opportunities created by larger potential markets, and this will depend, among other things, on the initial conditions and on the sector in question. Sapir (1996, p. 460), using the Her®ndahl index to measure changes in the degree of specialisation due to the SMP in the four largest EU member countries, concludes that `in general, the 1992 programme does not appear to have generated a signi®cant degree of specialisation in the largest EC member states'. Likewise Amiti (1996), using indexes of relative specialisation (like the ones used in the present study), ®nds that specialisation in the four largest EU member states remained fairly constant after the introduction of the SMP. In the literature several indexes have been devised to measure relative competitive strength and specialisation, and changes to them as conditions change. However, the indicator that is probably best known

Yannis Katsoulakos and Nicholas Tsounis 115

and most often used to measure relative competitiveness is the Balassa or revealed comparative advantage (RCA) index (Balassa and Bauwens, 1988, p. 7), which is given for sector i in period t by: RCAit ˆ

…XEit MitE † :100 …XEit ‡ MitE †

It is clear that ±100 < RCA < 100 and that when exports grow faster than imports there will be a positive change in RCA, indicating an increase in relative comparative advantage. In the literature, RCA values such as 100 > RCA > 0 are considered broadly as indicating strong relative competitiveness, whilst values such as 0 > RCA > ±100 are considered as indicating weak relative competitiveness. Another index that has often been used to evaluate a country's export specialisation in a particular sector, relative to that of its EU partners, is the relative specialisation index (SI), given by: SI ˆ

…Xi =X† :100 …XiE =XE †

where Xi are the total exports of sector i, X the total exports of the country, XiE the total exports of sector i from EU countries and XE the total exports of EU countries. When SI > 100 this suggests a relatively high degree of specialisation in sector i and SI < 100 suggests the opposite. Of course, we are also interested in the dynamics of sectoral export specialisation; that is, in
(SI), the change in SI over time. On the basis of the RCA, the SI and Grubel±Lloyd indicator analysis, a number of issues will be addressed. First, how did the competitive position of each sector develop prior to and after the implementation of the SMP? Second, how did the specialisation position of each sector develop over the same period? Third, have there been signi®cant changes in sectoral competitiveness and/or specialisation? Fourth, are the traditional sectors the strongest performers? Fifth, are these sectors subjects to strong international demand (what is their future growth potential)? Sixth, has the SMP increased product differentiation in Greek manufacture due to an increase in market size and scale economies?

7.4

The effects of SMP on Greek manufacture

As described in Section 7.2, changes in consumption shares have been utilised to examine the trade effects of the internal market programme on

116 Effects of European Integration on Greek Manufacture

the Greek economy. The consumption shares of 155 three-digit NACE manufacturing sectors have been examined (a description of the sectors is provided in Appendix 7.1). The period examined is 1981±92, and a search for changes in the pattern of consumption shares has been made for the years after 1985. A direct application of the method is to compute the shares between a base pre-integration year and a representative post-integration year and examine any changes according to the analysis presented above. However, this approach requires the very strong assumption that, in the absence of integration, nothing would have happened to alter systematically the pattern of shares in the base year. Three factors that in¯uence the three shares have been taken into account: (1) the cyclical changes of the shares over time, (2) the general trend exhibited in the shares of some sectors, and (3) the level of domestic demand pressure.1 The effects of cyclical changes in the shares can be neutralised by taking the average value of the shares over a pre-integration period and a postintegration period, rather than single years. This type of averaging, however, has the disadvantage that it does not take into account the general trends of the shares before and after integration. If, for example, before integration the domestic share had a tendency to increase at the expense of the partners' and non-partners' shares, and in the postintegration period this pattern was reversed, then by taking averages over consecutive pre-integration and post-integration years one could ®nd that the shares remained unchanged. The second factor that might create errors in the conclusions drawn concerns the assumption that, in the absence of integration, the patterns of the three basic shares would have remained the same as those exhibited in the speci®ed base year. In the absence of integration the domestic shares of expenditure on apparent consumption might have declined, while the partners' and the non-members' shares might have increased. Thus, the assumption that there is no trend in the shares, if such a trend actually exists, would lead to an overestimation of the degree of trade creation. Finally, the `shares in apparent consumption' analysis usually assumes that the pressure exerted by domestic demand will not change over the period in question. If there is a change in the productive capacity of the economy or a change in domestic demand, there will be a change in the shares that cannot be attributed to integration. To test for the existence of trends, the statistical signi®cance of the average changes in the shares over the two periods is examined. If the

Yannis Katsoulakos and Nicholas Tsounis 117

difference between the average rates of change of the shares in the two periods is statistically different from zero, the hypothesis of no trend or a change in the trend will be adopted (it will be explained later how the two cases can be distinguished from one another). Regarding the third assumption, to test whether domestic demand pressure remained constant during the pre-integration and post-integration periods, for each sector the following procedure is followed. A change in domestic demand pressure can be caused by the supply side, the demand side or both, hence measures representing both sides should be included in the test. The ratio of total `apparent' consumption (C = Q ± X + M) over GNP is used to measure domestic demand pressure. Changes on the demand side should be re¯ected in `apparent' consumption and changes on the supply side are taken to be measured by GNP. To test for a structural change in domestic demand pressure, the domestic shares of each sector are regressed on the ratio of (C/GNP) for the periods 1981±85 and 1986±92 and a Chow test is performed to test for the structural stability of the coef®cients with the F-statistic. The F-statistic is calculated from the following formula.2 Fˆ

…RSSR RSSl RSS2 †=k …RSS1 ‡ RSS2 †=…n1 ‡ n2 2k†

where RSS1 is the residual sum of squares from the regression: DSit ˆ a1 ‡ b1 …Ct =GNPt †

…7:1†

for the period t = 1981±85 and RSS2 is the residual sum of squares from the estimation of Equation 7.1 for t = 1986±92. RSSR is the residual sum of squares from the estimation of Equation 7.1 for the period t = 1981±l92; k = 2, n1 = 5 and n2 = 7.3 It is concluded that structural change in domestic demand pressure has taken place in ®ve out of 155 sectors (sectors 2480, 3110, 3240, 3500 and 3640). For these sectors nothing can be said about the effects of the SMP by analysing the changing pattern of consumption shares because these are affected by other exogenous factors, and therefore they will be excluded from the analysis. As has been noted, to eliminate cyclical ¯uctuations in the shares, averages are taken over the pre-integration and post-integration periods and the statistical signi®cance of the two means are tested via the t-test.4 The application of the means method, as described above, will not produce correct results in cases where there are general trends in the movement of the shares of a sector. As already discussed, there are two cases in which trends may be exhibited: there might be a general trend

118 Effects of European Integration on Greek Manufacture

(increasing or decreasing) in the shares in both the pre-integration and the post-integration period; and there might be a trend (increasing or decreasing) in the shares in the pre-integration period and a reverse trend in the post-integration period. In both these cases, averaging the shares will not produce correct results: in the ®rst case it will show a structural break in the postintegration period that has not in fact happened, and in the second case it will not show a structural break that has taken place, since the averages of the two periods could be almost equal. To overcome the problem of trends in the shares, the average rate of change in the shares over the two periods is calculated as well, and the t-statistic is used to test for the statistical signi®cance of the two means from the two periods. If the ®rst case applies to a sector, then it is expected that a structural break has happened only if the difference in the average rates of change in the shares in the two periods is statistically different from zero ± the average rates of change will have the same sign. If the second case applies to a sector, then it is expected that although the means of the shares in the two periods might be the same (their difference is not statistically signi®cantly different from zero), the signs of the average rate of change of the shares in the two periods will be different. Finally, in the case where no general trend is exhibited in a sector and a structural break has happened after integration, it is expected that the difference in the means of the shares during the pre-integration and postintegration periods will be statistically different from zero and the difference in the average rate of change in the shares over the two periods will not be signi®cantly different from zero. However, this case can be mistaken for the ®rst case, where no structural break has taken place and the difference in the means is due to a general trend in the shares. Therefore, for both cases one must examine the shares of that sector over the whole pre-integration and post-integration period. Table 7.1 summarises the pattern of change in the three shares. A minus sign indicates a decrease in the share, a plus sign an increase and a zero a constant share over the pre-integration and the post-integration periods. Sectors 4100 (food, drink and tobacco), 4130 (dairy products) and 4940 (toys and sporting goods) fall into the second category described above ± there was a tendency for the domestic shares to increase in the preintegration period but this was reversed in the post-integration period. Taking the averages of the shares in these periods would not be truly representative of the effects of integration because the reversed tendency would cause the means of the shares in the two periods to be the same

Table 7.1 Patterns of change of the shares of expenditure in apparent consumption (sign of share change) Domestic

EU

Non-EU

NACE

Domestic

1100

+

0

+

3500

1200 1300 1400 1500 1510 1520 1600 1610

0 ± ± 0 0 0 0 0

0 + + + 0 + 0 0

0 + + + 0 + 0 0

3510 3520 3530 3600 3610 3620 3630 3640

± 0 ± ± ± + 0

1620 1630 1700 2100 2110 2120 2200 2210 2220 2230 2240 2300 2310

0 0 0 + 0 0 ± ± ± ± ± + 0

0 0 0 ± 0 0 + + + + + 0 0

0 0 0 + 0 0 + + + + + + 0

3650 3700 3710 3720 3730 3740 4100 4110 4120 4130 4140 4150 4160

0 ± + 0 0 ± ± 0 + + ± + +

EU Change in domestic demand pressure + + + ± ± 0 0 Change in domestic demand pressure 0 ± 0 0 0 + + + ± ± + 0 ±

Non-EU

0 + + + + ± 0

0 + + + 0 + + ± ± + + 0 ±

119

NACE

Table 7.1 Patterns of change of the shares of expenditure in apparent consumption (sign of share change) (continued ) Domestic

EU

Non-EU

NACE

Domestic

EU

Non-EU

2320 2400 2410 2420 2430 2440 2450 2460 2470 2480

0 ± 0 ± ± 0 ± ± ±

+ + 0 + + 0 + + +

4170 4180 4190 4200 4210 4220 4230 4240 4250

0 0 ± ± ± 0 0 ± ±

0 + + + + + + 0 +

0 ± + + + ± + + 0

2500 2510 2550 2560 2570 2580 2590 2600 2601 3100 3110

± ± 0 + ± ± ± ± ± ±

+ + + ± ± + + + + +

4270 4280 4290 4300 4360 4380 4390 4400 4410 4420 4500

± ± ± ± 0 ± ± ± ± ± 0

+ + + 0 0 + + + + 0 ±

+ + + + 0 + + + + + +

3120

0

+ + 0 + + 0 + + + Change in domestic demand pressure + + + ± + + + + + + Change in domestic demand pressure 0

0

4510 4530

± ±

+ +

+ +

120

NACE

Table 7.1 Patterns of change of the shares of expenditure in apparent consumption (sign of share change) (continued ) Domestic

EU

Non-EU

NACE

Domestic

EU

Non-EU

3130 3140 3150 3160 3166 3200 3210 3220 3230 3240

± 0 0 ± 0 ± 0 ± 0

+ + + + 0 + + + 0

4550 4560 4600 4610 4620 4630 4640 4650 4660

0 0 ± ± ± ± ± 0 0

0 0 + + + + + + 0

0 0 + + + + + + 0

3250 3260 3270 3280 3300 3400 3410 3420 3430 3440 3450 3460 3470

± ± ± ± 0 ± + ± ± + 0 0 0

+ + + + 0 + + + 0 Change in domestic demand pressure + + + 0 0 + 0 + + ± + 0 0

± + ± + 0 + ± + + ± + + 0

4670 4700 4710 4720 4800 4810 4820 4830 4900 4910 4920 4930 4940 4950

0 ± ± ± ± 0 0 ± ± 0 0 0 ± 0

0 + + + + 0 0 + + 0 0 0 + 0

0 + + + + 0 0 + + 0 0 0 + 0

121

NACE

122 Effects of European Integration on Greek Manufacture

(their difference would not be signi®cantly different from zero) and the signs of the average rate of change in the shares in the two periods would be different. In these sectors, the last year of the pre-integration period (1985) is taken as the representative year for the construction of the anti-monde and is compared with the last available year of the post-integration period (1992). Trade creation occurred in 67 out of 155 sectors (a decrease in the domestic share) and trade diversion in 12 (an increase in the domestic share), but no statistically signi®cant change in the domestic share over the pre- and post-SMP period can be observed in the remaining 76 sectors. In six sectors, internal trade creation and external trade diversion occurred. This is what the traditional theory of customs unions predicts will happen after the formation of a customs union. Pharmaceutical products (2570), machinery for the iron and steel industries (3250), equipment for use in special branches (3270), oil and fat products (4110), starch and starch products (4180) and animal and poultry foods (4220) belong to that category. In most sectors (66), double trade creation (internal and external) occurred. This is in accordance with the ®ndings of the empirical studies mentioned in Section 7.2, which reinforces the view that economic integration creates rather than diverts trade. In four sectors ± other means of transport (3600), shipbuilding (3610), instrument engineering (3700) and footwear and clothing (4500) ± there was external trade creation and internal trade diversion. Of the 12 cases of trade diversion, ®ve ± `extraction of solid fuels (1100), extraction and preparation of metal ores (2100), non-metallic minerals, peat (2300), measuring and precision instruments (3710) and manufactured dairy products (4130) ± represented an increase in the share of nonEU countries at the expense of the EU countries. The remaining seven sectors where trade diversion occurred experienced a decrease in the share of both the EU countries and non-EU countries; chemical products for industry and agriculture (2560), insulated wires and cables (3410), telecommunication equipment (3440), railway rails and rolling stock (3620), the slaughtering and preparation of meat (4120), the processing and preserving of ®sh (4150) and grain milling (4160). There was no instance of trade diversion where the EU countries' share increased and non-EU countries' share decreased. To determine the effects of the SMP on the competitive position of each sector and the development of its specialisation position over the pre- and post-SMP period, the RCA and specialisation indexes are calculated. In

Yannis Katsoulakos and Nicholas Tsounis 123

addition, the Grubel±Lloyd indexes are used to examine whether the SMP led to an increased IIT in Greek manufacture due to an increase in market size and scale economies (see also Section 7.3). Table 7.2 summarises the results of the change of the RCA, specialisation and Grubel±Lloyd indexes.5 A plus indicates a positive change in the averages over the pre-SMP and post-SMP period, while a minus indicates a decrease in the averages. The data used for the calculations was extracted from the VISA database. The description of NACE sectors is reported in Appendix 7.1. The calculation of RCA indexes reveals that comparative advantage was strengthened over the second period in only 15 sectors: extraction of petroleum and natural gas (1300), extraction and preparation of metal ores (2110), iron and steel industry (2210), drawing and cold rolling of steel (2230), non-metallic mineral products (2400), cement, lime and plaster (2420), concrete, cement and plaster structures (2430), working of stones and non-metallic products (2450), grindstones and abrasive products (2460), forging, pressing and stamping (3120), insulated wires and cables (3410), furs and fur goods (4560), sawing and processing of wood (4610), other manufacturing industries (4900) and miscellaneous manufacturing industries (4950). Of importance is the increase in the RCA index for the non-metallic mineral products sector (sectors 2400, 2420, 2430, 2450 and 2460), which is one of the three most important sectors in the Greek economy in terms of output and employment (the others being food, drink and tobacco, and textiles). A weakening of Greece's competitive position relative to the other EU countries was experienced by 17 sectors: chemical products for industry and agriculture (2560), spaghetti and macaroni (4170), sugar manufacturing and re®ning (4200), ethyl alcohol and spirit distilling (4240), wine and cider (4250), textiles (4300), carpets, linoleum and ¯oor covering (4380), miscellaneous textiles (4390), leather products and substitutes (4420), footwear (4510), clothing (4530), household textiles (4550), timber and wooden furniture (4600), semi-®nished wood products (4620), wooden containers (4640), other wood manufactures, excluding furniture (4650) and jewellery (4910). For the remaining 123 sectors with Greek manufacturing involvement, no statistically signi®cant change in the RCA indexes is observed for the pre- and post-SMP period. This suggests that the SMP has had a minimal effect on the competitive position of the majority of Greek manufactures. To examine the development of specialisation in Greek manufacturing during the pre- and post-SMP period, relative specialisation indexes are used. It is found that 16 sectors became more specialised within the EU

NACE

RCA index

Specialisation index

Grubel±Lloyd index

NACE

RCA index

Specialisation index

Grubel±Lloyd index

1100 1200 1300 1400 1500 1510 1520 1600 1610 1620 1630 1700 2100 2110 2120 2200 2210 2220 2230 2240 2300 2310 2320 2400 2410

0 0 + 0 0 0 0 0 0 0 0 0 0 + 0 0 + 0 + 0 0 0 0 + 0

0 0 + + 0 0 0 0 0 0 0 0 ± + ± 0 + 0 + 0 ± 0 0 + 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 + 0 + 0 0 0 0 + 0

3500 3510 3520 3530 3600 3610 3620 3630 3640 3650 3700 3710 3720 3730 3740 4100 4110 4120 4130 4140 4150 4160 4170 4180 4190

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ± 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ± 0 ± ± + 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 + + 0 0 0 + 0

124

Table 7.2 Changes in the RCA, Grubel±Lloyd and specialisation indexes over the two periods6

Table 7.2

Changes in the RCA, Grubel±Lloyd and specialisation indexes over the two periods6 (continued) RCA index

Specialisation index

Grubel±Lloyd index

NACE

RCA index

Specialisation index

Grubel±Lloyd index

2420 2430 2440 2450 2460 2470 2480 2500 2510 2550 2560 2570 2580 2590 2600 2601 3100 3110 3120 3130 3140 3150 3160 3166 3200

+ + 0 + + 0 0 0 0 0 ± 0 0 0 0 0 0 0 + 0 0 0 0 0 0

+ 0 0 + + 0 0 0 0 0 ± 0 0 0 0 0 0 0 + 0 0 0 0 0 0

0 + 0 ± ± 0 0 0 0 0 ± 0 0 0 0 0 0 0 + 0 0 0 0 0 0

4200 4210 4220 4230 4240 4250 4270 4280 4290 4300 4360 4380 4390 4400 4410 4420 4500 4510 4530 4550 4560 4600 4610 4620 4630

± 0 0 0 ± ± 0 0 0 ± 0 ± ± 0 0 ± 0 ± ± ± + ± + ± 0

0 0 0 0 ± 0 0 0 0 0 0 ± 0 0 0 0 0 ± 0 ± + 0 0 0 0

0 0 0 0 ± ± 0 0 0 + 0 ± 0 0 0 ± 0 0 ± ± + ± + 0 0

125

NACE

126

Table 7.2 Changes in the RCA, Grubel±Lloyd and specialisation indexes over the two periods6 (continued) NACE

RCA index

Specialisation index

Grubel±Lloyd index

NACE

RCA index

Specialisation index

Grubel±Lloyd index

3210 3220 3230 3240 3250 3260 3270 3280 3300 3400 3410 3420 3430 3440 3450 3460 3470

0 0 0 0 0 0 0 0 0 0 + 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 + 0 0 0 0 0 0

0 + 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

4640 4650 4660 4670 4700 4710 4720 4800 4810 4820 4830 4900 4910 4920 4930 4940 4950

± ± 0 0 0 0 0 0 0 0 0 + ± 0 0 0 +

0 ± 0 0 0 0 0 0 0 0 0 + 0 0 0 + +

0 ± 0 0 0 + 0 0 0 0 0 + ± 0 0 0 +

Yannis Katsoulakos and Nicholas Tsounis 127

after the SMP: extraction of petroleum and natural gas (1300), mineral oil re®ning (1400), extraction and preparation of metal ores (2110), iron and steel industry (2210), drawing and cold rolling of steel (2230), nonmetallic mineral products (2400), cement, lime and plaster (2420), working of stones and non-metallic products (2450), grindstones and abrasive products (2460), forging, pressing and stamping (3120), insulated wires and cables (3410), starch and starch products (4180), furs and fur goods (4560), other manufacturing industries (4900), toys and sporting goods (4940) and miscellaneous manufacturing industries (4950). In 12 sectors Greece's relative export share decreased in the post-SMP period relative to the pre-SMP period: extraction and preparation of metal ores (2100), extraction and preparation of non-ferrous metal ores (2120), non-metallic minerals and peat (2300), chemical products for industry and agriculture (2560), processing of fruit and vegetables (4140), grain milling (4160), spaghetti and macaroni (4170), ethyl alcohol and spirit distilling (4240), carpets, linoleum and ¯oor coverings (4380), footwear (4510), household textiles (4550) and other wood manufactures, excluding furniture (4650). However, in the remaining 127 sectors no statistically signi®cant change in the specialisation indexes is observed for the pre- and post-SMP period. This result conforms with that derived from the examination of the RCA indexes and suggests that the effect of the SMP on the specialisation position of the majority of Greek manufactures has been minimal. To provide an overall picture of the pattern of change in the specialisation of sectors before and after the SMP, the change in the dispersion of the indexes of relative specialisation in Greek manufactures over the pre- and post-SMP period is examined. It is found that the dispersion of the value of the relative specialisation indexes increased. As a measure of dispersion, the standard deviation of the averages of the indexes in the two periods is calculated. For the 1981±85 period the standard deviation is 350.4, while for the 1986±92 period it is 482.6 (all sectors were included in the calculation). This means that Greece has become marginally more specialised since the introduction of the SMP. This partly con®rms Krugman's assertion that the 1992 programme was likely to increase specialisation in the EU, particularly in sectors operating under increasing returns to scale. In addition, specialisation may have increased because economic integration has revealed the full extent of comparative advantage (or disadvantage). Sapir (1996) argues that there are several possible explanations of why the SMP has not resulted in major intersectoral shifts in the patterns of specialisation of the four largest EU member countries: (1) the markets

128 Effects of European Integration on Greek Manufacture

were already largely liberalised prior to 1986 or there were delays in the implementation of the 1992 programme; (2) the liberalisation of manufactured goods markets may not have produced inter-industry resource shifts due to rigidities elsewhere in the economy; and (3) many of the effects of increased integration may have taken place within rather than across industries. As noted earlier, the SMP is expected to increase IIT. This was the case for 15 sectors of the Greek economy: iron and steel industry (2210), drawing and cold rolling of steel (2230), non-metallic mineral products (2400), concrete, cement and plaster structures (2430), forging, pressing and stamping (3120), machine tools for metalwork (3220), manufactured dairy products (4130), processed fruits and vegetables (4140), starch and starch products (4180), textiles (4300), furs and fur goods (4560), sawing and processing of wood (4610), manufacture of pulp, paper and board (4710), other manufacturing industries (4900) and miscellaneous manufacturing industries (4950). However, a decline in intra-industry trade occurred in an approximately equal number of sectors; that is, in 12 sectors (if the residual sectors 4900 and 4950 are excluded from the sectors where an increase has occurred, then the IIT increased in 13 sectors). These sectors are stone and non-metallic products (2450), grindstones and abrasive products (2460), chemical products for industry and agriculture (2560), ethyl alcohol and spirit distilling (4240), wine and cider (4250), carpets, linoleum and ¯oor coverings (4380), leather products and substitutes (4420), clothing (4530), household textiles (4550), timber and wooden furniture (4600), other wood manufactures, excluding furniture (4650) and jewellery (4910). For the remaining 128 sectors with Greek manufacturing involvement, no statistically signi®cant change in the Grubel±Lloyd indexes can be observed for the pre- and post-SMP period. This suggests that the SMP has had a minimal effect on intra-industry trade in Greek manufactures and the effect is con®ned to a limited number of sectors. Other investigators have found that the intensity of intra-industry trade in Greece is the lowest among the member states and is limited to Greece's traditional export products (Mardas, 1990). This phenomenon is explained by the poor ability of the Greek economy to create dynamic comparative advantages that are closely related to investor choice, specialisation pro®le and the complementarity of Greek trade within the EU. In the present study, about half of the manufacturing categories that experienced an increase in IIT belong to the food or textile sectors, which are also important sectors for the Greek economy. The result can be considered as `normal' since the products of these sectors are easy to

Yannis Katsoulakos and Nicholas Tsounis 129

differentiate, so the SMP could be expected to have a positive impact due to the increase in market size and greater economies of scale.

7.5

Conclusions

This chapter has analysed the change in the structure of trade and specialisation in Greece since the implementation of the SMP. The period examined was from 1981 (the year in which Greece became a member of the EU) to 1992 (the year of completion of the SMP). The data used were three-digit NACE. Several ®ndings emerged from the study. First, it was found that in most sectors double trade creation (internal and external) occurred: the consumption share of both intra- and extraEU imports increased. This is in accordance with the ®ndings of the empirical studies mentioned in Section 7.2, which reinforces the view that economic integration creates rather than diverts trade. Second, the results derived from the examination of the competitive position of Greek manufacturing sectors using the RCA index suggest that the effect of the SMP on the competitive position of the majority of Greek manufactures was minimal. Third, the effect of the SMP on intra-industry trade in Greece was minimal and was con®ned to a limited number of sectors. About half of the manufacturers that experienced an increase in IIT belonged to the food or textile sectors, which are important sectors for the Greek economy. Finally, it was found that during the SMP period Greece became marginally more specialised relative to the period before the SMP. This partly con®rms Krugman's assertion that the 1992 programme was likely to increase specialisation in the EU, particularly in sectors operating under increasing returns to scale. In addition, specialisation may have increased because the removal of export subsidies and other trade impediments revealed the full extent of comparative advantage (disadvantage).

130 Effects of European Integration on Greek Manufacture

Appendix 7.1

Description of sectors

NACE Description

NACE

Description

1100 1200 1300 1400 1500 1510 1520 1600 1610 1620 1630 1700 2100 2110 2120 2200 2210 2220 2230 2240 2300 2310 2320 2400 2410 2420 2430 2436 2440 2450 2460 2470 2480 2500 2510 2550 2560 2570 2580 2590 2591 2592 2600 2601 3100 3110 3120 3130 3140 3150 3160 3165 3166 3190 3200

3600 3610 3620 3630 3640 3650 3700 3710 3720 3730 3740 4100 4110 4120 4130 4140 4150 4160 4170 4180 4190 4200 4210 4220 4230 4240 4250 4270 4280 4290 4300 4310 4320 4330 4340 4350 4360 4370 4380 4381 4390 4400 4410 4420 4500 4510 4530 4550 4560 4600 4610 4620 4630 4640 4650

Other means of transport Shipbuilding Manufacture of railway rolling-stock Manufacture of cycles, motor cycles Aerospace equipment manufacturing Other transport equipment Instrument engineering Measuring and precision instruments Medical and surgical equipment Optical instruments, photo equipment Manufacture of clocks and watches Food, drink, tobacco industry Manufacture of oils and fats Slaughtering, preparation of meat Manufacture of dairy products Processing of fruit and vegetables Processing and preserving of ®sh Grain milling Manufacture of spaghetti, macaroni Starch and starch products Bread and ¯our confectionery Sugar manufacture and re®ning Cocoa and sugar confectionery Animal and poultry foods Manufacture of other food products Ethyl alcohol, spirit distillation Wine, cider Brewing and malting Soft drinks, natural water Manufacture of tobacco products Textile industry Wool industry Cotton industry Silk industry Preparation of ¯ax, hemp, ramie Jute industry Knitting industry Textile ®nishing Carpets, linoleum, ¯oor covering Carpets of all types of ®bre Miscellaneous textile industry Leather and leather goods Tanning, dressing of leather Leather products, substitutes Footwear and clothing industries Footwear Manufacture of clothing Manufacture of household textiles Manufacture of furs and fur goods Timber and wooden furniture Sawing and processing of wood Semi-®nished wood products Carpentry and joinery components Manufacture of wooden containers Other wood manufacturing (excluding furniture)

Extraction of solid fuels Coke ovens Extraction of petroleum and gas Mineral oil re®ning Nuclear fuels industry Extraction nuclear materials Production nuclear materials Electricity, gas, steam, hot water Generation, distribution of electric power Gasworks, gas distribution Products distribution steam, hot water Water supply Extraction, preparation metallic ores Extraction, preparation of iron ore Extraction, preparation of non-ferrous metallic ores Metals Iron and steel industry (ECSC) Manufacture of steel tubes Drawing, cold rolling of steel NF-metals Non-metallic minerals, peat Extract of building materials Salt, natural peat Non-metallic mineral products Clay products for construction purposes Cement, lime, plaster Structure concrete, cement, plaster Manufacture ready-mixed concrete Manufacture of asbestos Work of stones and non-metallic products Grindstones, abrasive products Manufacture of glass and glassware Manufacture of ceramic goods Chemical industry Basic industries chemicals, petrochemicals Paint, varnish, printing ink Chemical products for industry, agriculture Pharmaceutical products Soap, detergents, perfume Manufacture of other chemical products Manufacture of photographic products Polishes, household chemical products Man-made ®bres industry Chemical and man-made ®bres Manufacture of metal articles Foundries Forcing, pressing, stamping Treatment of metal Structural metal products Boilers, reservoirs, tanks Tools and ®nished metal goods Domestic heating appliances Manufacture of metal furniture Other metal workshops Mechanical engineering

Yannis Katsoulakos and Nicholas Tsounis 131 3210 3220 3230 3240 3250 3260 3270 3280 3300 3400 3410 3420 3430 3440 3450 3460 3470 3480 3500 3510 3520 3530

Agricultural machinery and tractors Machine tools working metal Textile machines, sewing machines Machines for food and chemical industry Machines for iron and steel industries Transmission equipment Equipment for use in special branches Other machinery and equipment Of®ce and DP machinery Electrical engineering Insulated wires and cables Manufacture of electrical machinery Electrical apparatus for industrial use Manufacture of telecomm. equipment Manufacture of radios and TVs Domestic electrical appliances Manufacture of electrical lamps and others Assembly of electrical equipment Motor vehicles and parts Manufacture, assembly of motor vehicles Manufacture of bodies for motor vehicles Manufacture of parts for motor vehicles

4660 4670 4700 4710 4720 4730 4740 4800 4810 4820 4830 4900 4910 4920 4930 4940 4950 5000 5010 5020 5030 5040 5100

Articles of cork, straw brushes Manufacturing of wooden furniture Paper, printing and publishing Manufacturing of pulp, paper board Products of paper and board Printing and allied industries Publishing Products of rubber and plastic Manufacturing of rubber products Rubber products Processing of plastics Other manufacturing industries Manufacturing of jewellery Manufacturing of musical instruments Photography and cinema laboratories Toys and sporting goods Miscellaneous manufacturing industries Building and civil engineering Construction of buildings Civil engineering Installation of ®ttings and ®xtures Building completion work Building without specialists.

Appendix 7.2 F-statistic for the Chow test for structural stability of the demand pressure on domestic shares in apparent consumption NACE

F-stat

NACE

F-stat

NACE

F-stat

1100 1200 1300 1400 1500 1510 1520 1600 1610 1620 1630 1700 2100 2110 2120 2200 2210 2220 2230 2240 2300

1.428 ± 1.608 0.508 ± ± ± ± ± ± ± ± 0.498 ± ± 0.945 0.581 0.867 3.216 0.690 1.310

3140 3150 3160 3166 3200 3210 3220 3230 3240 3250 3260 3270 3280 3300 3400 3410 3420 3430 3440 3450 3460

1.565 2.604 2.268 ± 0.659 0.634 0.654 ± 32.772 * 1.340 8.676 2.641 1.263 ± 2.318 0.121 1.495 5.746 0.934 6.913 2.854

4180 4190 4200 4210 4220 4230 4240 4250 4270 4280 4290 4300 4360 4380 4390 4400 4410 4420 4500 4510 4530

0.963 8.553 0.883 0.234 0.795 1.047 4.326 2.418 3.329 2.666 2.132 2.125 ± 2.827 1.576 5.997 6.092 0.098 0.390 0.963 0.933

132 Effects of European Integration on Greek Manufacture 2310 2320 2400 2410 2420 2430 2440 2450 2460 2470 2480 2500 2510 2550 2560 2570 2580 2590 2600 2601 3100 3110 3120 3130

± 1.311 3.616 ± 2.689 0.763 ± 0.371 3.771 1.986 13.845 * 4.530 4.291 1.421 2.500 0.095 1.220 5.798 1.346 4.589 3.202 32.840 * ± 3.958

3470 3500 3510 3520 3530 3600 3610 3620 3630 3640 3650 3700 3710 3720 3730 3740 4100 4110 4120 4130 4140 4150 4160 4170

± 9.575* 3.571 3.229 4.712 1.204 0.877 0.015 ± 11.163* ± 0.195 0.907 1.014 ± 2.695 2.186 0.938 0.154 0.342 2.683 0.952 0.664 0.947

4550 4560 4600 4610 4620 4630 4640 4650 4660 4670 4700 4710 4720 4800 4810 4820 4830 4900 4910 4920 4930 4940 4950 ±

0.499 ± 4.564 4.356 4.788 3.425 0.441 9.184 0.638 1.685 7.509 2.660 7.199 5.183 ± ± 5.017 1.469 ± ± ± 0.739 ± ±

*Statistically signi®cant at the 1 per cent level.

Notes 1. For a discussion of the factors affecting the change of shares in apparent consumption, see Section 7.3.1. 2. The results of the calculation of the F-statistic are given in Appendix 7.2. 3. DS is domestic share, C is total apparent consumption and i denotes sector i. 4. The results of the values of the shares, their averages in the pre-SMP and postSMP periods, the t-statistic of the test for the statistical signi®cance of the averages, their average annual rate of change and the t-statistic of the test for the statistical signi®cance of the average change are available upon request from the authors. 5. The values of the RCA, specialisation and Grubel±Lloyd indexes, their averages in the pre-SMP and post-SMP periods, the t-statistic of the test for the statistical signi®cance of the averages, their average annual rate of change and the t-statistic of the test for the statistical signi®cance of the average change are available upon request from the authors. 6. A plus indicates a positive change in the averages over the pre-SMP and postSMP period, while a minus indicates a decrease in the averages. Only the statistically signi®cant changes are presented in the table. Zero indicates that the change in the averages of the indexes over the two periods was not statistically signi®cant at 5 per cent level.

Yannis Katsoulakos and Nicholas Tsounis 133

References Amiti, M. (1996) `Specialisation Patterns in Europe', mimeo (London: LSE and Universitat Pompeu). Balassa B. and L. Bauwens (1988) Changing Trade Patterns in Manufactured Goods: An Econometric Investigation (Amsterdam: North-Holland). Baldwin, R. (1989) `The Growth Effects of 1992', Economic Policy, vol. 9, pp. 247±82. Baldwin, R. (1992) `The Growth Effects of Import Competition', NBER Working Paper no. 4045 (Cambridge, MASS: NBER). Baldwin, R. and A. Venables (1995) `Methodologies for an Aggregate Ex-Post Evaluation of the Completion of the Internal Market', Study commissioned by CEC, DGII (Brussels: European Commission). Buigues, P. and J. Sheehy (1995) `The impact of the Internal Market Porgramme on European Integration', in E. EMS (eds), 35 Years of Free Trade in Europe (Geneva: EFTA). de la Fuente, A. (1995) `Assessing the Effects of the Internal Market Programme: A Conceptual Framework', Study commissioned by CEC, DGII (Brussels: European Commission). Finger, J. M. (1975) `Trade Overlap and Intra-Industry Trade', Economic Enquiry, vol. 13, pp. 581±89. Greenaway P. and C. Milner (1983) `On the Measurement of Intra-Industry Trade', Economic Journal, vol. 94, pp. 900±8. Greenaway P. and C. Milner (1984) `A Cross-Section Analysis of Intra-Industry Trade in the UK', European Economic Review, vol. 26, pp. 319±44. Greenaway P. and C. Milner (1994) `Adjustment and the Measurement of Marginal Intra-Industry Trade', Welfwirtschaftliches Archiv, vol. 130, pp. 418±27. Grossman, G and E. Helpman (1992) Innovation and Growth in the World Economy (Cambridge, MASS: MIT Press). Grubel, H. G. and P. J. Lloyd (1975) Intra-Industry Trade (London: Macmillan ± now Palgrave). Italianer, A. (1994) `Whither the Gains from European Economic Integration?' Review EÂconomique, vol. 45, pp. 689±702. Jacquemin, A. (1990) `Horizontal Concentration and European Merger Policy', European Economic Review, vol. 34, pp. 539±50. Jacquemin, A. and A. Sapir (1988) `European Integration or World Integration?', Welwirtschaftliches Archiv, vol. 124, pp. 127±39. Kemp, M. C. and H. Y. Wan (1976) `An Elementary Proposition Concerning the Formation of Customs Unions', Journal of International Economics, vol. 6, pp. 95±7. Krauss, M. B. (1972) `Recent Developments in Customs Union Theory: An Interpretive Survey', Journal of Economic Literature, vol. 10, pp. 413±36. Krugman, P. (1991a) `Increasing Returns and Economic Geography', Journal of Political Economy, vol. 99, pp. 483±99. Krugman, P. (1991b) Geography and Trade (Cambridge, MASS: MIT Press). Krugman, P. and A. Venables (1990) `Integration and the Competitiveness of Peripheral Industry', in C. Bliss and J. B. de Macedo (eds), Unity with Diversity in the European Community (Cambridge: Cambridge University Press). Krugman, P. and A. Venables (1993) `Integration, Specialisation and Adjustment', CEPR Discussion Paper no. 886 (London: CEPR). Krugman, P. and A. Venables (1994) `Globalisation and the Inequality of Nations', CEPR Discussion Paper no. 1015 (London: CEPR).

134 Effects of European Integration on Greek Manufacture Leibenstein, H. (1966) `Allocative Ef®ciency vs. X-ef®ciency', American Economic Review, vol. 56, pp. 392±415. Lipsey, R. G. (1960) `The Theory of Customs Unions: A General Survey', Economic Journal, vol. 7, pp. 496±513. Mardas, D. (1990) `Country Report for Greece', in P. Bigues, F. Ilzkovitz and F. Lebrum (eds), The Impact of the Internal Market by Industrial Sector: The Challenge for the Member States, Study commissioned by the DG for Economic and Financial Affairs and DG for Employment, Industrial Relations and Social Affairs (Brussels: European Commission). Meade, J. (1955) The Theory of Customs Unions (Amsterdam: North Holland). Michaelly, M. (1963) `On Customs Unions and the Gains from Trade', The Economic Journal, vol. 75, pp. 577±83. Mundell, R. (1964) `Tariff Preference and the Terms of Trade', The Manchester School, vol. 32. Neven, D. and H. Roller (1991) `European Integration and Trade Flows', European Economic Review, vol. 35, pp. 1295±309. Pelkmans, J. (1993) `The Signi®cance of EC-1992', CEPS Working Document 82 (Brussels Centre for Economic Policy Studies). Romer, P. (1986) `Increasing Returns and Long-Run Growth', Journal of Political Economy, vol. 94, pp. 1002±37. Sapir, A. (1990) `Does 1992 Come Before or after 1990? On Regional versus Multilateral Integration', in R. Jones and A. O. Krueger (eds), The Political Economy of International Trade ± Essays in Honor of Robert Baldwin (Oxford: Basil Blackwell). Sapir, A. (1992) `Regional Integration in Europe', Economic Journal, vol. 102, pp. 1491±506. Sapir, A. (1996) `The Effects of Europe's Internal Market Program on Production and Trade: A First Assessment', Weltwirtschaftliches Archiv, vol. 132, pp. 457±75. Truman, E. M. (1969) `The European Economic Community: Trade Creation and Trade Diversion', Yale Economic Essays, vol. 9, pp. 201±57. Truman, E. M. (1975) `The Effects of European Economic Integration on the Production and Trade of Manufactured Products', in B. Balassa (ed.), European Economic Integration (Amsterdam: North-Holland), pp. 3±39. Tsounis, N. K. (1992) `The Effects of European Economic Integration on the Greek Economy and the Pattern of Greek International Trade', PhD thesis, University of Manchester. Verdoorn, P. S. and C. A. Van Bochove (1972) `Measuring Integration Effects: A Survey', European Economic Review, vol. 3, pp. 337±49. Viner, J. (1950) The Customs Union Issues (New York: Carnegie Endowment for Peace).

8

Economic Integration in Europe and the Pattern of German Foreign Direct Investment Florence Hubert and Nigel Pain1

8.1

Introduction

Economic integration in Europe has been accompanied by a rise in foreign direct investment (FDI) within the EU economies. In virtually all member states, real in¯ows and out¯ows of direct investment after the mid-1980s reached their highest level since the initial formation of the Common Market, and the proportion of the aggregate stock of world FDI located within EU member states is estimated to have risen from 31 per cent in 1985 to 39 per cent by 1995 (UNCTAD, 1996). This re¯ects rising levels of investment in the EU by non-EU nationals and EU ®rms themselves. The surge in cross-border investment largely occurred in the aftermath of the Single European Act of 1986/7, during the gradual implementation of the measures required to complete the internal market. This chapter investigates the empirical relationship between the industrial pattern of manufacturing FDI by German companies and the measures taken in the internal market programme to establish whether the moves towards greater European integration have generated a structural change in the pattern of intra-EU investment. Worldwide, German ®rms have the fourth largest stock of overseas assets of all investing countries. Within Europe, they are the second most important investors after the United States. At the end of 1994 just under 1.2 million workers were employed in the foreign af®liates of German ®rms located in Europe. One striking feature of German FDI is the extent to which it became increasingly concentrated within Europe from the latter half of the 1980s.

135

136 Economic Integration and German FDI

However, comparatively little econometric research exists on the foreign investment decisions of German corporations, in spite of their large portfolio of overseas investments. There is, thus, little formal evidence as to whether the observed acceleration in intra-EU multinational investment by German ®rms has arisen in response to the measures introduced in the internal market programme, or whether it has simply been part of a broader trend towards production relocation, with Germany increasingly being viewed as an unattractive business location. A useful summary of this debate is provided by Jungnickel (1995). We use a panel data set for investments in four industries in ten separate locations to obtain explicit estimates of the impact that economic integration in Europe has had on cross-border investment. We also investigate whether there is any evidence that foreign investments by German ®rms has been diverted into EU member states at the expense of non-EU locations. In Sections 8.2 and 8.3 we review recent trends in German FDI and give an overview of the relevant literature on regional liberalisation and intraregional investment. This suggests a number of speci®c hypotheses that can be examined in any empirical analysis. Sections 8.4 and 8.5 describe the basic model used and discuss ways of ascertaining the impact of the internal market measures. Section 8.6 contains the main empirical results and provides a quantitative evaluation of the impact of the internal market programme.

8.2

The regional and industrial pattern of German FDI

The level of new investment by European ®rms within the EU has risen markedly since the mid-1980s. In contrast to the long-term historical trend, new intra-EU FDI investment has exceeded new investment by European ®rms in North America since 1988, reaching a peak of 1.25 per cent of GDP in 1990. Over the ®ve years to 1994, intra-EU FDI was equivalent to 4.5 per cent of gross domestic ®xed capital formation in the EU on average (Barrell and Pain, 1997, provide a more detailed overview of the growth of FDI in Europe). A simple descriptive analysis of the changing pattern of intra-EU ¯ows of direct investment cannot provide any direct quantitative indication of the possible in¯uence of the measures taken to ease product market barriers in the internal market programme. The observed aggregate pattern of investment by EU ®rms might, in part, simply be a temporary adjustment to the removal of national capital controls in many member states. This is one reason for focusing on large individual investors with detailed statistics on the location of foreign investment. UK outward

Florence Hubert and Nigel Pain 137

direct investment is considered in Pain (1997); here we utilise German data. The capital markets of both countries have been free of foreign exchange controls for some time. This chapter uses a panel data set to analyse the determinants of FDI by German manufacturing corporations from 1977 to 1994. The panel consists of six EU countries (or country blocs) plus a separate bloc consisting of Austria, Sweden and Switzerland, which were all outside the EU during the sample period. Outside Europe we focus on three additional regional groups ± North America (the United States and Canada), Latin America (Argentina, Brazil and Mexico) and South-East Asia (Japan, Singapore and India) ± re¯ecting the primary locations of German FDI. Around 90 per cent of German manufacturing FDI is sited in the locations we consider. Within the EU we distinguish France, Italy, the UK, Belgium, the Netherlands and Spain. We examine investments undertaken in four broad sectors: chemicals, mechanical engineering, electrical engineering and road vehicles. Together, the investments in these four sectors account for around three quarters of outward manufacturing direct investment from Germany. Table 8.1 illustrates how the location of manufacturing direct investment by German companies has evolved over time in our panel. There has been a marked rise in the share of investment in North America and, since the mid-1980s, in the EU, offset by a gradual fall in the proportion in Latin America. The long-term decline in the proportion of investments sited in other developing countries suggests that the relatively high level of outward investment from Germany is not simply a move towards lowwage locations. Within the EU, France has consistently attracted the largest proportion of manufacturing investment, although the share of investment in the British Isles (the UK and Northern Ireland), and to a lesser extent Iberia (Spain and Portugal), rose noticeably over the sample period. Investment within the EFTA members ± Austria, Sweden and Switzerland ± has also risen sharply since the late 1980s. In 1994 some 88 per cent of German manufacturing direct investment was located in the ten host locations we consider. This was a little lower than in 1988, re¯ecting the rapid growth in investment in Central and Eastern Europe since that time. The vast majority of manufacturing investment in the former Soviet-bloc economies has been in the road vehicles sector, with around 8.25 per cent of the stock of investments in that sector located in these countries by 1994. German chemical companies have followed a consistent path towards globalisation, with the distribution of investment in each of the regional locations remaining broadly in line with the size of the overall market. In

138 Economic Integration and German FDI Table 8.1 DM)

Geographical and industrial distribution of German FDI stocks (million

Industry

Host country

Chemicals

European Union EFTA North America Latin America South-East Asia European Union EFTA North America Latin America South-East Asia European Union EFTA North America Latin America South-East Asia European Union EFTA North America Latin America South-East Asia

Mechanical engineering

Electrical engineering

Road vehicles

1976

1982

1988

1994

3 357 538 3 117 859 151 676 366 283 578 42 1 404 661 293 772 205 841 59 138 1 920 55

5 694 600 7 882 2 095 575 1 196 446 1 489 1 260 116 1 929 884 2 824 1 337 257 1 609 310 1 419 2 773 128

10 308 767 13 530 3 069 1 585 2 249 749 2 358 1 031 260 4 222 1 780 4 071 1 648 326 3 580 798 2 062 4 511 213

15 489 1 677 19 231 3 894 2 389 4 030 1 597 4 603 1 904 379 7 430 2 734 4 945 2 762 841 9 951 1 887 3 104 6 687 362

Notes: EFTA: Austria, Sweden and Switzerland; North America: USA and Canada; Latin America: Argentina, Brazil and Mexico; South-East Asia: Japan, Singapore and India. The chemicals sector has the highest level of foreign investment out of the four we consider. Source: Deutsch Bundesbank, Kapitalver¯echtung mit dem Ausland, various issues.

contrast, the distribution of investment in the road vehicles and electrical engineering sectors is more skewed, with investment becoming increasingly concentrated within the EU. The road vehicles sector is also notable for the high level of investment in Latin America.

8.3

The impact of integration on FDI

It has long been recognised that changes in regional trade arrangements and technical standards can have an important effect on both the level and the location of overseas investment (Eaton et al., 1994). The initial formation of the European Community prompted considerable empirical study into the question of whether investment was being diverted into the region (Yannopoulos, 1990). More recently, European integration has involved the removal of non-tariff barriers to market entry within the EU

Florence Hubert and Nigel Pain 139

region. For EU investors it is likely that such barriers have been the main impediment to market entry elsewhere in Europe over the time period we study, given that the potential investor and host country are members of a common customs union. Survey evidence suggests that institutional barriers, particularly in public procurement, have been a signi®cant barrier to investment in Europe (Millington and Bayliss, 1991). A wide variety of measures to ease non-tariff barriers have been introduced under the internal market programme. These include steps to harmonise technical standards and regulations, the removal of customs barriers, and moves to open public procurement and remove constraints on capital markets. Progress towards the implementation of the single market is described in Hoeller and Louppe (1994) and European Commission (1996). Some of the non-tariff barriers, notably customs controls, would previously have restrained trade linkages but not market entry by means of direct investment. Others, such as technical requirements and lack of competition in public procurement, would have affected both exporters and (potential) foreign subsidiaries. This implies that the impact of the removal of non-tariff barriers might be expected to vary across sectors and across countries. Early studies of the likely impact of the internal market (IM) programme suggested that the measures should bring about a considerable degree of industrial restructuring. This was largely expected to come about through greater industrial specialisation, with ®rms able to produce in a single location, exploit any economies of scale arising from the existence of ®rmspeci®c ®xed costs and serve the wider European market through trade (Emerson et al., 1988). Such a pattern is consistent with the ®ndings of Brulhart and Torstensson (1996), who report a signi®cant correlation between the geographical concentration of particular industries in Europe and a prior ranking of the importance of scale economies in those industries. Little mention was made of intra-EU FDI in the initial studies of the internal market. The implication of the specialisation argument is that intra-EU FDI might ultimately have been lower than would otherwise be expected as a result of the IM programme (Itaki and Waterson, 1990), although there could have been an initial transition period with higher cross-border investment if the restructuring process required new investment, to ensure that all stages of production were sited in the lowest-cost locations. Labour-intensive assembly activities would be concentrated in sites on the periphery of Europe with relatively lower labour costs. Other, more capital-intensive manufacturing activities would be located closer to the industrial core of Western Europe.

140 Economic Integration and German FDI

For North America there is some evidence that such a restructuring process has occurred in some industries following deeper regional integration, with a number of US multinationals closing subsidiaries in Canada and substituting exports for FDI as a result of the improvement in market access arising from the US±Canadian Free Trade Agreement, which entered into force in 1989 (Niosi, 1994). Trade liberalisation also appears to have altered the economic geography of Mexico (Hanson, 1997), with Mexico City declining in importance relative to areas on the US±Mexico border as trade barriers between Mexico and the United States were lowered. In contrast, the IM programme was widely expected to bring a higher level of inward investment in Europe from non-European ®rms by raising the relative locational advantages of the European market. An additional implication of this argument is that the internal market process may have served to divert investment into the EU at the expense of other locations (Baldwin et al., 1995). There is some support for these hypotheses in the econometric results obtained by Aristotelous and Fountas (1996) and Pain (1997). The former ®nd that the level of inward investment in Europe by US and Japanese ®rms since 1987 has been signi®cantly higher than might otherwise have been expected, whilst the latter ®nds some evidence that UK companies have increased their investment in Europe since 1990 at the expense of investment in the United States. There are a number of reasons why the argument in favour of greater concentration within national industries may understate the scope for intra-EU direct investment as product market barriers are removed. At a practical level, the IM legislative process has taken much longer than initially anticipated. By October 1996 an average of 91 per cent of all IM directives were on member states' statute books (European Commission, 1996, p. 8). Our sample period ends in 1994, by which time the internal market had been expected to be completed. The continuing existence of some impediments to trade at that time may make it dif®cult to obtain a full picture of the eventual impact of market integration. Moreover, it is clear that many national factors continue to impose costs on market access. Some of these are regulatory, arising from differences in environmental and health and safety provisions. Examples include the continuing use of national health insurance price controls within the pharmaceuticals industry, regulations on the movement of waste and standards of labelling and packaging. In other cases markets remain differentiated as a result of consumer preferences (Cantwell, 1992). In

Florence Hubert and Nigel Pain 141

such cases, direct investment is often made either to enter local markets or to establish facilities for adapting products to local needs. A number of models arising out of the new literature on economic geography and international trade also suggest that changes in technology and production costs can help to support the existence of multinationals, even at a time when barriers to trade are being reduced (Markusen and Venables, 1996). The key features of these models are interactions between ®rm-speci®c assets, economies of scale and transport (or trade) costs. Firm-speci®c, knowledge-based assets include process innovations, marketing skills and managerial expertise. All can be utilised in different plants under common ownership, giving economies of scale at the level of the ®rm rather than at plant level. Multinational ®rms thus have lower variable costs but higher ®xed costs than national ®rms in different locations. Markusen and Venables (ibid.) argue that continuing integration, and hence expansion in market size, may lead to a gradual substitution of `horizontal' foreign investment for intra-industry trade between countries within integrated regions. This is because the variable cost advantage of multinational ®rms arising from the use of a joint input across plants comes to dominate the higher ®xed costs of multiplant operations. However, such a result would depend upon the structure of both industries and countries. Finally, it is possible that direct investment may be as much motivated by strategic considerations as by a desire to seek out low-cost locations (Buigues and Jacquemin, 1994). If product markets are imperfectly competitive, the sunk costs incurred in undertaking FDI may be a means of achieving the market power required to exploit fully intangible assets such as brand names, managerial expertise and other ®rm-speci®c knowledge. Such factors imply that direct investment might continue to take place even as trade liberalisation occurs. A broad summary of the expectations at the start of the IM process was thus that the measures would raise inward FDI from outside the EU, but eventually lower intra-EU FDI in the tradeable goods sector. Dunning (1996) provides a more detailed discussion of these and other issues. Our brief review of the existing literature on the expected effect of the IM programme suggests a number of hypotheses worthy of further investigation in the econometric work. In particular, it is of interest to test whether intra-EU investment by German companies since 1986 has been signi®cantly different from what might otherwise have been expected, and if so, whether there is evidence of investment diversion into the EU and whether there are differential industrial effects.

142 Economic Integration and German FDI

8.4

A model of Foreign Direct Investment

In practice, there is a wide range of factors that may determine the pattern of specialisation and location over time. Multinational enterprises arise through a combination of industrial organisation motives that result in a number of activities being placed under common ownership and control, and comparative advantage reasons that cause these activities to be placed in separate countries (Krugman, 1995). Our analysis attempts to allow for these various in¯uences. The basic form of the model we employ can be expressed as: FDIij;t ˆ ij ‡ 1 FDIi;t

j

‡ 2 Yj;t ‡ 3 RELCj;t ‡ 4 RD7i;t ‡ 5 INTt

1

‡

6 GEQt ‡ 7 STRj;t ‡ 8 VOLj;t ‡ i Mij;t ‡ 1i;t where FDIij;t denotes the stock of foreign direct investment in sector i in country (or region) j at time t, Yj is a measure of market size, RELCj denotes average production costs in the host location relative to a weighted average of all possible hosts (including Germany), RD7i is an industryspeci®c measure of R&D expenditure in Germany, INT is a measure of interest gearing and GEO is the growth of German equity prices in real terms. Additional characteristics of potential hosts are captured by the number of labour disputes in their economies (STR) and the volatility of their bilateral nominal exchange rate against the Deutschmark (VOL). Finally, we include a set of industry and country dummies to investigate the impact of the internal market. We brie¯y describe each of these variables in more detail below. The sector- and location-speci®c ®xed effects, aij , allow for unobserved in¯uences that remain constant over the whole of the sample period. All other in¯uences will be contained in the disturbance term vijt . The ®xed effects may capture factors, such as contiguous borders and language, that are not re¯ected in the other variables. Conventional supply-side models of FDI emphasise the role of market size and relative costs (see, Barrell and Pain, 1996). Given that there is a cost advantage to producing outside the home country of the investor, a large market would be expected to raise the level of foreign investment. In the empirical work, we proxy market size by industrial production in the host location. To capture the role of relative costs of production, we use a weighted average of labour and capital costs in the host location relative to a weighted average of costs in the competing locations (including Germany).2 This provides a measure of the real exchange rate. For labour costs, we use `normalised' trend unit-labour-cost data for manufacturing produced by the IMF, converted into a common currency.3 Unit costs are

Florence Hubert and Nigel Pain 143

used to allow for differentials in productivity levels as well as wages and payroll taxes. Allowance also needs to be made for the user cost of capital in explaining the level of foreign investment, as this is part of the overall cost differential between operating abroad and at home. In practice, overseas capital investment by German companies appears to have largely been ®nanced from domestically generated funds,4 so it is unlikely that a signi®cant effect could be obtained from the relative costs of borrowing in alternate locations. However, even if multinational ®rms predominantly choose to raise ®nance from one source, international differences in corporate tax systems may still affect the cost of capital in different locations, due to differences in both the allowances available for capital investment and the tax rate on retained earnings. To allow for such differences our measure of relative costs includes a measure of the cost of capital for a German ®rm operating in the host economy, relative to the cost of capital faced by a German ®rm operating in other potential host countries. These variables are drawn from Pain and Young (1996) and assume that ®nance for the investment is raised in Germany.5 As already discussed, the decision to establish foreign operations may also re¯ect factors internal to the ®rm, in particular the existence of ®rmspeci®c, knowledge-based assets (Markusen, 1995). Such assets may be expected to stimulate FDI because of the variable cost advantage they can give a multinational ®rm, and because they affect the trade-off between direct investment overseas and the alternative of licensing. Licensing arrangements can give rise to a variety of transactions costs. Typically, these arise from the dif®culty of writing state-contingent contracts to ensure the maintenance of product quality and guard against the appropriation by potential competitors of the rents accruing from ®rmspeci®c knowledge. In this chapter we follow the basic approach of Pain (1997) and Barrell et al. (1996) and assume that the level of ®rm-speci®c assets is captured by the cumulated `stock' of research and development (R&D) undertaken by German companies in particular industries. Many German corporations, especially in industries such as automobiles and mechanical engineering, have been at the forefront of important innovations in business practice (Patel and Pavitt, 1989), the level of business enterprise expenditure on R&D having risen sharply over our sample period from 1.25 per cent of GDP in the early 1970s to more than 2 per cent of GDP by the late 1980s. In the empirical work, we use a seven-year, cumulative, sector-speci®c, moving average of German R&D expenditure (in real terms) as a measure of the `stock' of ®rm-speci®c assets. A cumulative measure is used as there

144 Economic Integration and German FDI

is likely to be some time lag before the full commercial potential of most research is realised. A number of studies also suggests that the investment decisions of companies, both at home and abroad, are affected by domestic ®nancial conditions, although there appears to be little agreement on how these are best measured. Carlin (1996) provides an overview of the impact of corporate pro®tability on the pattern of domestic investment in Germany, while Dinenis and Funke (1994) obtain a signi®cant effect from real equity prices on domestic ®xed investment. Heiduk and Hodges (1992) suggest that the planned foreign investment of some leading German multinationals is reduced before domestic investment in times of ®nancial distress. In this chapter, we concentrate on the impact of the growth in real equity prices in Germany and a measure of corporate interest gearing. Interest gearing captures information on the evolution of corporate debt and debt service burdens. Fluctuations in interest gearing have already been shown to have a signi®cant effect on the level of foreign investment by UK ®rms (Pain, 1997).6 We have also included two variables that are designed to capture other characteristics of potential host economies. It is sometimes argued that cross-border investment may depend on the ¯exibility and quality of the workforce as much as its basic cost. Eltis and Higham (1995) claim that improvements in the ¯exibility of the labour market have been a primary factor behind the high level of inward investment in the UK since 1979. Flexibility is affected by a large number of institutional features in labour, product and housing markets, and is thus dif®cult to capture in any empirical study of this sort. Here we follow Moore (1993) and utilise data on the number of days lost through strikes in each of the host economies. We expect that more strike-prone locations will receive less inward investment. The literature on FDI also suggests that currency variability might affect direct investment. Volatility in the exchange rate may directly contribute to uncertainty about the timing of planned transactions. If companies are risk-averse, uncertainty about future nominal exchange rate movements may act as a barrier to foreign purchases. Barrell and Pain (1996) illustrate that expectations of a future movement in the dollar have a signi®cant effect on the level of current investment by US parent companies. It may also be the case that the impact of currency variability on investment from a particular location is dependent on the importance of that location within the wider regional market (Barrell et al., 1996). This is particularly true of Germany, since the German market is the primary destination for many tradeable goods produced elsewhere, especially within Europe. A

Florence Hubert and Nigel Pain 145

testable implication of this is that German ®rms may prefer to produce in countries whose nominal Deutschmark exchange rates are relatively stable. We construct a measure of nominal bilateral exchange rate volatility using a three-year moving-sample standard deviation in the rate of change of the bilateral exchange rate of the host economy with Germany. Letting ej denote the nominal, bilateral Deutschmark exchange rate of the host economy, nominal volatility is given by: " VOLj;t ˆ

m X …1=m† ‰
1n…ej;t‡1 k †Š2

#0:5

kˆ1

This measure will be zero for any country whose exchange rate is fully pegged against the Deutschmark, and a constant for any country whose bilateral exchange rate changes at a constant rate. The basic model is also augmented by a lagged dependent variable. This allows for the potential existence of adjustment costs arising from factors such as delivery lags and delays in ®nding suitable investments overseas. We therefore estimate a dynamic, partial adjustment panel model. In contrast, previous studies of intra-EU FDI, such as those by Culem (1988) and Molle and Morsink (1991), simply relate the current change in the investment stock to factors that might be expected to determine the longterm desired stock of investment. Omission of the lagged stock ignores potentially valuable information and could result in misspeci®cation. Although the panel has a relatively rich time dimension, with 18 observations per panel member, the inclusion of a lagged dependent variable may induce small sample bias into panel estimates produced using OLS, so that an instrumental variable estimator has to be employed. There are a number of potential instruments that can be used for the lagged dependent variable. In this chapter, we employ two additional instruments: a simple linear time trend and the rank order of the lagged dependent variable (Durbin, 1954).7

8.5

Modelling the internal market

The most important question in an exercise of this kind is how to incorporate the effects of the IM programme into the speci®cation of the model. Baldwin and Venables (1995) provide a detailed overview of the range of techniques that have been employed in previous empirical studies. We have argued above that the impact of the IM programme can be expected to have varied by sector. Thus, it would not be appropriate to

146 Economic Integration and German FDI

include either a single dummy variable set to unity from 1987 across all sectors, or to draw ®rm conclusions from a set of individual sector dummies, since these could capture other unobserved in¯uences as well. Some indication of the relative sensitivity of particular sectors in the ®rst place is required. Here we follow Pain (1997) and use an ordinal variable (ranging from 1±3) for the level of non-tariff barriers in particular sectors. Sectors where the programme is expected to have a large impact are given a ranking of 3. Sectors with moderate and little impact are given rankings of 2 and 1 respectively. A related measure has also been used in studies of trade patterns in Europe (Sapir, 1996). Pain (1997) uses a measure constructed from European Commission estimates of the importance of non-tariff barriers at the industry level. An alternative strategy is to devise an indicator based on more speci®c information on industrial developments. In this chapter, we exploit data on the number of cross-border mergers and acquisitions within the European Community from 1982/3 to 1990/1. The Single European Act came into force in the middle of this period. We use a regression of mergers on output growth and sector ®xed effects for the period 1982/3± 1986/7 to generate forecasts of the level of cross-border mergers that might have been expected in subsequent years given economic growth.8 Sectors were then ranked on the basis of the difference between the actual and predicted number of mergers. On this basis, the road transport industry was given a ranking of 3, the chemicals and electrical engineering industries a ranking of 2 and the mechanical engineering industry a ranking of 1. Cross-border mergers in the ®rst three of these industries were higher than predicted after 1986. The econometric work in this chapter relies on the internal market indicator based on the pattern of mergers and acquisitions. Whilst this offers a consistent estimate for each sector, it should be recognised that the rankings are somewhat arbitrary, although the same criticism could probably also be made of most of the plausible alternative measures. In practice, the results obtained by Pain (1997) suggest that the choice of measure may make relatively little difference to the econometric ®ndings.

8.6

Empirical results

Estimation is undertaken over a sample period running from 1977 to 1994. With ten separate country (or regional) groupings and four separate manufacturing sectors, there is a total sample size of 720. Most of the main explanatory variables are entered in logarithmic form, permitting direct estimates of their elasticities. We use constructed data on the stock

Florence Hubert and Nigel Pain 147

of direct investment at constant prices. This was obtained by de¯ating the nominal (dollar) value of the FDI stock by the national GDP de¯ator of each host location. 8.6.1

The basic model

The econometric results are summarised in Table 8.2. The ®rst column reports the parameter estimates for the basic panel model. Overall, the coef®cients are in accordance with our priors, suggesting that foreign investment by German ®rms is driven by strategic factors and ®rm-speci®c competitive advantages as well as by a desire to relocate to lower-cost sites. There appears to be an important role for the accumulated sectoral level of R&D expenditure undertaken by German companies, suggesting that ®rms in research-intensive industries are more likely to engage in foreign investment. There are also well-determined effects from relative costs and market size, with both having a long-run elasticity that is greater than unity. There is also evidence that labour relations in the host country are of importance, with an implied long-run elasticity on the strike variable of ±0.14 per cent, so that a rise in the number of strikes will reduce inward investment. There are very signi®cant effects from both the growth in real equity prices and interest gearing, with a sustained 1 per cent rise in gearing ultimately predicted to reduce the stock of foreign investments by 0.37 per cent. These ®ndings suggest that ®nancial factors have had an important in¯uence on the timing and scale of direct investment by German companies, con®rming the ®ndings from the case studies cited by Heiduk and Hodges (1992) and the earlier ®ndings for the United States and the UK by Barrell and Pain (1996) and Pain (1997). These terms should be seen primarily as an indicator of the extent to which changes in domestic ®nancial conditions affect the timing and the size of the ¯ow of direct investment. As neither can be expected to trend permanently over time, they cannot be the primary factor behind the continuing upward trend in the stock of investment. We also obtain a signi®cant effect from our proxy variable to capture the impact of currency volatility. The reported negative coef®cient suggests that German corporations value nominal exchange rate stability, and is consistent with the notion that Germany may be the ®nal market for some goods produced elsewhere in Europe. An implication of this is that German ®rms may prefer to produce in countries that have made a commitment to link their nominal exchange rates to the Deutschmark. The separate sector dummies provide a test of whether there has been a common structural break in investment in any sector in EU locations

148 Economic Integration and German FDI

since the advent of the IM programme. The coef®cients indicate that German investment in the EU has been higher in all sectors from 1987 than might otherwise have been expected. Three out of the four sector Table 8.2 The determinants of German FDI: dependent variable: ln (FDI)ijt , sample period 1977±94 (1) ln(FDIij )t 1 ln(Yj )t ln(RELCj )t ln(RD7i )t ln(INT)t 1 GEOt ln(STRj )t 1 VOLjt IMCHjt IMMEjt IMEEjt IMTRjt IMjt NLIMit BGIMit UKIMit FRIMit ITIMit SPIMit IMEFjt IMNAjt IMLAjt IMSAjt EFIMit NAIMit LAIMit SAIMit R2 Standard error Serial correlation

(2)

(3)

(4)

0.4521 (9.1) 0.4370 (8.6) 0.4358 (8.7) 0.4337 (8.8) 0.7389 (3.3) 0.5415 (2.5) 0.7324 (3.4) 0.6437 (2.2) ±0.5766 (4.6) ±0.5632 (4.5) ±0.6078 (4.8) ±0.6440 (5.1) 0.3588 (3.2) 0.4829 (4.4) 0.4602 (4.2) 0.6443 (5.0) ±0.2018 (3.4) ±0.2246 (3.8) ±0.2062 (3.6) ±0.2735 (4.5) 0.3273 (4.3) 0.3316 (4.4) 0.3085 (4.1) 0.2641 (3.2) ±0.0764 (3.1) ±0.0775 (3.2) ±0.0471 (1.8) ±0.0583 (2.3) ±0.2834 (3.4) ±0.3019 (3.6) ±0.3299 (4.1) ±0.3429 (3.2) 0.0636 (1.2) ± ± ± 0.1088 (2.1) ± ± ± 0.1379 (2.4) ± ± ± 0.3533 (4.9) ± ± ± ± 0.0797 (4.3) 0.1102 (2.9) 0.0795 (2.1) ± ± ±0.0387 (0.4) ±0.0495 (0.5) ± ± ±0.2579 (2.8) ±0.2744 (2.8) ± ± 0.1762 (1.9) 0.1566 (1.6) ± ± ±0.1596 (1.8) ±0.1820 (2.0) ± ± 0.0282 (0.3) 0.0137 (0.1) ± ± ±0.1694 (1.8) ±0.1788 (1.8) ± ± ± 0.0024 (0.1) ± ± ± ±0.2647 (3.2) ± ± ± ±0.0378 (0.5) ± ± ± ±0.1587 (1.9) ± ± ± ±0.0664 (0.4) ± ± ± 0.3450 (1.9) ± ± ± 0.0008 (0.1) ± ± ± 0.2621 (1.3) 0.9711 0.9706 0.9718 0.9722 0.2453 0.2473 0.2422 0.2403 Chi(1) = 2.11 Chi(1) = 1.85 Chi(1) = 1.47 Chi(1) = 2.01

Notes: T-statistics in parentheses. IMCHj , IMEEj , IMEEj and IMTRj are dummies for the chemicals, mechanical engineering, electrical engineering and transport industries in the EU locations. All are zero before 1987. IMj = [IMME + 2*(IMCHj + IMEE j ) + 3*IMTRj ]. IMEFjt , IMNAj , IMLAj and IMSAj are equivalent series for the EFTA, North American, Latin American and South-East Asian blocs. BGIMi , UKIMi , FRIMi , ITIMi , NLIMi and SPIMi are country dummies for Belgium, the UK, France, Italy, the Netherlands and Spain respectively. Each is equal to 1 from 1987 across all sectors and zero otherwise. The equivalent dummies for the other four blocs are denoted EFIMi , NAIMi , LAIMi and SAIMi respectively.

Florence Hubert and Nigel Pain 149

dummies are individually signi®cant and the four dummies are jointly signi®cant [Chi(4) = 25.14], suggesting a change of some kind in the factors determining direct investment in Europe since 1986. The results provide little support for the hypothesis that German manufacturing companies have actively reduced their outward investment in order to exploit economies of scale from higher domestic production as trade barriers weaken. The size of the coef®cient on the lagged dependent variable suggests that there is some modest inertia in the pattern of outward investment. It is of interest to note that the coef®cient is signi®cantly different from unity, implying that the use of a model speci®ed in ®rst difference form, using investment ¯ows without any effect from the lagged stock level, would be rejected by the data. The parameters in column 1 of Table 8.2 appear to be well determined and there is no sign of signi®cant ®rst-order error correlation over time. The test statistic reported in the table was obtained using an auxiliary regression procedure, with the lagged residuals being added into the equation as an additional regressor and tested for signi®cance. Further details of this test are provided in Barrell et al. (1996, p. 15). We have argued above that it is not appropriate to conclude that the internal market has had a particular effect simply by inspecting the coef®cients on the individual sector dummies, since these could re¯ect other unobserved common factors as well. In the second column we replace the individual sector dummies with the weighted dummy (denoted IMij ) based on the prior sensitivity rankings. The dummy variable has a signi®cant positive coef®cient, implying that German investment in the EU has, on average across sectors and countries, been higher than might otherwise have been expected since 1987. The continued presence of the lagged dependent variable means that the model implies that the effects of the internal market have accumulated over time. The tests reported so far have explored the possibility of variation between sectors across countries. It is also possible to allow for variation between countries across sectors. A test for variation within countries is reported in the third column of Table 8.2. Here we have included separate dummies for each EU country, set to unity from 1987 onwards for all sectors. The IM sector dummy is also included. This speci®cation relaxes two implicit restrictions in the previous equations: the common magnitude of effects across countries and the common direction of effects across industries. The impact of the internal market on each sector, within each country, is given by the coef®cient on the country dummy

150 Economic Integration and German FDI

plus the coef®cient on the sector dummy multiplied by the ranking of that sector, as reported in Section 8.5 above. Some care is required when interpreting the ®ndings since the country dummies will also pick up any factors that are otherwise accounted for and have changed systematically within each host country since 1987. The national dummies are jointly signi®cant, suggesting that there is considerable variation across countries [Chi(6) = 32.16]. This was con®rmed by the rejection of an attempt at imposing identical coef®cients on the country dummies [Chi(5) = 31.31], implying that there have been signi®cantly different patterns of direct investment in European countries since 1987. In most cases, the dummies have a negative coef®cient. However, for all countries this is offset, in some sectors at least, by the positive coef®cient on the IM variable. Even so, the implied internal market effects differ markedly between countries. For the UK and Italy, the combination of the positive coef®cients on the country and sector dummies means that investment in all sectors has been higher since 1987. In contrast, the Belgium dummy (BGIM) has a signi®cant negative coef®cient, implying that German investment has declined in nearly all sectors. We return to this issue in more detail below. In column 4 of Table 8.2 we investigate the impact of the internal market on German direct investment in non-EU locations. We augment the previous model with a number of additional industry and sector dummies for the four remaining regional blocs ± the EFTA (EF), North American (NA), Latin American (LA) and South-East Asian (SA) blocs. All dummies have the same pattern as their EU equivalents and are set at zero for all EU countries. For all four non-EU locations there is some evidence that German investment has, on average, been lower than might have been expected in their manufacturing sectors since 1987. This is consistent with the hypothesis that the IM programme has diverted some investment into the EU (Baldwin et al., 1995). This appears to be particularly true of investments in North America. However, neither of the IM dummies for the EFTA bloc are signi®cant, possibly because it became clear at an early stage that these countries would implement the internal market legislation in full, initially via membership of the European Economic Area, established in 1992, and then, at least for Austria and Sweden, as a result of entry into the EU itself. The eight separate non-EU dummies are jointly signi®cant [Chi(8) = 17.83], providing evidence of signi®cantly lower investment than might otherwise have been expected in these locations since 1987. It is worth noting that the introduction of country dummies has led to some changes in the other coef®cients in the model compared with the

Florence Hubert and Nigel Pain 151

initial regression. In particular the size and signi®cance of the R&D measure has risen, with the long-run elasticity in column 4 being over unity. This suggests that its coef®cient was initially biased downwards as a result of the conjunction of the weak level of new investment outside the EU since the late 1980s and the continued high level of new R&D expenditure since that time. 8.6.2

Evaluating the impact of the internal market

The full matrix of coef®cients implied by column 4 for the impact of the internal market on investment by sector and country is shown in Table 8.3. This indicates that German investment appears to have risen much more rapidly since 1987 with regard to the UK, the Netherlands and Italy than might have been expected. Investment in Spain, France and especially Belgium has been lower than might have been expected, although in some cases the implied coef®cients are insigni®cantly different from zero. Outside Europe, investment appears to have been signi®cantly lower in three sectors in North America and in transport equipment in South-East Asia. Whilst investment has fallen in the EFTA and Latin American blocs, the coef®cients are not signi®cant. It is possible to use the estimated relationships with the sector and country dummies to calculate the direct effect of the IM programme on the stock of German FDI in any particular sector within a EU member state.9 Here we use Equation 8.4, although there are obviously a number of alternative equations upon which the calculations might be based.10 The methodology is based on that described in Pain (1997, Appendix 1), extended to allow for the country dummies as well. Table 8.3

The impact of the internal market on FDI, by country and sector

Netherlands Belgium UK France Italy Spain EFTA North America Latin America South-East Asia

Chemicals

Mechanical engineering

Electrical engineering

Transport equipment

0.1095 (1.4) ±0.1155 (1.6) 0.3155 (4.0) ±0.0230 (0.3) 0.1726 (2.3) ±0.0199 (0.3) ±0.0615 (0.8) ±0.1844 (2.3) ±0.0749 (0.7) ±0.0554 (0.6)

0.0300 (0.4) ±0.1950 (2.6) 0.2360 (2.9) ±0.1025 (1.4) 0.0931 (1.2) ±0.0994 (1.2) ±0.0640 (0.6) 0.0803 (0.7) ±0.0370 (0.3) 0.1034 (0.8)

0.1095 (1.4) ±0.1155 (1.6) 0.3155 (4.0) ±0.0230 (0.3) 0.1726 (2.3) ±0.0199 (0.3) ±0.0615 (0.8) ±0.1844 (2.3) ±0.0749 (0.7) ±0.0554 (0.6)

0.1889 (2.0) ±0.0361 (0.4) 0.3950 (4.2) 0.0564 (0.6) 0.2520 (2.8) 0.0596 (0.7) ±0.0591 (0.5) ±0.4490 (3.8) ±0.1127 (0.8) ±0.2141 (1.7)

Notes: Coef®cients derived from country and industry dummies in column 4 of Table 8.2. T-statistics reported in parentheses.

152 Economic Integration and German FDI Table 8.4

The impact of the internal market on German FDI in the EU 1994 stock ($bn,1990)

Belgium and Luxembourg UK France Italy Netherlands Spain and Portugal Sector Chemicals Mechanical engineering Electrical engineering Transport equipment Total

IM effect ($bn,1990)

3.61 5.14 4.91 2.94 1.15 5.44

±0.69 2.30 ±0.25 0.73 0.17 0.09

9.63 2.55 4.70 6.31 23.19

0.30 0.03 0.59 1.43 2.35

The results for each EU member state and each individual sector are summarised in Table 8.4. These indicate that, as of 1994, the IM programme is estimated to have raised the stock of German manufacturing FDI in the EU by some $2.4 billion, at constant 1990 prices, equivalent to 10 per cent of the aggregate stock level. The primary bene®ciary of the higher level of outward investment by German ®rms appears to have been the UK, where investment is some $2.3 billion higher than elsewhere, over one third of the reported stock level. As might be expected, given the coef®cients shown in Table 8.3, Italy and the Netherlands have also gained additional investment. The IM programme is estimated to have reduced German investment in Belgium and France overall, largely re¯ecting a drop in chemicals investments in both countries. In proportionate terms the IM programme has had a particularly marked effect on the transport sector, accounting for nearly a quarter of the outstanding stock of German direct investment in the EU. Economic integration appears to have had less of an impact on the chemicals and mechanical engineering industries. Overall, the effects found for Germany are somewhat smaller than those obtained for manufacturing FDI by UK companies in Pain (1997), whose results imply that the IM programme has raised UK manufacturing FDI in the EU by some $5.7 billion, at constant 1990 prices. This difference may imply that more German companies have been able to exploit economies of scale by concentrating production at home. This is especially true of the chemicals industry, where there is a greater tendency for production to be concentrated in a single location. The sector that

Florence Hubert and Nigel Pain 153

shows the fastest growth in investment `transport equipment' tends to have a greater degree of vertical integration in production, with dispersed production facilitated by lower barriers to cross-border trade. The evidence we obtain of greater cross-border investment is consistent with other developments since the advent of the IM programme, as reported by the European Commission (1996, ch. 6). Concentration ratios, measured as the market share of the four largest ®rms, rose at the EU level between 1987 and 1993, but not at the national level, and intraindustry trade has increased. Thus, adjustments to regional economic integration appear to have taken place within ®rms rather than within industries.

8.6

Conclusions

This chapter has investigated the factors determining the scale and location of German manufacturing FDI, with allowance being made for differential effects between regions and sectors. The approach used augments a conventional supply-side model of production location with measures to re¯ect both domestic ®nancial conditions and internal ®rmspeci®c developments within industries. Our results provide evidence that the comparatively high level of R&D expenditure by German corporations has had a signi®cant in¯uence on the level and pattern of outward investment. To this extent, high levels of outward investment by many German ®rms could be construed as a sign of competitive health rather than weakness. More importantly, it suggests that FDI may be an important vehicle for the transmission of ®rm-speci®c innovations throughout Europe. Our results also suggest that the IM programme has had a signi®cantly positive impact on the level of intra-EU investment by German corporations, with some evidence of investment diversion from locations outside the EU. The largest single bene®ciaries of the extra investment appear to be the UK, the Netherlands and Italy, and there is evidence of a modest increase in investment in the southern periphery of the EU. Overall, our results suggest that regional economic integration may help to stimulate the diffusion of new technologies and ideas within the region and thus, as Barrell and Pain (1997) argue, help to raise technical progress and economic growth. Future research might usefully seek to ®nd out more about the forms of activity undertaken by foreign subsidiaries in the EU. Our ®ndings are consistent both with models of horizontal direct investment, such as that by Markusen and Venables (1996), in which ®rms produce similar

154 Economic Integration and German FDI

products in different locations, and models in which ®rms aim to broaden their international division of labour. To address this question it would be of considerable interest to seek to augment the ®ndings from sectoral econometric analyses with those from more detailed case studies of individual multinational ®rms. Notes 1. We are grateful to Ray Barrell, John Dunning and Melanie Lansbury for helpful comments and suggestions, and to the ESRC for ®nancial support (grant number R000236306). 2. Costs are given by wa ± c…1 a† , where w denotes unit labour costs, c the cost of capital and (1±a) is the average capital income share in the German business sector over our sample period. 3. For Latin America, we use constructed data on manufacturing unit labour costs in Argentina, Brazil and Mexico. For South-East Asia, we use equivalent series for Singapore and India plus the normalised trend unit cost series for Japan. 4. Jungnickel (1995) estimates that the ®xed assets of the foreign af®liates of German companies totalled DM226 billion in 1992, compared with a total outward direct investment stock of DM288 billion. The FDI stock was around 15 per cent of the total (®xed plus current) assets of foreign af®liates, close to the share in 1980. 5. Data is available for all locations apart from Latin America. 6. If capital markets were perfect then, given the Modigliani±Miller theorems, the cost of capital should be independent of the capital structure of the ®rm. However, if there are dead-weight costs associated with bankruptcy, then companies' real decisions may not be independent of their ®nancial situation. 7. The rank order is clearly strongly correlated with the variable being instrumented, but has been `cleaned' of the lagged disturbance term. However it may be a weak instrument if substantial measurement error is present in the instrumented variable, and hence in the associated rank order. 8. The basic regression was: Mergersi;t = 0.7013 Growthi;t 1 + Fixed effects; (3.02) R2 = 0.38, where Mergersi;t denotes the number of mergers in sector i in year t, and Growth denotes value-added output in the EU. The number of mergers appeared to be unrelated to the ¯ow of new investment in the EU by German companies, as the latter variable was not signi®cant when added to the basic regression. 9. The internal market may also have had an indirect effect to the extent that the programme has affected both economic growth and cost competitiveness in the host location. 10. Again, it should be emphasised that such estimates need to be treated with a degree of caution because the country dummies may be picking up additional effects unrelated to the internal market.

References Aristotelous, K. and S. Fountas (1996) `An Empirical Analysis of Inward Foreign Direct Investment Flows in the EU with Emphasis on the Market Enlargement Hypothesis', Journal of Common Market Studies, vol. 34, pp. 571±83.

Florence Hubert and Nigel Pain 155 Baldwin, R., R. Forslid and J. Haaland (1995) `Investment Creation and Investment Diversion: Simulation Analysis of the Single Market Programme', CEPR Discussion Paper, no. 1308 (London: CEPR). Baldwin, R. E. and A. J. Venables (1995) `Regional Economic Integration', in G. M. Grossman and K. Rogoff (eds), Handbook of International Economics Volume III, (Amsterdam: North-Holland). Barrell, R. and N. Pain (1996) `An econometric analysis of US foreign direct investment', Review of Economics and Statistics, vol. L X X V I I I , pp. 200±l7. Barrell, R. and N. Pain (1997) `Foreign direct investment, technological change and economic growth within Europe', Economic Journal. Barrell, R., N. Pain and F. Hubert (1996) `Regionalism, innovation and the location of German direct investment', National Institute of Economic and Social Research Discussion Paper no. 91, (London: NIESR). Brulhart, M. and J. Torstensson (1996) `Regional integration, scale economies and industry location in the European Union', CEPR Discussion Paper no. 1435 (London: CEPR). Buigues, P. and A. Jacquemin (1994) `Foreign direct investment and exports to the European Community', in D. J. Encarnation and M. Mason (eds), Does Ownership Matter? Japanese Multinationals in Europe (Oxford: Oxford University Press). Cantwell, J. (1992) `The Effects of Integration on the Structure of Multinational Corporation Activity in the EC', in M. W. Klein and P. J. J. Welfens (eds), Multinationals In The New Europe and Global Trade (Berlin: Springer-Verlag). Carlin, W. (1996) `West German growth and institutions, 1945±90', in N. Crafts and G. Toniolo (eds), Economic Growth In Europe Since 1945 (Cambridge: Cambridge University Press). Culem, C. G. (1988) `The locational determinants of direct investments among industrialised countries', European Economic Review, vol. 32, pp. 885±904. Dinenis, E. and M. Funke (1994) `Factor prices, employment and investment in UK and West German manufacturing', The Manchester School, vol. L X I I , pp. 412±24. Dunning, J. H. (1996) `The European Internal Market Program and Inbound Foreign Direct Investment', University of Reading Discussion Papers in International Investment and Business Studies, no. 217 (Reading: University of Reading), Journal of Common Market Studies. Durbin, J. (1954) `Errors in variables', Review of the International Statistical Institute, vol. 22, pp. 23±32. Eaton, B. C., R. G. Lipsey and A. E. Safarian (1994) `The theory of multinational plant location in a regional trading area', in L. Eden (ed.), Multinationals In North America (Calgary: The University of Calgary Press). Eltis, W. and D. Higham (1995) `Closing the UK competitiveness gap', National Institute Economic Review, vol. 154, pp. 71±84. Emerson, M., M. Aujean, M. Catinat, P. Goybet and A. Jacquemin (1988) The Economics of 1992: The EC Commission's Assessment of the Effects Of Completing The Internal Market (Oxford: Oxford University Press). European Commission (1996) The Single Market And Tomorrow's Europe (London: Kogan Page). Hanson, G. H. (1997) `Increasing returns, trade and the regional structure of wages', Economic Journal, vol. 107, pp. 113±33.

156 Economic Integration and German FDI Heiduk, G. and U. W. Hodges (1992) `German multinationals in Europe: patterns and perspectives', in M. W. Klein and P. J. J. Welfens (eds), Multinationals in the New Europe and Global Trade (Berlin: Springer-Verlag). Hoeller, P. and M.-O. Louppe (1994) `The EC's Internal Market: Implementation and Economic Effects', OECD Economic Studies, vol. 23, pp. 55±108. Itaki, M. and M. Waterson, (1990) `European multinationals and 1992', University of Reading Discussion Papers in International Investment and Business Studies, no. 141 (Reading: University of Reading). Jungnickel, R. (1995) `Foreign direct investment, trade and employment: the experience of Germany', in Foreign Direct Investment, Trade and Employment (Paris: OECD). Krugman, P. R. (1995) `International trade theory and policy', in G. Grossman and K. Rogoff (eds), Handbook of International Economics Volume III (Amsterdam: Elsevier). Markusen, J. R. (1995) `The boundaries of multinational enterprises and the theory of international trade', Journal of Economic Perspectives, vol. 9, pp. 169±89. Markusen, J. R. and A. J. Venables (1996) `The Increased Importance of Direct Investment in North Atlantic Economic Relationships: A Convergence Hypothesis', in M. B. Canzoneri, W. J. Ethier and V. Grilli (eds), The New Transatlantic Economy (Cambridge: Cambridge University Press). Millington, A. I. and B. T. Bayliss (1991) `Non-tariff barriers and UK investment in the European Community', Journal of International Business Studies, vol. 91, no. 4, pp. 695±710. Molle, W. and R. Morsink (1991) `Intra-European direct investment', in B. Bu È rgenmeier and J. L. Mucchielli (eds), Multinationals and Europe 1992: Strategies For The Future (London: Routledge). Moore, M. O. (1993) `Determinants of German manufacturing direct investment 1980±1988', Weltwirtschaftliches Archiv, vol. 129, pp. 120±38. Niosi, J. (1994) `Foreign Direct Investment in Canada', in L. Eden (ed.), Multinationals In North America (Calgary: The University of Calgary Press). Pain, N. (1997) `Continental drift: European integration and the location of UK foreign direct investment', The Manchester School. Pain N. and G. Young (1996) `Tax competition and the pattern of European foreign direct investment', Paper presented at 52nd Congress of the International Institute of Public Finance, Tel Aviv, August, and the IFS conference on Public Policy and the Location of Economic Activity, November. Patel, P. and K. Pavitt (1989) `A comparison of technological activities in West Germany and the United Kingdom', National Westminster Bank Quarterly Review, May, pp. 27±42. Sapir, A. (1996) `The Effects of Europe's Internal Market Programme on Production and Trade: A First Assessment', Weltwirtschaftliches Archiv, vol. 32, no. 3, pp. 457± 75. Thomsen, S. and S. Woolcock (1993) Direct Investment and European Integration: Competition Among Firms and Governments (London: Pinter). UNCTAD (1996) World Investment Report 1996 (Geneva: United Nations). Yannopoulos, G. N. (1990) `Foreign Direct Investment and European Integration: The Evidence From the Formative Years of the European Community', Journal of Common Market Studies, vol. XXVIII, pp. 235±59.

9

Spanish Unemployment ± a Hiring Function Approach: GARCH-M Model Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos

9.1

Introduction

The high and persistent unemployment in Spain, and to a lesser extent in Europe as a whole, has become the central theme in macroeconomic policy. Reaching a rate of 25 per cent in 1995 and standing at 21.7 per cent in the ®rst quarter of 1997, Spanish unemployment continues to be a highly important topic for many studies. These attempt to explain the main factors behind the phenomenon, as well as the characteristics of the economies of Spain and other European countries that make them different from countries with a more favourable employment climate, such as Japan and the United States. The purpose of this chapter is to investigate this unemployment puzzle using a matching function as the main tool of analysis. This should enable one to study the evolution of the reemployment rate in terms of the average duration of vacancies and unemployment. The reason why vacancies and unemployment coexist over time in the labour market lies in the following argument. In an economy where there is continuous job creation and job destruction, even if the labour market is very ef®cient there always exists a searching period for job seekers and ®rms. It is thus important to point out that, even in rigid labour markets such as that of Spain, there is a high rotation of workers and jobs. For instance, the gross labour ¯ow1 in Spain (hirings given unemployment, plus redundancies given employment) was 40 per cent in 1987.2 Such a sizeable ¯ow rate suggests that the coexistence of unemployment and vacancies is a relevant phenomenon that deserves further research. The empirical literature on this issue seems to lean towards static, longrun, Cobb±Douglas set-ups that relate, in a log-linear form, hirings with

157

158 Spanish Unemployment ± a Hiring Function Approach

vacancies and unemployment. The main innovation in our methodology is the use of a GARCH-M model, which allows us to study the evolution of the conditional variance of the exit rate over time. There are a number of advantages to using this speci®cation:

. It ®ts the data well.3 All the Cobb±Douglas static models we estimated decisively rejected the hypothesis of constant variance;

. It explicitly models the variance of the exit rate. Intuitively, it is

probable that the interventions of the Spanish government in the labour market and cyclical economic activity have an impact not only on the expected mean of the probability of the unemployed to ®nd a job but also on its expected variance. For instance, this might be possible through the variable strength and durability of these shocks to the country's labour market.

Our sample period (the second quarter of 1979 to the fourth quarter of 1996) is very representative because it includes interesting episodes and covers two fundamentally different epochs. The early 1980s were characterised by the heritage of the Franco regime and weak economic activity, but from the mid-1980s onwards the socialists' labour market reforms began to take effect. Interestingly, the latter period also experienced both strong growth (1986±91) and recession (1991±94).

. By deriving the conditional variance of the exit rate we can introduce it

as an exogenous variable to explain the mean hazard rate of unemployment. Therefore, the elasticity of the exit rate with respect to its variance can be derived. Here we attempt to describe this measure as a ¯exibility proxy of the labour market. Our rationale is as follows: in a very ¯exible labour market the probability of a person ®nding or losing a job should be high. Hence, greater variance in the exit rate is associated with a higher overall exit rate. In our set-up, the latter is a testable hypothesis.

The chapter is structured as follows. Section 9.2 describes the theoretical background underlying the matching function, starting with basic concepts and followed by a standard discussion of Spanish labour market symptoms. Section 9.3 presents the model, the methodology used and the results. An association of our main ®ndings with several well accepted facts about the Spanish hazard rate of unemployment will be also offered. Section 9.4 concludes.

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 159

9.2 9.2.1

The matching function Basic theoretical concepts

The matching function has been proposed as a useful tool for linking the aggregate hazard or job-®nding rate to labour market conditions. Such a function captures the process whereby job seekers and jobs combine to form employment relationships, mapping stocks of the unemployed into a stock of new matches.4 The empirical evidence for Europe (see Layard et al., 1994; Pissarides, 1986) shows that changes in ¯ows out of unemployment are a crucial determinant of changes in the unemployment rate. Thus, it is very interesting to analyse the out¯ow pattern in Spain. If we assume that all ¯ows out of unemployment are due to reemployment, hence ignoring out¯ows from the labour force, then the matching or hiring function is a way to model out¯ows. Let us suppose that all the unemployed are available for work and are looking for it with varying degrees of intensity. Each newly unemployed person will start to look for work within a short period of time. This period may be thought of as the intensity of search by or `choosiness' of a given unemployed person. Firms enter the market for workers by opening up job vacancies. Only a fraction of ®rms at any point in time will be actively engaged in search and recruitment. This fraction describes the intensity of search by or `choosiness' of employers. We can ease the uncertainties in the labour market regarding the search process by imagining that there is a relationship that links the number of workers actively looking for jobs and the number of job vacancies on the one hand, with the number of new jobs that arise at any point in time on the other. The numbers of unemployed workers and job vacancies are U and V respectively, and H denotes the number of successful job matches. We make the reasonable assumption that H increases with both U and V. As H represents a ¯ow out of unemployment, it goes up when either the number of unemployed or the number of un®lled vacancies goes up, for given search intensities. If there are more unemployed workers, this in general means that a ®rm's cost of recruitment is less due to the larger labour supply. A larger number of unemployed people relative to job vacancies means that vacancies are ®lled more quickly, so there are fewer ®rms competing for the available workers and, therefore, it is less costly to ®nd the right worker (the total cost of a vacancy is higher, the greater the number of job vacancies relative to the number of unemployed). In this case, a ®rm can ®ll its vacancies more quickly at the same cost, hence the

160 Spanish Unemployment ± a Hiring Function Approach

®rm will create more vacancies to take advantage of the lower cost. This gives a positive relationship between vacancies and unemployment. Hence, ignoring mismatch,5 we can write the hiring function as: H ˆ h…cU; gV†

…9:1†

where c and g are the effectiveness of the unemployed and the choosiness of the ®rm respectively. With H workers leaving unemployment within the period, the average probability, p, that an unemployed worker will leave unemployment during the period is: p ˆ H=U ˆ h…cU; gV†=U

…9:2†

This is the probability that is behind most of the changes in unemployment. Each hiring requires a marriage between a vacancy and an unemployed person. For this reason, it is sensible to think that in a reasonably sized market the number of marriages will double as the numbers of men and women double. The same reasoning can applied to the labour market. Thus, we can assume that h(
) is normal and linearly homogeneous in V and U. However, this assumption could be questioned. For example, as in the famous coconut parable, when U and V are high (that is, when many people are carrying coconuts to trade) it is more likely that any given person will ®nd a match. Pure increases in scale, corresponding to increases in the number of searchers, improve the ef®ciency of the search of any one worker or employer.6 The theory supports both constant and increasing returns to hiring, but constant returns to scale is an attractive hypothesis that is often well argued in empirical studies.7 Assuming constant returns to scale, we can rewrite Equation 9.2 as: H=U ˆ cgh …V=U; 1†8

…9:3†

which says that the out¯ow rate from unemployment (H/U) only depends on the ratio of vacancies to unemployment and the `effectiveness' of both the unemployed and ®rms. This is an aggregate relation determining out¯ow rates, and it is particularly relevant for the light it sheds on increases in the aggregate unemployment rate, since it explicitly includes the exogenous variable (V/U) that determines the out¯ow or exit rate.9 9.2.2

Background to the arguments behind spanish unemployment

To put ourselves into the Spanish context, we shall outline the possible causes of the high unemployment.

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 161

The Spanish bene®t system is based on the insurance and assistance principles found in most Western economies, but it is far more generous than that of the average European country. Up to 1993 the replacement ratio (the ratio of unemployment bene®t to a person's expected wage) was much higher than for Europe as a whole. For instance, some10 put public expenditure on unemployment compensation in 1987 at 2.50 per cent of GDP, compared with 1.66 per cent in the UK, 1.26 per cent in France and 1.49 per cent in Italy.11 The 1993 reform removed one of the most obvious distortions of the bene®t system by making bene®ts taxable, thus requiring bene®t recipients to pay part of their social security contributions. Mainly for this reason,12 nowadays the level of insurance in Spain (on a replacement ratio basis) is similar to that elsewhere in Europe. However, unemployment bene®t, judged on an eligibility basis, is far more generous than the European average. Indeed, according to Blanchard et al. (1995), along with France, Belgium and Germany, Spain had the highest gross coverage rate in the EU in 1992, despite having the highest unemployment rate in Europe. Moreover, the duration of bene®t payments nowadays ranges from four to 24 months: the ®rst 12 months of contributions provide four months of bene®t and each additional six months of contributions provide two further months of bene®t. Until the 1993 reform, the Spanish insurance system mainly acted as a secondary form of income, making life easier for those who did not have a job, reducing their search intensity and in turn reducing the hazard rate of unemployment. Another signi®cant factor in reducing the intensity of the search for jobs, or even the incentive to accept job offers, is extended family protection. Unlike in most North European countries, singleperson households are only a small minority of the total (some 10 per cent). The fact that offspring tend to remain at home until they marry, plus the high unemployment rate for married women, implies that ± when considered at the household level ± the incidence of unemployment is much less important. In fact, Blanchard et al. (1995) ®nd that there are only 25 per cent of Spanish households (excluding those whose members are out of the labour force) where all the active members are unemployed. An important source of in¯exibility (and thus of choosiness on the ®rm's part) in the Spanish labour market is the excessive employment protection granted to `permanent' workers.13 This, combined with the organisation and structure of collective bargaining, has led to the presence of two classes of worker:

162 Spanish Unemployment ± a Hiring Function Approach

. The have-nots: workers on ®xed-term contracts and the unemployed; . The haves: workers with high unemployment protection. These play a dominant role in collective bargaining.

As a consequence, wage settlements are largely insulated from labour market developments. The Spanish case is, therefore, an outstanding example of the insider±outsider model.14 High unemployment does not raise a risk for the haves, as it has only a small effect on wage settlements and in¯ation. Labour mobility and turnover is, thus, greatly reduced due to the insider±outsider problem.15

9.3

The models

Our basic speci®cation gives new hiring as a Cobb±Douglas function of vacancies and unemployment, with all the variables de®ned as in the previous sections, Ht ˆ AV a1 Uta2

…9:4†

where H refers to hirings, V refers to vacancies and U represents the unemployment level. The Cobb±Douglas function is characterised by the following properties:

. The parameters a1 and a2 are the elasticity of hirings with regard to vacancies and unemployment respectively;

. The hiring function is linearly homogeneous (the degree of homo-

geneity depends on the returns to hirings). a1 + a2 = 1 implies constant returns to hirings (CRTH), a1 + a2 > 1 implies increasing returns to hirings (IRTH), a1 + a2 < 1 implies decreasing returns to hirings (DRTH), . There are constant unitary elasticities between the two variables, which implies that: ˆ

d…U=V† h1 =h2 ˆ1 d…ht =h2 † U=V

…9:5†

where h1 = H/V, h2 = HU In order to make feasible its estimation, we take logs of Equation 9.4 and write it as: ln Ht ˆ a0 ‡ a1 1n Vt ‡ a2 1n Ut The latter equation will be estimated ®rst.

…9:6†

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 163

A usual variation of the above model requires the imposition of the CRTS hypothesis (a1 + a2 = 1). This should be a formally testable assumption. From Equation 9.3 we derive the second equation to be estimated: ln …H=U†t ˆ a0 ‡ a1 ln …V=U†t

…9:7†

The literature (see, Blanchard and Diamond, 1989, 1990; Layard et al., 1991; and Pissarides, 1986) suggests that the time trend could be a very useful addition to the previous relationship. Intuitively, we hope that it will account for missing crucial variables in explaining the variation of the hazard rate of unemployment. Hence the third equation to be estimated is: ln …H=U†t ˆ a0 ‡ a1 ln …V=U†t ‡ a3 T

…9:8†

Our main departure from the empirical literature is to introduce an additional equation to account for the complex variance of the hazard rate of unemployment. Therefore, a GARCH-M model will be formulated. It is reasonable to expect that the probability of an unemployed person obtaining a job is not constant over time. This is a particularly appealing hypothesis in our case because the period concerned (1979±96) includes two fundamentally different political epochs for Spain. At the start of the 1980s the Spanish labour market was plagued by anachronistic state institutions, the shaky post-Franco heritage and weak economic activity.16 In 1982, the socialists came to power with strong restructuring intentions. In the socialist period (during which Spain experienced strong growth from 1986±91 and a marked recession from 1991±94), a number of political measures was taken to reduce unemployment. Irrespective of their absolute outcome, we could expect them to have affected the variance of the hazard rate of unemployment. Let the constant returns hiring function be a Cobb±Douglas type, written as: p ln…H=U†t ˆ a0 ‡ a1 ln…V=U†t ‡  ht ‡ a3 T ‡ ht and ht ˆ b0 ‡

X i

b1 e2t…

i†

‡

X

cj h…t

j† ;

i; i ˆ 1; . . . ; n:

…9:9†

j

where ai , bi , c1 and  are parameters The above system is the fourth estimation and suggests that variance depends on three factors: a constant, last quarter's forecast variances (the

164 Spanish Unemployment ± a Hiring Function Approach

GARCH terms) and last quarter's news about variance, which is taken to be the squared residual from the previous quarter (the ARCH term). This speci®cation of the variance equation incorporates the phenomenon of volatility clustering. Hence, a large hazard rate of unemployment is more likely to be followed by a large hazard rate variance. Obviously, the less apparent term in the ®rst equation is the standard deviation term, pht . It is considered to be a shift variable, since it is natural to test whether the expected probability of an unemployed person obtaining a job is proportional to its expected variance. Measuring this by the conditional variance leads to the ARCH-M (ARCH in mean) model. The estimated coef®cient indicates the variance trade-off; in this case we attempt to interpret it as a ¯exibility proxy. If a higher  causes greater elasticity of the exit rate with respect to its variance, then, by the same token, greater variance in the exit rate is associated with a higher probability of an unemployed person ®nding a job, which implies a ¯exible labour market. 9.3.1

Data properties and discussion

The data we use for our estimations is quarterly and comes from Dolado and Gomez (1996), who used the same information on unemployment and vacancies for their study on the relationship between unemployment and vacancies in Spain. The source of their data on unemployment was the EPA (Labour Force Survey). Job vacancy data were obtained from the National Employment Institute, which reports the number of unemployed people registered at public employment of®ces. Unemployment, U, is de®ned as the number of unemployed persons; vacancies, V, is de®ned as the number employment offers open at the end of each month (as reported by INEM); and hirings, H, is the number of people who leave unemployment to take up a job. Therefore, we implicitly ignore another factor in out¯ows: people dropping out of the labour force. Likewise, we do not account for the hiring of people already in the workforce. Figure 9.1 provides a detailed picture of our data properties. It is apparent that the assumption of normally distributed errors is valid for the (log) hazard rate of unemployment. Statistically, this is veri®ed by the third and fourth moments of distribution, skewness and kurtosis respectively. In our sample, skewness is 0.05, which is close to zero, as required by symmetrical distribution (such as the normal distribution). Kurtosis is less than three (2.20), which is the standard for normal distribution. Since we have a kurtosis problem we performed the Jarque± Berra test, which tests whether a series is normally distributed. The

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 165 12

Series: HZUNT Sample 1979:2 1996:4 Observations 71

10

Mean Median Maximum Minimum Std. Dev. Skewness Kurtosis

8 6 4 2

–2.075246 –2.093068 –1.078532 –2.916623 0.478699 0.052967 2.203771

Jarque–Berra 1.908723 Probability 0.385058

0 –3.0 –2.8 –2.6 –2.4 –2.2 –2.0 –1.8 –1.6 –1.4 –1.2 –1.0

Figure 9.1 Hazard function and Spanish unemployment

statistic in our case is 1.90, and is distributed as chi-square distribution with two degrees of freedom. Therefore, we accept the null hypothesis of normality. 9.3.2

Econometric considerations and main ®ndings

To estimate the preceding four models we need to add a stochastic error term ± "  i.i.d. N (0, 1) ± to account for optimisation mistakes or the myopic decisions of agents. Equations 9.6, 9.7 and 9.8 are estimated using the ordinary least squares (OLS) technique. We performed several parameter tests, for instance checking the CRTS assumption using the Wald test. Moreover we thoroughly investigated these equations for autocorrelation and heteroskedasticity biases. The GARCH-M model cannot be estimated by OLS since the associated parameters must be determined jointly, which might involve crossparameter restrictions that OLS could not account for. Hence maximum likelihood estimation appears to be more appropriate. Since the errors are conditionally normally distributed (since the variance is nonlinearly formulated, the likelihood function must be estimated using iterative algorithms (such as the Marquardt algorithm). Our strategy for this model is to design a parsimonious model free of autocorrelation and heteroskedasticity problems, starting off from a general hazard function. Using the likelihood ratio test we reduce the lag length (i, j) to a GARCH(2,2)-M (the testing-down results are not presented here). The resulting conditional squared residuals should improve upon the normality statistics of the original series presented in Figure 9.1.

166 Spanish Unemployment ± a Hiring Function Approach

Table 9.1 summarises our main ®ndings from modelling the Spanish hazard rate of unemployment for the period 1979 (second quarter) to 1996 (fourth quarter). Column 1 presents the estimated coef®cients for the unrestricted equation (9.6). The coef®cients of U (a2 ) and V (a1 ) are 0.615 and 0.397 respectively. They are highly signi®cant, showing a tstatistic of 6.89 and 4.53 respectively. Thus, the share of unemployment in total hirings is 61 per cent whereas the share of vacancies is 39 per cent. These results are consistent with those of Layard et al. (1991), who estimated the male unemployment out¯ow rate in Britain from 1969 to 1988 and obtained parameter values of 0.3 for vacancies and 0.7 for unemployment. For the United States, Blanchard and Diamond (1989) obtained rather different coef®cient estimates: a1 = 0.6 and a2 = 0.4. This might be because they used a different dependent variable ± wheareas we de®ne hirings as the number leaving unemployment (as in Layard et al., 1991), theirs is total hirings, including the hiring not only of unemployed people but also employed people and people who were previously outside the labour force. The next step is to test for constant returns to scale. The null hypotheses of constant returns to hirings (a1 + a2 = 1) could not be rejected (2 …1† = 0.05923, with a probability of 80 per cent that we have not committed a type 1 error). This is consistent with what the theory usually assumes and with the results of Blanchard and Diamond (1989), where the hypothesis of constant returns to hirings was not rejected. Studies of Europe provide rather controversial results. Layard et al. (1991) rejected the CRTH hypothesis in a slightly different speci®cation. This could be because they included (endogenised) the effectiveness of the unemployed, c, in their model. Pissarides (1986) tested the requirement that the ¯ows out of unemployment are characterised by increasing returns in vacancies and unemployment. He argues that if this is the case, a multiple equilibria should arise. However, in most of his regressions he found CRTH, and in some cases decreasing returns. Burda and Wyplosz (1993) estimated the hiring function for France, Germany, Spain and the UK. The CRTH hypothesis was rejected at the 5 per cent and 1 per cent level for Spain and France respectively. Column 2 of Table 9.1 reports the parameter estimates with the CRTH preimposed (Equation 9.7). The share of vacancies in total hirings is 40 per cent and the calculated unemployment share is 60 per cent. Thus, unemployment seems to outweigh vacancies in explaining hirings. It is important to comment on some statistical properties of our results. The above regressions are potentially subject to simultaneity bias. Despite the fact that the estimated disturbance term is negatively correlated, it

Table 9.1 Regression results Parameters

Eq. 9.6 unrestricted

Eq. 9.7 (CRTH)

Eq. 9.8 (with trend CRTH)

Eq. 9.9 (GARCH-M)

±1.541 (±1.207) 0.397 (4.535) 0.615 (6.89) ±

±4.031 (±6.696) 0.403 (3.397) 0.5971 ± ±

±4.1935 (±136.1323) 0.14631 (11.18847) 0.8551 ± 0.02834 (35.49264) 2.5452 (3.9044) 0.000375 (0.44711) 0.7028 (2.827498) 0.10307 (1.540313) 0.6278 (8.086246) ±0.1699 (±2.54632) 0.911 ± 6.4388 ± 1.069574



±

±

±3.831 (±9.865) 0.244 (3.004) 0.7561 ± 0.016 (9.314) ±

b0

±

±

±

b1

±

±

±

b2

±

±

±

c1

±

±

±

c2

±

±

±

0.141 64.1813 ± 63.82 4.45

0.619 60.70 ± 53.887 1.983

a0 a1 a2 a3

R2 LM (5) LM(10)-Q2 ARCH (5)3 Jarque±Berra4

0.666 65.482 ± 48.359 3.0620

167

Notes: 1. These parameters are not estimated but calculated as 1 ± a1 2. The Langrangian multiplier test for higher order of autocorrelation (5 and 10 lags) 3. The Langrangian multiplier test for higher order of heteroskedasticity (5 lags) 4. This is a normality test

168 Spanish Unemployment ± a Hiring Function Approach

may be the sum of a large, negatively correlated measurement error and a positive, serially correlated disturbance term standing for omitted factors in the hiring function. In this case, the estimated coef®cients for vacancies and unemployment are likely to be biased downwards, as a positive disturbance to hiring leads, all other things being equal, to a decrease in unemployment and vacancies in the following quarter, thus a negative correlation between the hiring disturbance and both unemployment and vacancies. For instance, DW is 0.055 and the Breusch±Godfrey LM test for high autocorrelation (up to ®ve lags) gives 64.18813, which decisively rejects the null hypothesis of no autocorrelation. Likewise, the ARCH test (up to ®ve lags) provided a ®gure of 63.82331, which is also substantially above the critical value of the chi-square distribution (the degrees of freedom are equal to ®ve). Column 3 of Table 9.1 updates the restricted equation by adding a time trend to the hazard rate of unemployment speci®cation (Equation 9.8). Our results are not consistent with the literature. For example Blanchard and Diamond (1989), Layard et al. (1991) and Burda and Wyplosz (1993) report a negative time trend, which is supposed to stand for factors that might reduce the rate of hirings over time, such as search effectiveness. Furthermore, we found a positive time trend coef®cient, implying that we are omitting some variables that may have a positive effect in hirings. Whereas the size of the coef®cient (0.016) is fairly low, it appears to be highly signi®cant (the t-value is equal to 9.31). Moreover, the ®t of the regression clearly improves as we include the time trend since R2 rises from 0.14 to 0.61, showing that the time trend is an important variable that should not be omitted. The share of the V/U ratio has decreased to 24 per cent, implying an increase in the weight of unemployment in explaining hirings. Since the above models (Equations 9.6, 9.7 and 9.8) persistently exhibit autocorrelation and heteroskedasticity problems, these speci®cations are not acceptable (our ®rst attempt to correct these anomalies was to run an autoregressive model, but unfortunately it could not account for the heteroskedasticity biases). Column 4 of Table 9.1 presents the GARCH-M (2,2) results of the maximum likelihood estimation. This model is free of autocorrelation and heteroskedasticity problems. The LM test gives a statistic of 6.4, which allows us to accept the null hypothesis of no autocorrelation. Moreover, R2 has risen to 0.91, implying an almost perfect ®t, since 91 per cent of the variation in the hazard rate of unemployment is explained by our model. We observe that the V/U coef®cient is even lower than in the previous

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 169 0.04

0.6

0.03

0.4

0.02

0.2

Recession

Strong growth of the Spanish economy

0.01

0.0 Benefits reform

0.00

–0.2 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

Conditional variunce of H/U

Figure 9.2

Growth of H/U (yearly)

The hazard/unemployment ratio and the Spanish economy

regressions (it has decreased from 0.24 to 0.14).17 This implies that unemployment dominates vacancies, and explains most of the `matchings' that take place in the economy. This might explain the strong positive effect (2.54) of the spread of the hazard rate of unemployment on the probability of a jobless person a ®nding a job. Explicitly, it suggests that greater variability in the H/U ratio implies greater ¯exibility in market conditions via an increase in circulation.18 Figure 9.2, which plots the conditional variance along with the yearly growth of the job ®nding rate, should help us to visualise the above discussion. Several points in Figure 9.2 are worthy of note. First, the expected mean of the exit rate and its variance follow a very similar path. This seems to support the argument of ¯exibility, since, when the variance rises, the hazard rate of unemployment rises. Second, the variance increases at times of government intervention (1984, 1994) and improved economic activity (1987±91). Similarly, it drops with recession (1991±94). Closer examination of these ®ndings would be necessary if one wished to determine the underlying relationship between, for instance, the unemployment growth rate and conditional variance of the exit rate. 9.3.3

Discussion of the results and `stylised' facts

To assist the discussion we plot the evolution of hirings (in terms of hazard rates) over time (Figure 9.3). One main feature to be observed is

170 Spanish Unemployment ± a Hiring Function Approach

0.04

0.03

0.02

0.01

0.00 1980

1982

1984

Hazard rate U

1986

1988

1990

1992

1994

1996

Hazard rate of V

Figure 9.3 Hazard rate of unemployment

that both hazard rates (or the overall/average hazard rate) seem to rise gradually after 1984. This is due to the reduction of ®ring costs and severance payments via the introduction of the ®xed-term contracts, which increased ®rms' willingness to expand at a comparatively lower employment cost. As a result, higher labour turnover took place, induced by an increase in search intensity on the part of ®rms and the unemployed. The rise in the hazard rates could also re¯ect the effects of national agreements that took place between 1978 and 1986 to restrict wage increases. This should also have increased labour market ¯exibility. Another important feature is that after 1984 the hazard rate of unemployment, which is the probability of an unemployed person ®nding a job, rises above the hazard rate of vacancies (the probability of a job vacancy being ®lled). This is highly consistent with our results, which persistently report a very low V/U ratio. For instance, the last regression gives a hazard rate of vacancies of 0.14. This implies that ®rms are having trouble recruiting workers, since the average duration of a vacancy is higher than the average duration of unemployment. The rate of hirings appears to be signi®cantly determined by the supply side of the labour

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 171

market, rather than by the demand side (see, Blanchard and Diamond, 1986), which is contrary to what an aggregate demand model would predict. Several pieces of evidence suggest that the introduction of ®xed-term contracts has added ¯exibility to ®rms' hiring/®ring decisions, and has led to an increase in labour turnover. The ratio of in¯ow of registered job demands (at employment of®ces) to employment increased from 3.4 during the 1980±84 recession to 6.5 during the 1991±92 recession. This is re¯ected in our results by an increasing hazard rate of unemployment induced by high search intensity on the part of the unemployed. An important variable affecting search intensity is the ease with which a person expects to ®nd a job. If the availability of jobs increases, he or she expects to ®nd a job more easily for each extra hour spent searching, so search intensity increases. One might wonder why the introduction of ®xed-term contracts did not affect the hazard rate of vacancies as much as the hazard rate of unemployment, given that, theoretically, changes in employment legislation should have had a greater effect on ®rms `choosiness' than on job search intensity. In order to understand this, it is important to take into account that the new legislation had an adverse effect in that it increased the power of the `haves' (the insider±outsider hypothesis seemed to be in effect). Empirical evidence suggests that the power of insiders has grown since 1984 as a result of the introduction of ®xed-term contracts, which have helped to insulate permanent workers (the insiders) from the effect of cyclical ¯uctuations. Generous job protection for those with permanent contracts has crowded new entrants out of the labour market and fostered a vicious circle of rotating short-term jobs.19 As a result, ®rms have reduced the supply of vacancies since the insiders have forced wages above the level of workers' marginal productivity. This has reduced the probability of a given vacancy being ®lled, since employers have become more choosy about whom they employ. The induced reduction in labour demand has led to high and persistent unemployment, off-setting the higher labour market ¯exibility induced by the introduction of ®xedterm contracts. The above discussion suggests that, in line with Figure 9.3, the introduction of ®xed-term contracts in 1984 has had an adverse effect on hirings due to the insider±outsider problem. On the other hand, the reform of the bene®t system has had a signi®cant effect on the hazard rates of vacancies and unemployment. For instance, the in¯uence of the insurance system on job search intensity became obvious towards the end

172 Spanish Unemployment ± a Hiring Function Approach

of 1994. By then bene®ts had been made taxable, thus removing one of the largest distortions in the system. Out¯ow rates were further boosted by the 1994 abolition of the Ordenanzas Laborales and the relaxation of restrictions on the transfer of workers between regions, which contributed to increased labour mobility and ¯exibility. As can be seen in Figure 9.3, both hazard rates rose sharply after 1992, re¯ecting the positive effect of the insurance system reforms. However, whereas the hazard rate of unemployment continued to rise until the end of our sample period, the hazard rate of vacancies fell sharply after 1995. This could be due to the adverse effects of the insider±outsider problem.20 Further research is required to provide an explanation of this empirical observation.

9.4

Conclusions

In this chapter we have used the hiring function as our main analytical tool to study the evolution of Spanish unemployment over time. We consider that the coexistence of unemployment and vacancies in spite of high gross market ¯ows is a relevant feature of the Spanish labour market that could be appropriately analysed using a matching function approach. Our methodology allowed us to decompose the job ®nding rate into two probabilities: the hazard rate of unemployment and the hazard rate of vacancies. The main difference between our methodology and that of the mainstream was the adoption of the GARCH-M (2,2) model to estimate the Spanish hazard rate of unemployment. This approach helped us to identify the following empirical points:

. Our ®ndings for the period 1979±96 strongly suggest that the various

government interventions in the labour market should have affected not only the expected mean of the exit rate, but also its expected variance. Unfortunately, we were unable to pin down the policy instruments and, therefore, to evaluate their precise effects on the expected mean. However, having modelled the conditional variance and extended it over time, we recognise the importance of conducting a closer investigation of the variance of the exit rate. . The GARCH-M model also proved that the variance of the hazard rate of unemployment can be used as yet another exogenous shift variable in explaining the expected mean of the Spanish exit rate. Indeed, we estimated that its elasticity with respect to its variance is very signi®cant (2.54 per cent). Hence a 1 per cent rise in variance will cause a 2.54 per cent increase in the probability of an unemployed person ®nding a job. Here, this elasticity is interpreted as an indicator

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 173

of labour market ¯exibility, since greater dispersion of the exit rate is associated with a higher hazard rate of unemployment. Although a bold statement, in our mind this is a plausible explanation that requires further investigation. For instance, one could associate the exit rate variance with GDP growth or the unemployment growth rate. . The Cobb±Douglas type equations (Equations 9.6, 9.7 and 9.8) of the matching function overstate the importance of vacancies in explaining total Spanish hirings, as our GARCH-M model indicates. For instance, from almost 0.4 in the unrestricted model (Equation 9.6), its elasticity falls to 0.14 in the former model. This is primarily due to proper speci®cation of the variance structure of the exit rate. In conclusion, we believe that further investigation should focus on changes in the variance of the probability of someone obtaining a job over time. For example, do policy interventions such as the introduction of ®xed contracts have a permanent or a temporary effect on the labour market? If the elasticity of the hazard rate with respect to its variance is a proxy of ¯exibility, which macroeconomic fundamentals seem to alter it? Another interesting line of enquiry is how this measure of ¯exibility varies across the Spanish regions. All these matters are left for later investigation. Notes 1. According to Burda and Wyplosz (1993). 2. Dolado and Gomez (1996) estimate that gross job ¯ows in the manufacturing industry between 1984 and 1992 (excluding ®rm openings and bankruptcies) amounted to 7 per cent of annual employment. 3. In this chapter we do not consider the stationarity of hirings, vacancies and unemployment as these variables are not expected to be stationary. Indeed, using the ADF test we could not reject non-stationarity. This matter deserves proper consideration but the literature cited here ignores it (apart from Burda and Wyplosz, 1993). Perhaps a fully dynamic structure is necessary. Nevertheless our results are not invalid. 4. See Note 1 above. For further discussion of changes in unemployment in relation to in¯ows and out¯ows, see Layard et al., 1994. 5. To see how one can account for it in this model, see Pissarides (1986). 6. See, Hall (1986). 7. See, among others, Blanchard and Diamond (1989) for the United States, and Burda and Wyplosz (1993) for Europe. 8 9 8 9 H h…cU; gV† U V V : ; > ; ˆ cgh> :1; > ; ˆ ˆ cg h> 8 U U U U U 9. The exogeneity of V/U is not necessarily a correct assumption. For instance, Blanchard and Diamond (1989) and Burda and Wyplosz (1993) endogenise it by using instrumental variable estimation techniques.

174 Spanish Unemployment ± a Hiring Function Approach 10. See, Layard et al. (1994) 11. These ®gures should be weighted with the number of unemployed people to obtain a more meaningful picture. That is, an interesting intercountry comparison should include calculation of each government's expenditure per unemployed person on constant purchasing parity units. 12. Other important factors, for example the bene®t reform, contributed to this change. Since 1994, the assistance bene®t has been reduced on both replacement ratio and eligibility basis. Assistance bene®t is mostly paid to those who have exhausted their insurance bene®t. It is also paid to those who have contributed for six to twelve months. Those who have only worked for three months but have family responsibilities also receive assistance bene®t. The 1994 reform restricted the notion of `family responsibilities' to mean spouse and dependent children. Moreover, the amount of the cash bene®t has been reduced from 100 per cent of the SMW to 75 per cent of the SMW. 13. Until 1984, by law the employment relationship was a permanent one. Workers could be dismissed only under certain limited conditions, and dismissals typically involved sizeable redundancy payments. Even after the introduction of temporary contracts, many workers remained under permanent contract and were thus isolated from external labour market developments. This diminished somewhat with the 1994 and 1997 reforms. 14. See, Linbeck and Snower (1988). 15. A number of recent studies have attempted to measure the importance of insiders in Spanish labour market ¯exibility and wage determination. 16. In that era, Spain was governed by the UCD (Centre Right Union), which to some extent re¯ected the slow transition towards democracy. 17. A 1 per cent increase in the V/U ratio would only increase the H/U ratio by 0.14 per cent. 18. However, during the ®rst years of the post-Franco era, the variability of the hazard rate of unemployment did not imply greater ¯exibility in the labour market and, therefore, the strong positive effect of the volatility breaks down. Since the big spread during this period is due to adverse political and labour institutions, the greater volatility should be associated with a negative rather than a positive job-®nding growth rate. 19. See, Financial Times, 10 April 1997, p. 2. 20. It could be argued that the smaller replacement ratio induced by the 1993 reform increased the search effectiveness of the unemployed. This should increase the probability of an unemployed person ®nding a job. But the job offers available are likely to be of a ®xed-term nature, which again reinforces the power of insiders and reduces the probability of a given vacancy being ®lled.

References Blanchard, O. J. and P. Diamond (1989) `The Beveridge Curve', Brookings Papers on Economic Activity, vol. 1, pp. 1±60. Blanchard, O. J. and P. Diamond (1990) `The aggregate matching function', Productivity, Growth and Unemployment: Essays in Honour of Robert Solow's Birthday (Cambridge, MASS: MIT Press). Blanchard et al. (1995) `Spanish Unemployment: Is there a Solution?' (London: CEPR).

Isabel Figuerola-Ferretti and Yannis G. Paraskevopoulos 175 Burda, M. and C. Wyplosz (1990) `Gross Labour Market Flows in Europe: Some Stylised Facts', Discussion Paper, series no. 439 (London: CEPR). Burda, M. and C. Wyplosz (1993) `Gross Labour Market Flows in Europe', Discussion Paper series no. 868 (London: CEPR). Burns, T. (1997) Financial Times, 10 April, p. 2. Coles, Melvyn G. and E. Smith (1996) `Cross Section Estimation of the Matching Function: Evidence from England and Wales', Econometrica, vol. 63, pp. 589±97. Dolado, J. and R. Gomez (1996) The relationship between unemployment and vacancies in Spain: Aggregate and Structural shocks (Spain: Bank of Spain). Layard R., N. Nickell, and R. Jackman (1991) Unemployment, Macroeconomic Performance and the Labour Market, (Oxford: Oxford University Press). Pissarides, C. (1986) `Unemployment and vacancies in Britain', Economic Policy, vol. 3, pp. 499±559. Pissarides, C. (1990) Short Run Equilibrium Dynamics of Unemployment, Vacancies and Real Wages: Equilibrium Unemployment Theory (Oxford: Basil Blackwell).

10

Economic Growth, Labour Effort and Fertility Choice in Greece: Evidence and Implications for Economic Integration1 George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou

10.1

Introduction

Many economists believe there is a relationship between family behaviour and economic growth. According to Becker (1988), `family behaviour is active, not passive, and endogenous, not exogenous. Families have large effects on the economy, and evolution of the economy greatly changes the structure and decisions of families.' Malthus (1798) was the ®rst to support the idea that population growth is a potential determinant of output growth, and Adam Smith was the ®rst to realise that the relevant measure of growth is output per capita and not aggregate output. Extending Malthus' work, researchers such as Blaug (1962), Schumpeter (1954), and Smith (1937) developed the so-called `classical' model. They adopt the view that economic growth is determined exogenously and population growth must adjust to it in the long run. However, they argue that, in the short term, there is a positive relationship between deviations of per capita income and the rate of economic growth from their long-term values. An extension of the classical model is the neoclassical growth model (Solow, 1956). According to this model, economic growth is an endogenous variable that depends on population growth, while fertility remains an exogenous variable. Becker (1960, 1973), in his pioneering studies, supports the notion that fertility growth is an endogenous variable in the economic system, and he develops a theoretical framework to explain that the relationship between the two variables depends on a number of socioeconomic factors, such as the incentive to have children, the `quality' of children, the ef®ciency of private capital markets, and intergenerational transfers within the family. More recently, economists 176

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 177

such as Barro and Becker, (1989), Becker (1988, 1992), Becker and Barro (1988), Becker et al. (1990), Ehrlich (1990), Ehrlich and Lui (1991) and Wang et al. (1994), based on microfoundations of economic theory, have treated both population growth and income growth as endogenous variables in an attempt to develop a coherent model of economic growth and explain the process of dynamic economic growth. A major trend in the literature today is the development of theoretical dynamic models that treat population growth and development as endogenous variables that are simultaneously determined, rather than separate outcomes of different economic systems. Since the mid-1970s, most of the work on endogenous population and economic growth has been theoretical. Only a few empirical studies (for example, Ehrlich and Lui, 1991; Wang et al., 1994) have examined the effects of population growth and fertility on economic growth, mainly for the United States. However, there is limited empirical evidence for the countries of Western Europe and some developing countries (for example, Brander and Dowrick, 1993; Winegarden and Wheeler, 1992). This chapter conducts an empirical investigation of the dynamic interaction between fertility choice and economic growth, in order to explain systematically the process of economic development and the associated demographic changes. The relevance of endogenous fertility choice is investigated, and its dynamic response to structural shocks is estimated for the case of Greece. In addition, the chapter extends previous empirical work, mainly for the United States, to a medium-sized country, thus enriching our understanding of the development process and the low fertility growth rate in Greece. The results obtained have important policy implications for the consolidation of the Greek social security system, and especially for the EU convergence process in respect of Greece. Since the beginning of the 1990s, Greece has been pursuing a mediumterm adjustment programme (the Convergence Programme 1993/4±99), aimed at achieving nominal convergence towards the other EU member states. Attainment of this objective is a precondition for Greece's participation in the single European currency. Also, the success of the European integration process requires real convergence of the EU economies; that is, a reduction in the income inequality that exists among the EU member states. Real convergence of the Greek economy can only be attained in the medium to long term and presupposes the achievement of GDP growth rates that are appreciably higher than the EU average. Economic and monetary union (EMU) requires a high degree of price stability, a reduction of public de®cit and debt-to-GDP ratios to the

178 Economic Growth, Labour Effort and Fertility Choice

Figure 10.1 Live births per 1000 inhabitants, Greece, 1960±95

reference values laid down in the Maastricht Treaty, exchange rate stability and the convergence of long-term interest rates.2 These targets are outlined in the Greek convergence programme. An additional goal of the convergence programme is to stabilise social security fund surpluses, and further growth of surpluses is not anticipated in the short term.3 Furthermore, the widening of the primary de®cit of the major social security and welfare funds underlines the chronic and structural nature of the problems faced by the Greek social security system, and the efforts required for its consolidation. In the medium term, problems with the social security system will be aggravated by the transition of many of the large funds to a stage of maturity and the upward trend in health costs. Also, as recognised by the majority of empirical studies,4 adverse developments in demographic factors, such as low fertility rates and the ageing of the population, are putting further pressure on the Greek social security system. Figure 10.1 shows the total fertility rate in Greece for the period 1960± 95, de®ned as the number of births per 1000 inhabitants. It is clear that there has been a decline in fertility, especially from 1980 onwards, during which time the economy developed at a slow pace. Moreover, the fertility

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 179

rate in Greece, especially since 1985, has been lower than in most of the EU countries. This important demographic issue seems not to have received appropriate attention. Policy measures to reverse the downward trend of the fertility rate will facilitate the ®nancing and resolve the chronic de®ciencies of the Greek social security system. The purpose of this chapter is fourfold. First, using annual data for the period 1960±95, we test the stationary properties of the data and the order of integration. Second, the Johansen maximum likelihood technique is applied to search for cointegration among fertility choice, employment and output. The use of the Johansen technique controls the endogeneity and the complicated short-term dynamics, while focusing on long-term relationships (cointegration) among non-stationary variables. Third, the vector error-correction model, variance decomposition analysis and impulse response functions are used to investigate the response of employment, fertility and output growth to employment, fertility and output shocks. This is aimed at testing the proposition that fertility choice should not be considered exogenous to labour market developments or to the growth process. Finally, since fertility choice is considered an endogenous variable that also affects the growth process, policies aimed at increasing the low fertility rate are outlined and discussed. The chapter proceeds as follows. Section 10.2 deals with methodological issues and the data used in the empirical analysis. Section 10.3 presents the empirical results, and in Section 10.4 the conclusions of the analysis are summarised and the policy implications are discussed.

10.2

Methodological issues and data

In the empirical analysis, we test for the endogeneity of fertility growth and the responses of fertility, employment and output to fertility choice, labour market and output disturbances in order to capture the short-term dynamics of the variables following the theoretical framework of the `neoclassical model'. The three macroeconomic variables employed in the empirical analysis are the growth rates of labour effort, fertility choice and output. Testing for the existence of statistical relationship among the variables is conducted in four steps. The ®rst step is to verify the order of integration of the variables, since the causality tests are valid if the variables have the same order of integration. Standard tests for the presence of a unit root based on the work of Dickey and Fuller (1979, 1981), Perron (1988), Phillips (1987), Phillips and Perron (1988)5 and Kwiatkowski et al. (1992)6 are used to investigate the degree of integration of the variables used in the empirical analysis.

180 Economic Growth, Labour Effort and Fertility Choice

The combined use of the three tests employed to investigate the degree of integration of the series may result in four possible outcomes. First, rejection by the ADF and PP statistics and non-rejection by the KPSS test would provide strong evidence of stationarity. Second, non-rejection by both ADF and PP and rejection by the KPSS would provide a strong indication of I(1). Third, non-rejection by all tests would suggest that the data is not suf®ciently informative on the long-term characteristics of the series. Fourth, rejection by all tests would indicate that the series is neither an I(1) nor an I(0) process. The gain from testing both stationarity and non-stationarity using the three tests and comparing their outcomes is that the inference is not uncertain. More speci®cally, the inference is either stationarity, non-stationarity or we do not know which is preferable to being uncertain. The second step involves testing for cointegration using the Johansen maximum likelihood approach (Johansen, 1988; Johansen and Juselius, 1990, 1992). The Johansen±Juselius estimation method is based on the error correction representation of the VAR(P) model with Gaussian errors. This approach has several advantages over the Engle±Granger (1987) technique employed in most recent empirical studies. First, the Johansen± Juselius method tests for all the cointegrating vectors between the variables. These tests are based on the trace statistic test and the maximum eigen-value test of the stochastic matrix of equation. Second, it treats all variables as endogenous, thus avoiding an arbitrary choice of dependent variable. Third, it provides a uni®ed framework for estimating and testing cointegrating relations within the framework of a vector errorcorrection model. The third step involves utilisation of the vector error-correction model (VECM) and testing for exogeneity. Engle and Granger (1987) show that in the presence of cointegration there always exists a corresponding error-correction representation, which implies that changes in the dependent variable are a function of the level of disequilibrium in the cointegrating relationship, captured by the error-correction term (ECT), as well as changes in other explanatory variables. Thus, through the ECT, the VECM model provides an additional way of examining Granger causality, ignored initially by the Granger±Sims tests. The F-test, applied to the joint signi®cance of the sum of the lags of each explanatory variable, and the t-test of the lagged error-correction term will imply statistically the Granger exogeneity or endogeneity of the dependent variable. The dependence of the ECT is referred to as weak exogeneity, while strict exogeneity is dependent on the sum of joint lagged differenced variables.

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 181

However, the above tests do not provide an indicator of the dynamic properties of the system and do not measure the relative strength of the Granger causal chain or the degree of exogeneity among the variables beyond the sample period. Therefore, in the fourth step of the empirical analysis, we attempt to investigate these properties of the system, estimating the variance decomposition and the impulse response function. The purpose of the investigation is to establish how each variable responds to shocks from the other variables in the system. The forecast-error variance decomposition analysis reveals information about the proportion of the movements in the sequence due to its `own' shocks versus shocks to other variables. If the shocks do not explain any of the forecast error variance of one macroeconomic variable, Yt in all forecast horizons, then Yt is an exogenous variable. At the opposite side, if shocks can explain all the forecast error variance of Yt at all forecast horizons, then Yt is an entirely endogenous variable. The impulse responses provide an estimate of the response of a variable in the case of innovation in another variable. Plotting the impulse response functions is a practical way to explore the response of a variable to a shock immediately or with various lags. The empirical analysis is carried out using annual data for the period 1960±95. The labour effort variable (HOURS) is total weekly hours worked in manufacturing (in establishments employing ten persons or more), the fertility choice variable (FERT) is the live birth rate per 1000 inhabitants and the output variable (GDP) is real GDP. The variables HOURS and GDP are obtained from the OECD's Main Economic Indicators, while data on the fertility rate are taken from the Statistical Yearbook of the National Statistical Service of Greece. All variables are expressed in logarithmic form (LHOURS, LFERT, LGDP).

10.3 10.3.1

Empirical results Unit root tests

Table 10.1 presents the ADF, PP and KPSS tests for the three variables (employment, fertility rate and real GDP) used in the analysis in levels and ®rst differences. The ADF statistic suggests that all variables are integrated of order one, I(1), whereas the ®rst differences are integrated of order zero, I(0). Therefore, the hypothesis that the time series contain an autoregressive unit root is accepted in all cases. Although employing the Phillips±Perron test gives different lag pro®les for the various time series and sometimes lowers the level of signi®cance, the main conclusion is qualitatively the same as reported above by the Dickey±Fuller tests. In

182

Table 10.1 Tests of the unit roots hypothesis Augmented Dickey±Fuller Variable

I

LHOURS ±1.23 LGDP ± LFERT 0.49
LHOURS ±
LGDP ±
LFERT ±

q

II ±1.11 ±1.38 ±2.38 ±5.25** ±5.08** ±5.50**

0.82 14.41** 3.76 13.86** 13.02** 15.66**

Phillips±Perron k

T

0 0 2 0 1 0

±1.25 ± 0.99 ± ± ±

Tt ±1.19 ±1.39 ±1.32 ±5.25** -6.16** ±5.51**

KPSS p

0.89 15.69** 2.01 13.84** 19.16** 15.69**



1=1

1.234** 1.737** 1.759** 0.173 1.247** 0.305

T

0.239** 0.462** 0.352** 0.130 0.085 0.067

Notes: The relevant tests are derived from the OLS estimation of the following autoregression for the variable involved: k X i
xt 1 ‡ ut
xt ˆ 0 ‡ 1 …Time†t 2 xt 1 ‡ iˆ1

 0.848** 1.202** 1.210** 0.153 0.982* 0.246

1=2

r

0.170* 0.323** 0.249** 0.116 0.097 0.056

…10:1†

I is the t-statistic for testing the signi®cance of 2 when a time trend is not Equation 10.1 and TT is the t-statistic for testing the signi®cance of 2 when a time trend is included in Equation 10.2. The calculated statistics are those reported in Dickey±Fuller (1981). The critical values at 5 per cent and 1 per cent for n = 50 are ±2.93 and ±3.58 for T and ±3.5 and ±4.15 for TT , respectively. The 3 statistics are computed as the F-test for testing the hypothesis t = ±2 = 0. The critical values at 5 per cent and 1 per cent for n = 50 are 6.73 and 9.31 respectively (Dickey and Fuller, 1981, Table 6, p. 1061). When the trend is signi®cant, we retest for the presence of a unit root (2 = 0) using the standardised standard normal distribution (Enders, 1995). The lag length structure of i of the dependent variable xt is derived using a recursive procedure in the light of a Lagrange multiplier (LM) autocorrelation test (for orders up to two), which is asymptotically distributed as chi-squared distribution and the value of the t-statistic of the coef®cient associated with the last lag in the estimated autoregression. The critical values for the Phillips±Perron unit root tests are obtained from Dickey±Fuller (1981).  and T are the KPSS statistics for testing the null hypothesis that the series are I(0) when the residuals are computed from a regression equation with only an intercept and intercept and time trend, respectively. The critical values for t and T at 5 per cent are 0.463 and 0.146 and at 1 per cent are 0.739 and 0.216 respectively (Kwiatkowski et al., 1992, Table 1). ** and * indicate signi®cance at the 1 per cent and 5 per cent levels respectively.

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 183 Table 10.2 Johansen±Juselius cointegration test: employment, fertility choice and real output, 1960±95 VAR = 3 variables: LHOURS, LFERT, LGDP Null

Alternative

Eigenvalue

Critical values

Maximum eigenvalues: r=0 r<1

r=1 r=2

24.34** 7.83

95% 20.96 14.07

90% 18.6 12.07

Trace statistic r=0 r<1

r>1 r=2

36.48** 12.14

95 % 29.68 15.41

90 % 26.79 13.33

Notes: r indicates the number of cointegrating relationships. Maximum eigenvalue and trace test statistics are compared with the critical values from Johansen and Juselius (1990). ** and * indicate rejection of the null hypothesis at the 95 and 90 per cent critical values respectively.

particular, the Phillips±Perron test, based on the 5 per cent and 1 per cent critical values, supports the hypothesis that all series contain a unit root. Thus, both tests support the unit root hypothesis in all time series. Finally, the KPSS statistics test for lag-truncation parameters one and four (l = 1 and l = 2)7 , since it is unknown how many lagged residuals should be used to construct a consistent estimator of the residual variance. The KPSS test rejects the null hypothesis of level and trend stationarity for both lag truncation parameters. The KPSS statistic does not reject the I(0) hypothesis for the ®rst differences of the series at different levels of signi®cance. Therefore, the combined results from all the tests (ADF, PP, KPSS) suggest that all the series under consideration appear to be I(1) processes. 10.3.2

Cointegration analysis and error correction model

Table 10.2 summarises the results of the cointegration analysis using the Johansen maximum-likelihood approach, employing both the maximum eigenvalue and trace statistic. VAR = 2 is used in the Johansen estimation procedure.8 The estimation procedures assume that there is a linear deterministic trend in the variables but there is no trend term in the underlying datageneration process. Also, a constant term is not included in the estimation. The two test statistics give similar results. Both cause us to reject the null of zero cointegrating vectors in favour of one cointegrating vector at the 5 per cent level. On the basis of the results, a long-term relationship among fertility choice, labour effort and output ®nds

184 Economic Growth, Labour Effort and Fertility Choice

statistical support in the case of Greece for the period under examination. The long-term equilibrium is given by the following equation: LFERTt ˆ

4:76 LHOURSt ‡ 0:22 LGDPt

In addition, we employ the likelihood ratio test, which follows the chisquare distribution to test the hypothesis that the estimated coef®cients are statistically signi®cant. Estimates of the test indicate that, jointly, the estimated coef®cients are different from zero at the 5 per cent level of signi®cance (2 = 6.32). Having veri®ed that the variables are cointegrated, vector errorcorrection models (ECM) can be applied. The lagged residuals from the cointegrating regression with the appropriate number of lags are included in the Granger causality test structure. The lag length structure depends on the restricted error-correction models. The restricted error-correction models pass a series of diagnostic tests, including serial correlation based on inspection of the autocorrelation functions of the residuals as well as the reported Lagrange multiplier. Table 10.3 reports the ®ndings for causality, based on error-correction models of fertility choice, labour effort and output containing one error-correction term measuring the proportion by which the long-term imbalance in the dependent variable is corrected in each short-term period. The size and statistical signi®cance of the error-correction term indicate the extent to which each dependent variable has the tendency to return to each long-term equilibrium. Estimates of the parameters show that the error-correction term measuring the long-term disequilibrium is signi®cant only for the fertility choice and output equations. This implies that fertility and output move to restore equilibrium and take the brunt of the shock to the system. Conversely, in the labour effort equation the error-correction term is insigni®cant, which indicates that the long-run adjustment to correct disequilibrium primarily occurs through the fertility and output channel and not through the labour effort channel. The t-tests of the errorcorrection term and the F-test of the independent variables indicate that fertility and output are mainly endogenous variables, while labour effort is affected by the fertility choice variable. 10.3.3 Variance decomposition analysis and impulse response functions Having veri®ed the long-term relationship, exogeneity, endogeneity and direction of causality among the three variables, we proceed to the variance decomposition analysis and estimate the impulse response functions in order to investigate the dynamic properties of the system.

Table 10.3 Summary causality results based on vector error-correction models of fertility choice, labour effort and output (test of restrictions, F-statistics) Short-run dynamics
LHOURS
LFERT
LGDP


LFERT 4.80** ± 2.91*


LHOURS ± 5.81** 0.81

Joint (short-run dynamics and ECT)
LGDP 0.18 2.88* ±

ECT coef®cient ±0.0003 ±0.001*** 0.0015***


LFERT and ECT 2.76* ± 5.53***


LHOURS and ECT ± 9.66*** 5.47***


LGDP and ECT 1.74 9.73*** ±

Notes: The lagged ECT is derived by normalising the cointegrating vector on fertility choice (LFERT). In the short-run dynamics, the values in parentheses are the probabilities indicating the level of signi®cance to reject the Ho that there is no Granger causal relationship between the two variables. The coef®cients of the lagged ECTs are tested using the t-statistics testing the null hypothesis that the estimated coef®cient is equal to zero. ***,**,* indicate signi®cance at the 1 per cent, 5 per cent and 10 per cent levels respectively.

185

186 Economic Growth, Labour Effort and Fertility Choice Table 10.4 Percentage of forecast variance explained by innovations in variance decomposition of labour growth Period 1 2 4 8 12

Employment Fertility shock Output shock shock 100.0 90.8 87.9 85.5 84.7

0.0 8.4 10.7 11.3 11.4

0.0 0.8 1.5 3.3 3.9

Variance decomposition of fertility growth: 1 2 4 8 12

12.5 20.1 22.4 21.6 21.2

87.5 78.9 72.8 69.7 68.9

0.0 1.0 4.8 8.7 9.9

Variance decomposition of output growth: 1 2 4 8 12

4.4 11.5 11.4 10.6 10.3

1.4 1.2 2.8 3.8 4.5

94.2 87.3 86.9 85.6 85.1

The results of the variance decomposition for the three variables are presented in Table 10.4. The reported statistics indicate the percentage of the forecast error in each variable that can be attributed to innovations in other variables at six different time horizons: one year ahead (the shortterm period), two to four years ahead (the medium-term period) and more than four years ahead (the long-term period). The ordering used in the analysis for the Choleski factorisation is labour growth, fertility growth and output growth. The results of the variance decomposition analysis and the impulse response function provide the same conclusions regardless of the order of decomposition. The analysis of variance decomposition tends to suggest that each of the variables used in the empirical analysis can be explained by disturbances in the other variables. Speci®cally, in the short-term period 100 per cent of the variability in employment changes is explained by its own innovations, while in the medium term, 10.7 per cent of the variability in labour growth changes is explained by innovations in fertility changes and 1.5 per cent in output growth changes. Finally, in the long-term period about 84 per cent of the variability in employment changes is explained

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 187

by employment shock, 11 per cent by fertility changes and 4 per cent by output growth changes. Table 10.4 also presents the variance decomposition of fertility growth. In the short-term period, 87.5 per cent of the variability in fertility choice is explained by its own disturbances. In the long-term period, more than 21.0 per cent of the variability in fertility change is attributed to employment changes and approximately 10 per cent to technological changes. Finally, Table 10.4 presents the variance decomposition of output. In the short-term period, 4.4 per cent of output variability is attributed to fertility shock, while 1.4 per cent is due to changes in labour efforts. During the medium-term period, 11.4 per cent of output change is attributed to labour efforts, while 2.8 per cent is due to fertility changes. In the long-term period, output variations are mainly due to technological shocks, while approximately 4.5 per cent are attributed to fertility changes and more than 10 per cent to labour effort changes. An alternative way of obtaining the information on the relationship among the three variables included in the variance decomposition analysis is through the impulse response functions. Figure 10.2 presents the impulse response functions for labour growth, fertility growth and output growth from a one-standard deviation shock to employment, fertility choice and output shocks. As shown by the labour growth graphs in Figure 10.2, labour effort is mainly affected by a positive shock to labour growth in the short-term period. After an increase in the ®rst year, labour effort decreases, then rises, and the effect levels out completely after four years. In response to fertility, labour effort initially decreases due to the reallocation of time from labour activities to raising children, and this negative effect levels out in the long-term period. Finally, in response to output growth shock, labour effort increases in the second year and then returns to the initial level. The fertility growth graphs in Figure 10.2 show the response of fertility to the three different shocks. A positive shock from labour effort initially increases fertility; in the medium term, as might be expected, it decreases fertility since time is reallocated to labour; and in the long term the effect of the shock diminishes. In response to an output shock, fertility increases and stays at a higher level even at the long-term period. The empirical results indicate that fertility is an endogenous variable since it is affected by disturbances in labour and output. Finally, the output-growth graphs in Figure 10.2 show the response of output to the three alternative shocks. Output increases in response to a positive innovation in employment and technology. In the case of

188 Economic Growth, Labour Effort and Fertility Choice

Figure 10.2 Response to one standard deviation innovations ( 2 SE)

employment, the shock dies down in the long-term period, while, in the case of technological shock, the increase is sharp and raises output to a higher level. A positive shock in fertility decreases output in the shortterm period, since this type of disturbance affects labour effort and capital accumulation. However, in the long-term period the positive shock increases output growth.

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 189

Three conclusions can be drawn from the empirical analysis. First, employment shocks are responsible mainly for the variations in labour growth, fertility choice and output changes. Second, fertility choice should not be considered exogenous to the labour market or the growth process. Finally, fertility changes are partially responsible for changes in employment and output movements, indicating that a lower population growth results in slower economic growth in the long term, although in the short term lower population growth performs better than moderate population growth. These results are similar to those of Sapiro and Watson (1988), Simon (1992, ch. 2) and Wang et al. (1994) for the United States.

10.4

Conclusions

This chapter has analysed a growth model where fertility choice is treated as an endogenous variable. The statistical relationship among fertility choice, labour effort and output growth has been estimated for Greece during the 1960±95 period, thus adding to the current concepts of economic growth. The chapter ®rst examined the existence of a relationship among employment, fertility choice and output over the period 1960±95. The various unit root pretests show that the hypothesis that the variables contain an autoregressive unit root can be accepted for all variables. In addition, the empirical results, derived from the Johansen±Juselius maximum-likelihood estimation technique of cointegrating vectors, support the hypothesis that a long-term relationship exists among the three variables. The results of the empirical model (using the error-correction model estimation, variance decomposition analysis and the impulse response function) suggest that fertility should be considered as an endogenous variable to the labour market and the growth process. In particular, employment and output shocks are responsible for variations in the fertility rate. A negative employment shock causes a reallocation of time from labour effort to childbearing, thus increasing fertility growth, while at the same time an increase in fertility growth is responsible for higher economic growth in the long term. The empirical results have several important policy implications, especially for Greece (where the fertility rate is declining), given that Greece is aiming at real and nominal convergence towards other EU economies. The empirical results de®nitely suggest that the convergence process would be facilitated by suitable policy measures.

190 Economic Growth, Labour Effort and Fertility Choice

More speci®cally, reallocating time from work to leisure, thus reducing working time, would act as a negative employment shock that would have a positive effect on fertility growth. Reduced working time does not necessarily imply a reduction in output since measures could be taken to improve productivity. This process would give rise to higher growth in the medium term, facilitating real convergence and, therefore, bridging the gap between the standard of living in Greece and that in the other EU member states. Also, childbearing incentives and measures to increase provisions to families would boost fertility preference and capital accumulation, and thus accelerate output growth in the long term. These measures could well include subsidies to families with more than two children, free day-care provision, state education, paid maternity leave, pensions for mothers with a large number of children, tax reductions for families with many dependent children, and so on. Achievement of the Maastricht criteria relating to the consolidation of public ®nances is closely linked to consolidation of the Greek social security system. The proposed family policy measures would not jeopardise these objectives and could be pursued without hindering the achievement of the Maastricht targets. The family policies could be carried out more ¯exibly and ef®ciently by local authorities, social agents or voluntary organisations without burdening the general government ®nances. The state would not necessarily need to fund these policies, its only role being to provide the appropriate legal framework. Furthermore, the expected rise in the fertility rate would boost revenues and resolve the chronic de®ciencies of the social security system. Notes 1. The views expressed in this chapter are those of the authors and not necessarily of the Bank of Greece. 2. For a further discussion of these issues, see, Bank of Greece (1996, ch. 2). 3. This re¯ects the existing constraints, given that the positive effects of increases in the social security contribution rates have faded, and revenue from the investment of social security fund surpluses has declined as a result of the drop in interest rates. 4. See, inter alia, Provopoulos (1987), Voloudakis (1980). 5. This version of the test is an extension of the Dickey±Fuller test, which makes a semiparametric correction for autocorrelation and is more robust in the case of weakly autocorrelated and heteroskedastic regression residuals. According to Choi (1992), the Phillips±Perron test extension appears to be more powerful than the ADF tests for aggregate data. For more details, see, Perron (1990). 6. The KPSS procedure assumes that the univariate series can be decomposed into the sum of a deterministic trend, random walk and stationary I(0) disturbance, and is based on a Lagrange multiplier score-testing principle. This test reverses

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 191 the null and the alternative hypothesis. A ®nding favourable to a unit root in this case requires strong evidence against the null hypothesis of stationarity. The KPSS test is de®ned as:  ˆ T 2  S2t :=s2 …k†; where St ˆ

t X

ei …1; 2; . . . ; t†

iˆ1

and ei is the partial sum of the residuals from regressing the series on an intercept and possibly a time trend; s2 (k) is a consistent non-parametric estimate of the disturbance variance and T is the sample size. 7. The KPSS statistics are known to be sensitive to the choice of truncation parameter l and tend to decline monotonically as l increases. 8. To determine the appropriate lag length of the structural VAR model, three versions of systems are estimated. The ®rst is a four-lag version (unrestricted model) and the second and third a three- and two-lag version respectively (restricted). The likelihood ratio test was employed to test the null hypothesis that the three versions are statistically equivalent. The likelihood ratio tests a lag four versus two, a four versus three and a three versus one. The test statistic follows the chi-square distribution and indicates acceptance of the null hypothesis.

References Bank of Greece (1996) Annual Report for the Year 1995 (Athens: Bank of Greece). Barro, R. J. and G. S. Becker (1989) `Fertility Choice in a Model of Economic Growth', Econometrica, vol. 57, pp. 481±501. Becker, G. S. (1960) `An Economic Analysis of Fertility', in Demographic and Economic Change in Developed Countries: A Conference of the Universities, NBER Committee of Economic Research (Princeton, NJ: Princeton, University Press). Becker, G. S. (1973) `A Theory of Marriage: Part I', Journal of Political Economy, no. 81, pp. 813±46. Becker, G. S. (1988) `Family Economics and Macro Behaviour', American Economic Review, vol. 78, pp. 1±13. Becker, G. S. (1992) `Fertility and the Economy', Journal of Population Economics, vol. 5, pp. 185±201. Becker, G. S. and R. J. Barro (1988) `A Reformulation of the Economic Theory of Fertility', Quarterly Journal of Economics vol. 103, pp. 1±25. Becker, G. S., K. Murphy and R. Tamura (1990) `Human Capital Fertility and Economic Growth', Journal of Political Economy, vol. 98, pp. S12±S37. Blaug, M. (1962) Economic Theory in Retrospect, (London: Heinemann). Brander, J. A. and S. Dowrick (1993) `The Role of Fertility and Population in Economic Growth: Empirical Results from Aggregate Cross-National Data', NBER Working Paper, no. 4270 (Cambridge, MASS: NBER). Campbell, J. Y. and P. Perron (1991) `Pitfalls and Opportunities: What Macroeconomists should know About Unit Roots', in Blanchard and Fischer (eds), NBER Macroeconomics Annual 1991 (Cambridge, MASS: MIT Press), pp. 141±201. Choi, I. (1992) `Effects of Data Aggregation on the Power of Tests for a Unit Root', Economic Letters, vol. 40, pp. 397±401.

192 Economic Growth, Labour Effort and Fertility Choice De Jong, D. N., J. C. Nankervis, N. E. Savin and C. H. Whiteman (1989) `Integration Versus Trend Stationarity in Macroeconomic Time Series', University of Iowa Working Paper (Iowa: University of Iowa) pp. 89±99. Dickey, D. A. and W. A Fuller (1979) `Distributions of the Estimators for Autoregressive Time Series with a Unit Root', Journal of the American Statistical Association, vol. 74, pp. 427±31. Dickey, D. A. and W. A. Fuller (1981) `The Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root', Econometrica, vol. 49, pp. 1057±72. Ehrlich, I. (1990) `Introduction', Journal of Political Economy, vol. 98, pp. S1±S11. Ehrlich, I. and F. T. Lui (1991) `Intergenerational Trade, Longevity, and Economic Growth', Journal of Political Economy, vol. 99, pp. 1029±59. Enders, W. (1995) Applied Econometric Time Series, (New York: John Wiley). Engle, R. F. and C. W. J. Granger (1987) `Cointegration and Error-Correction: Representation, Estimation and Testing', Econometrica, vol. 55, pp. 251±76. Fuller, W. A. (1976) Introduction to Statistical Time Series (New York: John Wiley). Johansen, S. (1988) `Statistical and Hypothesis Testing of Cointegration Vectors', Journal of Economic Dynamics and Control, vol. 12, pp. 231±54. Johansen, S. and K. Juselius (1990) `Maximum Likelihood Estimation and Inference on Cointegration ± with Applications to the Demand for Money', Oxford Bulletin of Economics, vol. 52, pp. 169±210. Johansen, S. and K. Juselius (1992) `Testing Structural Hypotheses in a Multivariate Cointegration Analysis at the Purchasing Power Parity and the Uncovered Interest Parity for the UK', Journal of Econometrics, vol. 53, pp. 211±44. Kwiatkowski, D., P. C. B. Phillips, P. Schmidt and Y. Shin (1992) `Testing the Null Hypothesis of Stationarity Against the Alternative of a Unit Root', Journal of Econometrics, vol. 54, pp. 159±78. Malthus, T. R. (1798), An Essay on Population (Reprints of economic classics, 1965), (New York: Augustus Kelley). Perron, P. (1988) `Trends and Random Walks in Macroeconomic Time Series: Further Evidence from a New Approach', Journal of Dynamic and Control, vol. 12, pp. 297±332. Perron, P. (1990) `Tests of Joint Hypotheses in Time Series Regression with a Unit Root', in G. F. Rhodes, et al. (eds), Advances in Econometrics: Co-Integration, Spurious Regression and Unit Roots, vol. 8 (New York), pp. 10±20. Phillips, P. C. B. (1987) `Time Series Regression with a Unit Root', Econometrica, vol. 55, pp. 277±346. Phillips, P. C. B., and P. Perron (1988) `Testing for a Unit Root in Time Series Regression', Biometrica, vol. 75, pp. 335±46. Provopoulos, G. (1987) The Social Security System in Greece: Macroeconomic Facets of the Financial Problem (Athens: Institute of Economic and Industrial Research) (in Greek). Sapiro, M. D. and W. W. Watson (1988) `Sources of Business Cycle Fluctuation', NBER Macroeconomics Annual, (Cambridge, MASS: MIT Press), pp. 111±56. Schumpeter, J. A. (1954) History of Economic Analysis (New York: Oxford University Press). Schwert, G. W. (1987) `Effects of Model Speci®cation on Tests for Unit Roots in Macroeconomic Data', Journal of Monetary Economics, vol. 20, pp. 73±101. Simon, J. L. (1992) Population and Development in Poor Countries (Princeton, NJ: Princeton University Press).

George Hondroyiannis, Sarantis Lolos and Evangelia Papapetrou 193 Smith, A. (1937) An Inquiry into the Nature and Causes of the Wealth of Nations (New York: Modern Library). Solow, R. M. (1956) `A Contribution to the Theory of Economic Growth', Quarterly Journal of Economics, vol. 70, pp. 65±94. Voloudakis, E. A. (1980) `A Cross-regional Econometric Investigation of Fertility Behaviour in Greece', Greek Economic Review, vol. 2, pp. 255±72. Wang, P., C. K. Yip and C. A. Scotese (1994) `Fertility Choice and Economic Growth Theory and Evidence', The Review of Economics and Statistics, vol. 46, no. 2 pp. 255±66. Winegarden, C. R. and M. Wheeler (1992) `The Role of Economic Growth in the Fertility Transition in Western Europe: Econometric Evidence', Economica, vol. 59, pp. 421±35.

This page intentionally left blank

Part 3 Institutions and Policies of Integration

This page intentionally left blank

11

Institutional Development and the Harmonisation of Technological Policy in the European Union1 George M. Korres

11.1

Introduction

During the past half a century or so, many arrangements for international economic integration have come into existence. The most important of these, the European Union is, in fact, an amalgamation of three separate communities: the European Coal and Steel Community (ECSC), established by the Treaty of Paris in 1952 and valid for 50 years; the European Economic Community (EEC), created in 1957 by the Treaty of Rome for an unlimited period; and the European Atomic Energy Community (EURATOM), founded by another Treaty of Rome in 1957 and also of unlimited duration.2 The EEC Treaty provided for a drastic form of integration: a common customs union, the free movement of persons and capital, and integrated policies in a number of areas, such as agriculture, transportation, research and technology. Technological change has been widely considered for many years as an engine of growth and an important factor in the development process. Today, there is keen technological competition among the EU, the United States and Japan. The aim is to reinforce technological capabilities and international competitiveness. European technology policy also aims to increase convergence among the member states and to reduce the disparities of the EU's less favoured regions. European technology policy is implemented through various rolling framework research programmes. These consist of various research projects and cover various sectors and scienti®c subjects. Within the EU there is a large technological gap between the advanced and the less favoured regions. This chapter assesses the development of European technology policy, and the way in which it works and is implemented. It also examines the

197

198 Institutional Development and the Harmonisation of Technological Policy

implications of the harmonisation of policies and the transfer of scienti®c, technological and ®nancial resources.

11.2

Evaluation of European technology policy

The countries of Europe have a long cultural and scienti®c tradition. The major scienti®c discoveries and the main developments in technology are products of European civilisation. The Treaty of Rome did not endow the Commission with explicit power to make research and technology policy. The Commission operated only through the unanimous decisions of the Council of Ministers. In the ®rst phase of the Community's research policy only eight articles from the Euratom treaty were devoted to the promotion of research activities, and the treaty did not provide a framework for a general research policy. However, the Community's research activities were developed within this framework and provided the basis for the work being done today. The ECSC and EEC treaties do not contain such detailed provisions as the Euratom treaty. Thus, during the ®rst period, 1953±74, there was no clear common framework for Community's research policy. The Community's research programmes for this period concentrated mainly on the nuclear, steel and agricultural sectors, and it was not until the passing of the Single European Act (1986) that the Commission's competence in technological subjects was extended and its role in these ®elds was strengthened. In 1965 the three Communities (the European Coal and Steel Community, Euratom and the European Economic Community) set up a joint executive committee to examine the merits of coordinated research and development programmes and to prepare for a proposed meeting of the Councils of Ministers on this subject. In 1966 the vice president of the EEC (M. Marjolin) addressed the European Parliament on the importance of technology. He proposed that scienti®c research should be regarded as an integral part of economic policy. At the same time there was another proposition for a technological Marshall aid scheme, based on NATO. Both aimed to ®ll the existing technological gap. In the end, the establishment of a European Technological Community was preferred. During the 1960s several attempts were made to develop cross-national research groupings. For instance, in 1962 Siemens, Olivetti, Elliott Automation (these later formed the core of ICL) and Bull tried to create a cross-European research grouping, but were unsuccessful. In 1969 the Eurodata consortiums (ICL, CII, Philips, AEG, Telefuken, Saab and Olivetti) established the European Space Research Organisation (ESRO) for computer requirements, but this also failed.

George M. Korres 199

In the 1960s nuclear power was one of the most important areas of new technology; the Commission's power in this ®eld derived from the Euratom treaty of 1957. Four research centres were established and the areas of research included high-temperature gas reactors, nuclear ship propulsion and nuclear applications in medicine and agriculture. Later, in the early 1970s, research conducted at joint research centres (JRCs) focused on additional ®elds, such as the environment and solar energy and materials. The most successful example of European collaboration was aerospace. In the 1970s the European Space Agency (ESA) was developed with the collaboration of all West European countries. This helped to create a research space community of scientists, engineers, policy makers and industrialists. In November 1971 the COST programme (European cooperation in the ®eld of scienti®c and technical research) was established. It consisted of 19 OECD Western European members (including Switzerland). COST was a useful framework for the preparation and execution of pan-European projects in applied scienti®c research. The European Patent Of®ce (EPO) was founded in 1973 following the European Patent Convention of 1973, which provides a single-application process for a European patent. This remains valid in all the signatory states for twenty years, and saves the time and expense of applying to individual material patent of®ces. In 1974 the Council decided to work towards a common policy in the ®eld of science and technology, the aim being to coordinate the policies of the member states and implement research programmes and projects of EC interest. During the period 1974±82 there was a somewhat shaky technology policy that failed to produce results. However, there was a move to increase the allocation of funds to R&T activities and, in July 1978, the Commission launched the FAST (forecasting and assessment in the ®eld of science and technology) experimental programme. The main objective of FAST was to de®ne the long-term priorities and objectives of the Community's technology policy. European technology slowed down after the energy crises of the 1970s, but it came into its own in the 1980s. The Eureka project was launched in 1985, and by 1990 it had acquired a total committed investment by governments, companies and research institutes of more than 8 billion ecus, deriving from almost 500 projects. Eureka membership encompasses the EU, the EFTA countries and Turkey. The EU's research programmes and Eureka are complementary. Eureka projects have, through sheer size, gained widespread awareness and created a favourable image for initiatives such as Jessi, HDTV and Prometheus. In June 1991 there were 470 ongoing Eureka

200 Institutional Development and the Harmonisation of Technological Policy Table 11.1 Eureka projects, 1992 Number of active projects Medical and biotechnology Communications Energy Environment Information technology Lasers New materials Robotics and automation Transport Total

102 28 22 117 77 14 53 104 22 539

Cost (m.ecu) 830 1452 490 881 2067 405 317 1291 1110 8843

Percentage of total expenditure 9.4 16.4 5.5 9.9 23.4 4.6 3.6 14.6 12.6 100.0

Source: Eureka (1992).

projects in nine technology areas, varying greatly in scope and ®nancial impact. However, the percentage of funds allocated to technological activities was still very low ± just 2 per cent of the total EC budget expenditure. The joint research centres (JRCs) received about 25 per cent of these funds. There also exist direct-order research programmes, which are more suitable for the technologically advanced member states. Eureka has a large nominal budget and Table 11.1 lists the Eureka projects that were under way in September 1992. Until the late 1980s, EC research policy was mainly concerned with coordinating the technology policies of the member states rather than pursuing a coherent technology policy, and most of the research policy criteria were based on quality rather than needs. However, during the period 1982±90 a more coherent and clear technology policy began to develop. The European Single Act and the Maastricht Treaty worked in this direction. The Single European Act (SEA) explicitly legitimised scienti®c and technical cooperation in Europe by giving the EC formal power in the ®elds of research and technology. Articles 130f±g of the SEA embody a research and technology policy that enjoys equal status with other EC areas, such as economic, social and competition policy. The SEA made substantial amendments to the Treaty of Rome. It contains provisions that are intended to speed up European integration by completing the single market and strengthening economic and social cohesion and cooperation in ®nancial matters. The SEA is aimed at

George M. Korres 201

developing social and environmental policies and establishing a genuine European research and technology community. The principles introduced by the SEA were con®rmed and extended in the text agreed at Maastricht, and the resulting treaty gave a double perspective to technology policy. The coordination of national technology policies has essentially ceased to be entrusted solely to the good intentions of member states, and there is expected to be consistency between national and EU policies. Nowadays, the EU research policy is implemented through speci®c programmes. In the 1980s, the ®rst step towards a Community-wide research policy was the formulation of the ®rst framework programme (1984±87). This introduced the medium-term planning of research activities at the EC level.3 The research framework programmes aimed to strengthen the international competitiveness of European industry in the high technology sectors, particularly in relation to the United States and Japan. The ®rst framework programme covered seven high priority areas, and these formed the basis of a large number of projects in industry, universities and research centres. The second framework programme started in September 1987. Although about 60 per cent of the funds were allocated to industrial research, most of this was to promote the introduction of new technologies in traditional sectors such as engineering and construction. The decisive move towards a comprehensive political strategy for technology sectors came with the second and third framework programmes (1987±91 and 1990±94 respectively). These programmes were based on the Single European Act and were intended to improve the international competitiveness of European industry and reinforce economic and social cohesion. The idea behind the framework programmes is simple: to carry out at the EU level all research that for one or other reason is more usefully and economically dealt with there than at the national level. In April 1990, the Council of Ministers adopted the third framework programme, which overlapped the second one by two years. The total allocation of 5.7 billion ecus was split into two parts: 2.5 billion ecus for 1990±2 and 3.2 billion ecus for 1993±94. The Commission proposed a total amount of 14 700 million ecus for the fourth framework programme (1994±98). The latter complied with the provisions of the Maastricht Treaty and therefore covered all the EU's research and development activities. International scienti®c cooperation was made part of the programme and some research activities that were outside the third programme were included in the fourth.

202 Institutional Development and the Harmonisation of Technological Policy

Table 11.2 provides a sectoral breakdown of the four programmes. As can be seen, from 1984 there was a rapid rise in research expenditure. The amounts included in the ®gures correspond to the resource allocation for four activities: evaluation, coordination, concentration and JRCs. It should be noted that the ®gures shown are for the EU's ®nancial contribution only ± if national contributions were taken into account, then the total budget would be approximately double. The research framework programmes comprise various projects that are allocated on a competitive basis, which implies that the participation of individual member states depends on the quality and strength of their applications. The objectives of the programmes are: (1) to enhance European industrial competitiveness; (2) to set up a vast uni®ed market by promoting standardisation and open procurement; (3) to improve the effectiveness of scienti®c and technical cooperation in the EU; (4) to promote agricultural competitiveness; (5) to speed up the marketing of new technologies by carrying out programmes for the application of information technologies; (6) to help the least favoured regions of the EU to gain access to new technologies; (7) to encourage small- and medium-sized enterprises; and (8) to provide continuing education and training. Table 11.3 compares the ®rst, second and third research framework programmes. As we can see, information technologies accounted for 42 per cent of the total budget for the second programme, while spending on energy research declined from 50 per cent in the ®rst programme to just 14 per cent in the third. Research on life science and industrial and material technologies increased in the third programme and accounted for about 13 per cent and 16 per cent of the budget respectively. Tables 11.4 and 11.5 present, respectively, the R&TD expenditure for the period 1987±98, and the areas of concern and budgetary allocation in the fourth framework programme. In the 1990s, the EU's research plans were aimed at enforcing joint research and changing the approach of cooperation between theoretical and industrial research by providing a new learning environment. The main problem for the less favoured member states is the poor rate of diffusion of new technologies, and lack of access to information on new techniques and technologies. The EU intends to overcome this problem by speeding up the diffusion of technologies between and within the member states. For instance, the Sprint programme was established to reinforce dissemination activities. Sprint of®cially accounted for 90 million ecus in the 1989±93 period, however, in reality, it represented a total expenditure of 180 million ecus as it only supported the costs of an

George M. Korres 203 Table 11.2 EU framework research programmes, 1984±98 Focal areas

Sum in million ecus

Prop. of total budget

First framework programme (1984±87): Agricultural competitiveness Industrial competitiveness Improving raw materials Improving energy resources Stepping up development aid Improving working conditions Improving the S/T potential Horizontal action Total budget

130 1060 80 1770 150 385 85 90 3750

3.50 28.20 2.10 47.20 4.00 10.30 2.30 2.40 100.00

Second framework programme (1987±91): Quality of life Towards an information society Modernisation of industry Biological resources Energy S/T for development Exploiting marine resources Improvement of S/T cooperation Total budget R&D programmes already adopted or in hand Total budget

375 2275 845 1173 280 80 80 288 5396 1084 6840

6.90 42.30 15.60 5.20 21.70 1.50 1.50 5.30 100.00

3000 1200

38.90 15.60

700 1000 1100

9.10 13.00 14.30

700 7700

9.10 100.00

Third framework programme (1990±94): Enabling technologies Information technologies and communications Industrial and material technologies Management of industrial resources Environment Life sciences and technologies Energy Management of intellectual resources Human capital and mobility Total budget

Fourth Framework programme (1994±98): Area I: technologies and demonstration of R&T programmes 11 600 Area II: cooperation of non-EU countries and international. organisations 1400 Area III: circulating and exploiting research fundings 700 Area IV: improving training and mobility for researchers 1000 Horizontal support measures (1600) Total 14 700 Note: Horizontal support measures can be applied to four activities. Source: CEC, EC Research Funding, 1991 (CEC: Brussels).

78.91 9.52 4.76 6.80 100.00

204 Institutional Development and the Harmonisation of Technological Policy Table 11.3 Framework programmes for R&TD activities: a comparison of budget share, 1984±94 (per cent) 1984±87 Information and communication technologies Industrial and materials technologies Environment Life science and technologies Energy Human capital Total

25 11 7 5 50 2 100

1987±91 42 16 6 9 23 4 100

1990±94 39 16 9 13 14 9 100

Source: CEC, EC Research Funding (CEC: Brussels, 1991).

Table 11.4 Framework programmes of EU activities in R&TD, 1987±98

Framework programme 1 Framework programme 2 Framework programme 3 Framework programme 4

Period

Budget (million ecu)

1984±88 1987±91 1990±94 1994±98

3 750 5 396 6 600 12 300

Source: CEC, EC Research Funding (CEC: Brussels, 1991).

activity up to a maximum level of 50 per cent, the remaining funds being provided by the participants. In summary, we can say that there are at least three major bene®ts from technological collaboration within the EU: (1) cost savings for both research and production; (2) reinforced competitiveness against the United States and Japan; and (3) technological convergence of the member states.

11.3 The implementation of policies and the institutional framework The transfer of ®nancial and technological inputs has had a considerable effect on technological performance in the member states. Transferred resources are intended to reinforce the industrial infrastructure and enhance the research activities of small- and medium-sized enterprises. Most of these resources come from the structural funds, which make an important contribution to the development of research activities in the small member states. In June 1988, the European Council decided to reform the structural funds and plan their future coordination.4

George M. Korres 205 Table 11.5 Fourth framework programme, 1994±98 Budget (million ecu) Information and communications technology Industrial technologies Environment Life sciences and technologies Clean and ef®cient energy sources Nuclear safety and security Controlled thermonuclear fusion Research for European transport policy Socioeconomic research Cooperation with third countries and international organisations Stimulation of training and mobility of researchers Dissemination and utilisation of results Total

3 1 1 1 1

Percentage R&D

405 995 080 572 002 414 840 240 138 540

27.7 16.2 8.8 12.8 8.1 3.4 6.8 2.0 1.1 4.4

330

2.7

744 12 300

6.0 100.0

Source: CEC, EC Research Funding (CEC: Brussels, 1991).

Table 11.6 shows the planned EU budgetary expenditure from 1992 to 1997. Most of these funds went to agricultural activities (the Common Agricultural Policy), with R&D initiatives (including internal policies) accounting for only a small proportion of the total. IMPs (integrated Mediterranean programmes) were approved by the European Council in July 1985 to help the less favoured Mediterranean regions (France, Greece and Italy). The IMPs successfully emphasised productive investment, the development of local human resources and the transfer of new and more advanced technologies. The IMP budget amounted about 3.8 billion ecus and covered the period 1985±92. Since the beginning of the European integration process, three structural funds have been set up to promote harmonious economic and social development in Europe: the European Social Fund (ESF) in 1958, the Guidance Section of the European Agricultural Guidance and Guarantee Fund (EAGGF) in 1964 and the European Regional Development Fund (ERDF) in 1975. In 1986, the Single European Act provided for close cooperation between the three funds so that they could contribute effectively to the achievement of ®ve priority objectives: objective 1 related to regions that were lagging behind in terms of economic development; objective 2 related to industrial regions with high unemployment; objectives 3 and 4 concerned long-term and youth

206

Table 11.6 Planned budgetary expenditures, 1992±97 (million ecus, in 1992 prices)*

Common Agricultural Policy Structural operations (Structural funds) (Cohesion funds: IMP/Pedip.) Internal policies External action Administrative expenditures Reserves Total

1992

1993

35 348 18 559 (17 965) (594) 3 991 3 645 4 049 1 000 66 592

35 340 21 270 (19 770) (1 500) 4 500 4 070 3 310 1 500 69 990

* Initiative amounts for R&D policy. Source: CEC, Research after Maastricht: a strategy (Brussels: CEC, 1992).

1994 37 480 22 740 (20 990) (1 750) 5 035 4 540 3 465 1 600 74 860

1995 38 150 24 930 (22 930) (2 000) 5 610 5 060 3 720 1 200 78 670

1996 38 840 27 120 (24 870) (2 250) 6 230 5 650 3 850 1 300 82 990

1997 39 600 29 300 (26 800) (2 500) 6 900 6 300 4 000 1 400 87 500

George M. Korres 207

unemployment; objective 5a concerned agricultural restructuring; and objective 5b related to rural regions with a high level of income but dif®culties with development. Table 11.7 shows average EC expenditure in 1985±89 by category and member state. The greatest proportion went to the agricultural funds and only a small amount to research and technology, despite the fact that Europe was suffering from a major technological imbalance. Between 1985 and 1990, the EC trade surplus in manufactured products dropped from 116 billion ecus to 50.5 billion ecus. In 1990, high technology products represented almost one third of US exports (31 per cent), more than a quarter of Japanese exports (27 per cent) and less than one ®fth of European exports (17 per cent). Table 11.8 lists the technology sectors in which Europe dominates in relation to its competitors (the United States and Japan). As can be seen, Europe is stronger than the United States and Japan in ¯exible computerintegrated manufacturing but behind in optoelectronics and highperformance computers. Europe is behind the United States in medical equipment, biotechnology and arti®cial intelligence, and behind Japan in advanced materials and high-density data storage. However, this does not mean that European products are inferior in terms of quality and innovations. The examples of Ariane, Airbus, telecommunications systems, Scandinavian robots, and so on prove this point. The problem for Europe is the weak integration of research and innovations into an overall strategy that both exploits and orients them. Although it is beyond the scope of this chapter, comparisons can be made and the technological level across countries evaluated with the help of different measures (OECD and UN).5 Finally, Tables 11.9 and 11.10 illustrate the distribution of direct action projects carried out by the JRCs, and also the main European programmes to promote new technologies and R&TD activities. In summary, we can say that any positive effects have been mainly due to ®nancial transfers from the structural funds. What is really necessary for EU technology policy is for more attention to be paid to speci®c strategies that take into account the particular needs of the less favoured regions. The participation of member states in the EU research framework programmes via individual research institutes, organisations, universities and enterprises would give a push to technological ¯ows in these countries, and provide enterprises and scientists with costless and easy access to new technologies and know-how.

208

Table 11.7 Categories of EU expenditure in member states (average 1985±89, per cent) Belg. EAGGF Guarantee 39.0 EAGGF Guidance 1.0 Fisheries 0.0 Social Fund 2.4 Regional funds 1.3 Transport 0.1 Research, energy and industry 4.8 Administrative expenses 46.7 Miscellaneous 4.6

Den. 87.3 1.5 0.4 3.5 1.2 0.1 1.8 1.1 3.1

Ger. 80.1 2.3 0.0 2.7 1.9 0.0 3.2 1.0 8.6

Greece 68.6 5.1 0.0 6.2 15.5 0.2 0.3 0.2 3.9

Spain 33.4 2.2 0.5 10.1 15.6 0.1 0.5 0.5 37.2

France 80.1 3.2 0.2 4.9 4.6 0.0 2.3 0.7 4.0

Source: CEC, The Community Budget: the facts in ®gures, 3rd edn (Brussels: CEC, 1990).

Ire. 67.1 5.5 0.1 13.0 8.4 0.1 0.7 0.5 4.6

Italy 69.8 3.5 0.1 7.7 10.7 0.1 3.4 0.5 4.2

Lux. 0.6 0.6 0.0 0.3 0.6 0.0 0.5 95.3 2.1

Neth. 90.1 0.1 0.1 1.5 0.6 0.1 2.0 0.5 4.7

Port. 14.7 14.7 0.5 20.6 32.7 0.1 0.8 0.8 21.6

UK 55.7 55.7 0.2 13.7 14.7 0.2 4.8 0.7 7.6

209 Table 11.8 Status of European technologies vis-aÁ-vis those of the United States and Japan Europe

Vis-aÁ-vis the US

Vis-aÁ-vis Japan

Ahead:

Digital imaging technology Flexible computer-integrated manufacturing

Flexible computer-integrated manufacturing Software engineering technology

Level:

Advanced semiconductors High-density data storage Sensor technology Advanced materials Software engineering technology

Arti®cial intelligence Digital imaging technology Sensor technology Superconductors Biotechnology Medical equipment

Behind:

Arti®cial intelligence High-performance computers Optoelectronics Biotechnology Medical equipment

Advanced semiconductors High-performance computers High-density data storage Optoelectronics Advanced materials

Source: US Department of Commerce, republished in CEC, Research after Maastricht: a strategy (Brussels: CEC, 1992).

Table 11.9 Distribution of direct action projects, 1977±80

Sectoral policies

Expenditure on direct action Total expenditure projects (million (million units of units of account) account)

Energy Industrial Environment Resources and raw materials Transport Agriculture Social Development aid Public services and other Total

188 ± 28 7 ± ± ± ± 127 350

Percentage of total expenditure

566 137 42 28 19 14 9 4 144 963

58.8 14.2 4.3 2.9 1.9 1.5 0.9 0.5 15.0 100.0

Note: Direct action projects were carried out by the JRCs. Indirect action projects were 50 per cent funded by the EC. Source: CEC, Science, Research and the European Community (CEC: Brussels, 1988).

Acronym and full name

Period

Budget (million ecu)

ESPRIT (European Strategic Programme for R&D in IT) RACE (R&D in advanced communications technologies in Europe) TELEMATICS

I II III De®nition, 1985±87 RACE I, 1990±94 RACE II, 1992±94 1990±94

BRITE/EURAM (Basic Research in Industrial Technologies/Advanced materials for Europe)

BRITE I, 1985±88/ EURAM, 1986±88 BRITE/EURAM I, 1989±92 BRITE/EURAM II, 1992±96 BRITE/EURAM III 1996± 1985±89

450 663 402 75

BRIDGE (Biotechnological Research for Innovation Development and Growth in Europe) BIOTECH

1990±93

100

1992±96

164

ECLAIR (European Collaborative Linkage of Agriculture and Industry through Research)

1989±94

80

BAP (Biotechnology Action Programme)

750 1600 1350 21 460 489 380 100

210

Table 11.10 Major European programmes promoting new technology Prime objectives To promote EU capabilities and competitiveness in IT, especially microelectronics systems Help EU competence in broadband communications equipment, standards and technology Develop telematics, i.e. in health, transport, public administration Support for R&D that upgrades technological or materials base of production Develop infrastructure in biotechnology, especially research and training As for BAP, but for large projects, i.e. molecular modelling, advanced cell culture Prenormative research, more basic than BRIDGE, includes safety Applying advanced biotechnology in agro-industrial sector, especially using raw materials from agriculture

Table 11.10 Major European programmes promoting new technology (continued ) Acronym and full name FLAIR (Food-linked Agroindustrial Research COMETT (Community Action Programme for Education and Training for Technology) VALUE SPRING (Strategic Programme for InnovationTechnology Transfer) EURET STRIDE (Science and Technology for Regional Innovation-Development in Europe) MONITOR

Period 1989±94

Budget (million ecu) 25

COMETT I, 1987±89 COMETT II, 1990±94

30 200

VALUE I VALUE II, 1992±94 SPRINT I, experiment SPRINT II, 1986±89 SPRINT III, 1989±94

30 66 9 90

1990±93 1990±93

1994 27 400

1989±93

22

Prime objectives As for ECLAIR, but food oriented in manufacture and processing Training programmes between university and industry, via enterprise partnerships and international staff exchange Disseminate and exploit research results of EU programmes Promote innovation and technology transfer, especially among SMEs internationally. Merged with VALUE Rail, sea and air transport research DGXVI programme to promote R&D in assisted regions Identify R&D policy priorities

Note: The budget ®gures do not include industrial contributions to ESPRIT and other shared-cost programmes. In such cases the actual expenditure on the programme will be approximately twice the budget ®gure shown above. Source: CEC, Guide to Community initiatives, (Brussels: CEC, 1990, 1991).

211

212 Institutional Development and the Harmonisation of Technological Policy

11.4

Conclusions

European technology policy seeks to create new interinstitutional links and new forms of collaboration, as well as establishing capabilities in speci®c sectors. The EU technology policy has the following objectives: (1) industrial modernisation and competitiveness; (2) quality and productivity improvements in agriculture; (3) dissemination of new technologies; (4) exploitation of human and natural resources; (5) a better quality of life; and (6) regional convergence. EU research programmes also attempt to bring about cooperation between theoretical research (through the different research bodies of the public sector, such as research institutes and universities) and industrial research (through private enterprises). We can summarise the main conclusions and policy implications of this chapter as follows. Technology policy has been heavily concerned with the gap between the EU's progress in technology vis-aÁ-vis Japan and the United States. However, a gap also exists among the EU countries. Meanwhile, the competition among Japan, the United States and the EU remains intense. According to the EU, the following criteria are essential when it comes to selecting and implementing technology policy: (1) research must contribute to a strengthening of the economic performance, social cohesion and competitiveness of the EU countries, and secure their convergence; and (2) research must contribute to the solidi®cation of the common market and to the uni®cation of European scienti®c and technical areas, in order to establish EU-wide norms and standards. Decision making in technology policy has been largely transferred to the EU level: the scope for national decision making has been reduced and the responsibility of the EU has increased accordingly. EU measures to improve the technological activities of small member states should be considered carefully in order to bring about greater ef®ciency. They should be better targeted and more concentrated. The various research measures can be combined with other ®nancial instruments and institutional regulations to create a signi®cant degree of action and improve the productive capabilities of these countries. The EU research framework programmes are designed to meet the speci®c needs of the weaker member states. Financial and technological ¯ows through the CSFs and the framework programmes should be expected to reduce the disparities between member states and improve the opportunities for the less favoured regions. In the light of these remarks, EU technology policy should be reinforced and reoriented in order to make it more coherent and more effectively

George M. Korres 213

targeted at raising the technological capabilities of the small member states. Coordination with the broader EU instruments and resources (such as the CSFs) would have a favourable effect on the productive capabilities of these countries. Furthermore, the traditional industries of the weaker states should be supported by appropriate research and technology programmes and speci®c programmes for the diffusion of technology. Finally, human capital formation should have a particular place in EU policies vis-aÁ-vis the smaller countries. EU technology policy aims to boost the international demand for research activities and, consequently, to reinforce the weak internal market demand of the small member states. This provides an opportunity to expand activities that might otherwise remain at much lower levels. Notes 1. This chapter was ®rst presented at a workshop at the University of London (Queen Mary and West®eld College) and at the London School of Economics. 2. The Treaty of Rome states that the aim of establishing the EEC was `to promote throughout the Community a harmonious development of economic activities, a continuous and balanced expansion, an increased stability, an accelerated raising of the standard of living and closer relations between its member states' (Article 2). `In order to pursue this aim the EEC member states will consider their economic policy as a matter of common interest. They will consult with each other and with the Commission on measures to be taken in response to current circumstances' (Article 103). 3. The EC research programme covered the following: (1) the research and technological programme of JRCs; (2) direct order research programmes in collaboration with and co®nanced by the governments of the member states; (3) training research programmes; (4) international research programmes. 4. There are three structural funds to support the restructuring of economies: the Agricultural Guidance and Guarantee Fund EAGGF, the Social Fund (ESF) and the Regional Development Fund (ERDF). 5. The most widely used international measure indicators are the following: (1) research and technological expenditure; (2) patent registrations; (3) the technological balance of payments; (4) high technology product output and market share; (5) high technology product imports and exports; (6) employment of scienti®c and technical staff. R&D is a better measure of rates of change in real resources over time. However, it measures technological activities in small ®rms only very imperfectly. US patenting is a better measure of the technological activities of small ®rms and can be broken down quite ®nely into speci®c ®rms and speci®c technical ®elds. Neither measure is satisfactory for some categories, such as software technology, but no alternative measure yet exists. There are some important problems with measuring the technological balance of payments; for instance, the heterogeneous nature and the poor comparability of international ®gures. See CEC, European Economy, no. 25, ch. 4 (Brussels: CEC, 1992).

214 Institutional Development and the Harmonisation of Technological Policy

References Aked N. H. and P. J. Gummett (1976) `Science and technology in the European Communities: the history of the cost projects', Research policy, no: 5, pp. 270±94. Cadmos (1991) `European scenarios on technological change and social and economic cohesion', FAST, vol. 16, (Brussels: CEC). CEC (1980) The European Community's research policy, (Brussels: CEC). CEC (1982) The current state of European research and development (Brussels: CEC). CEC (1983) Towards a European research and science strategy (Brussels: CEC). CEC (1983) The Community and small and medium sized enterprises (Brussels: CEC). CEC (1983) The Europe, United States and Japan controversy (Brussels: CEC). CEC (1983) European economy, nos 15±18 (Brussels: CEC, July). CEC (1984) The European Community and new technologies (Brussels: CEC). CEC (1985) European research policy (Brussels: CEC). CEC (1985) The European Community and culture (Brussels: CEC). CEC (1985) The European Community research policy (Brussels: CEC). CEC (1985) Community R&D policy up to 1984 (Brussels: CEC). CEC (1985) ERDF in ®gures (Brussels: CEC). CEC (1986) New technologies in Europe (Brussels: CEC). CEC (1986) `National regional development aid' (Brussels: CEC, Economic and Social Consultative Assembly). CEC (1986) Incentives for industrial R&D and innovation, 2nd edn (Brussels: CEC). CEC (1986±87): `Research and technological development in less favoured regions of the Community' (STRIDE), ®nal reports, two vols (Brussels: CEC). CEC (1987) The Single European Act (Brussels: CEC). CEC (1987) Research and technological development for Europe (Brussels: CEC). CEC (1988) Research and technological development policy (Brussels: CEC). CEC (1988) `Technology in Europe, 1988±1991', Eurotech (Brussels: CEC). CEC (1988) Science, research and the European Community (Brussels: CEC). CEC (1988) Science and technology in EEC member states (Brussels: CEC). CEC (1988) Science and technological policy, scienti®c potential policies in EEC member states (Brussels: CEC) CEC (1988) `Small countries, science and technology and EC cohesion', FAST programme II (1984±1987), exploratory dossier 14 (Brussels: CEC). CEC (1989) `Science and technology for Europe', JRC (Brussels: CEC). CEC (1989) Managing technological change: a key element in technology transfer (Brussels: CEC). CEC (1989) `Statistical analysis of extra-EUR12 trade in high technology products', Eurostat, (Brussels: CEC). CEC (1989) First survey on state aid in the European Community (Brussels: CEC). CEC (1989) Guide to the reform of the Community's structural funds (Brussels: CEC). CEC (1989) European economy, no 41 (Brussels: CEC, July). CEC (1989) `Science, technology and society in Europe', FAST II, report (Brussels: CEC). CEC (1989) European Communities ± Financial reports 1986±1990, various issues (Brussels: CEC). CEC (1989) `The regional impact of Community policies', European Parliament Papers (Brussels: CEC). CEC (1989) Enterprises in the European Community (Brussels: CEC). CEC (1989±90) The Community budget in ®gures (Brussels: CEC).

George M. Korres 215 CEC (1989±91) General reports on the activities of European Communities (Brussels: CEC). CEC (1990) `SPRINT ± innovation and technology transfer', Fifth annual report 1988 (Brussels: CEC). CEC (1990) Trends in scienti®c R&TD in the EEC (by A.Quinn) (Brussels: CEC). CEC (1990) Community Support Framework 1989±1993, Greece (objective 1) (Brussels: CEC). CEC (1990) An empirical assessment of factor shaping regional competitiveness ± the problem regions (Brussels: CEC). CEC (1990) `The regional impact of Community policies', European Parliament Papers (Brussels: CEC). CEC (1990, 1991) EC research funding, 2nd and 3rd edns (Brussels: CEC). CEC (1991) `European Community direct investment 1984±1988', Eurostat (Brussels: CEC). CEC (1991) `A new strategy for social and economic cohesion after 1992', European Parliament Papers (Brussels: CEC). CEC (1991) `Government ®nancing of R&D 1980±90', Eurostat (Brussels: CEC). CEC (1991) Guide to Community initiatives 1990, 1st edn (Brussels: CEC). CEC (1991) Second survey on state aid in the European Community in the manufacturing and certain other sectors (Brussels: CEC). CEC (1991) Annual report of the implementation of the reform of the structural funds ± 1989 (Brussels: CEC). CEC (1991) `European scenarios on technological change and social and economic cohesion', FAST, vol.16 (by Cadmos) (Brussels: CEC, June). CEC (1991) Aid element of government R&D contracts (Brussels: CEC). CEC (1992) Information and communications technologies in Europe, 1991 (Brussels: CEC). CEC (1992) Research after Maastricht: a strategy (Brussels: CEC). CEC (1992) `IMP ± Progress report for 1990', SEC(92)690 ®nal (Brussels: CEC). CEC (1992) SPRINT ± The European network for technological inter-®rm co-operation (Brussels: CEC). CEC (1992) A speci®c programme for R&TD and demonstration in the ®eld of agriculture and agro-industry, a list of ®nanced projects (Brussels: CEC). CEC (1992) `The Community's Structural Interventions', Statistical Bulletin, no. 3 (Brussels: CEC). CEC (1992) Second annual report on the implementation of the reform of the Structural Funds (Brussels: CEC). CEC (1992) European economy, no 25, ch. 4 (Brussels: CEC). EIB (1982±91) Annual reports, various issues (Luxembourg: EIB). Giannitsis, T. (1991) Globalisation and the small less advanced Countries: the case of Greece (FAST), vol. 21, University of Athens (Brussels: CEC, April). Grahl, J. and P. Teague (1988) `The EUREKA project and the industrial policy of the European Community', International Review of Applied Economics. Grahl, J. and P. Teague (1990) 1992 ± The big market: the future of European Community (Lawrence & Wishart). Hodbay, M. (1989) `The European semiconductor industry: resurgence and rationalisation', Journal of Common Market Studies, vol. XXVIII, no 2. Krieger, L., A. Mytelka and M. Delapierre (1987) `The alliance strategies of European ®rms in the information technology industry and the role of ESPRIT', Journal of Common Market Studies, vol. 25, no 2.

216 Institutional Development and the Harmonisation of Technological Policy Naxakis Xarris (1996) `The globalisation of production, markets and technology', Oikonomika Xronika, no. 94 (in Greek) OECD (1968) Fundamental research and the universities (Paris: OECD). OECD (1988) Industrial revival through technology (Paris: OECD). OECD: (1989) `Science and technology indicators: R&D production and diffusion of technology', report no. 3 (Paris: OECD). Peterson, J. (1991) `Technology policy in Europe: explaining the framework programme and Eureka in theory and in practice', Journal of Common Market Studies, vol. XXIX, no. 3. Richonnier, M. (1984) `Europe's decline is not irreversible', Journal of Common Market Studies, vol. 22, no. 3. Sharp, M. (1991) in Technology and the future of Europe, ed. C. Freeman, Marie Jahoda, Keith Pavitt, Margaret Sharp and William Walker (London: Pinter). Williams, R. (1973) European technology (London: Croom Helm). Woolcock, S. (1984) `Information technology: the challenge to Europe', Journal of Common Market Studies, vol. 22, no. 4. Yuill, D. and K. Allen (1989) European regional incentives: a review to member states (Bower-Saur in association with the European Policies Research Centre, University of Strathclyde).

12

Reform of the EU Institutional Framework for Coastal Shipping: Concepts Concerning Greek Coastal Shipping Policy George P. Vlachos and Maria Lekakou

12.1

Introduction

Regular coastal shipping is a vital means of transport in Greece because of the large number of island groups that form part of the country. Important urban centres are situated on the coast and the Greek people have a long maritime tradition. In effect, the coastal shipping lines make up the arteries that connect various parts of the country together, and for this reason the Greek state has a long history of involvement in the shipping industry. In the Greek coastal services there are many market structures, ranging from state monopoly to perfect competition,1 but the predominant structure is differentiated oligopoly, regulated by the state. The existing institutional framework responds to the speci®c market structure, the characteristics of which can be summarised as follows:2

. A small number of interdependent enterprises; . A large number of independent users; . Incomplete knowledge of the market on the part of all involved (enterprises, employees, passengers and the state);

. Various purposes. Enterprises aim to realise pro®ts and maintain their

share of the market. Passengers seek transport for speci®c reasons (leisure, business, and so on) and emphasise quality of service. The state, through its regulation policy, aims to protect social groupings, promote coastal industry and improve the transport system; . Differentiated service; . Very severe institutional and economic barriers to entry; . Limited mobility of enterprises; 217

218 Reform of the Institutional Framework for Coastal Shipping

. Indivisibility of supply; . Inadequate foresight; . High ®xed costs, low variable costs and almost ®xed marginal costs. This institutional and economic situation seems to have reached its limits. Internal reasons, as well as the obligation to comply with EU Regulation 3577/92, demand the reform of the operational framework.

12.2

The existing institutional framework

The existing framework covers all the main aspects of the operation of regular passenger/ferry ships.3 A review of the most important points is conducted in this chapter. 12.2.1 Regulations laid down in the Code of Public Marine Law (regular passenger/ferry lines) The articles that address issues concerning regular/ferry lines are Articles 164±8 of the Code of Public Marine Law.4 The main concerns of these articles are as follows: passenger/ferry ships' decommissioning age, maritime cabotage, ship types, conditions of entry into the ¯eet, scheduling periods, schedules, sailing prohibitions, manning conditions, immobilisation, non-pro®table lines and freight rates. It is clear from the above that the state is involved in all aspects of coastal shipping, from the building of a ship to its decommissioning. 12.2.1.1

The age of passenger ships

Greek passenger ships5 must be withdrawn from service at the age of 35 years. This rule, which has been valid since 1970, came into effect because the numerous accidents that had occurred were attributed to the age of the vessels. The imposition of this age limit without the simultaneous application of other criteria was arbitrary. 12.2.1.2

Maritime cabotage

Only Greek ships have the right to offer passenger transportation between Greek ports. Exclusive maritime cabotage was granted to Greek ships in 1927 and has remained valid ever since. In our opinion, this situation arose because it was considered that foreign shipping ®rms could outcompete great ®rms as a result of state subsidies and the overcapacity that characterised the period after 1922.6

George P. Vlachos and Maria Lekakou 219

12.2.1.3

Greek passenger ship classi®cations

Passenger ships are classi®ed according to place and time of building, the ®nancial institution that ®nanced them, time of entry into the ¯eet, and total or partial compliance with the technical speci®cations laid down by the Mercantile Marine Ministry. 12.2.1.4

Entry into the passenger/ferry ¯eet

Shipowners who wish to introduce a new vessel into the Greek ¯eet are obliged to obtain a safety certi®cate.7 The entry of new ships is theoretically determined on the basis of statistical data and the characteristics of its intended route (distance and so on). 12.2.1.5

Schedule periods

There are three schedule periods:

. Winter: 1 November to 31 March; . Summer: 1 April to 31 October; . Annual: all 12 months.

This distinction is very important due to the high seasonality of the traf®c. 12.2.1.6

Line categories

There are four basic line categories.8 The various lines form a crisscrossing, extensive network, with the port of Piraeus at centre. We believe that this type of network does not provide the island inhabitants or visitors with the best service, and from the shipowners' point of view the large number of ports of call increases their operational costs. A more decentralised network would serve passengers and companies better.9 12.2.1.7

Immobilisation of vessels

Under the Code of Public Marine Law the immobilisation of vessels is permitted for the following reasons, subject to the approval of the Mercantile Marine Ministry: dry docking, annual inspection, repairs and maintenance, conversion or modi®cation, bad weather conditions, unexpected repairs and so on. Immobilising their vessels during the winter season bene®ts coastal shipping enterprises as it reduces their total costs. However, it results in a signi®cant reduction in the number of scheduled services, provoking justi®ed complaints by passengers, particularly those from small islands. 12.2.1.8

Manning of vessels

Shipowners10 are obliged to comply with the rule governing crew numbers, which states that all scheduled vessels must be fully manned

220 Reform of the Institutional Framework for Coastal Shipping

for the entire year. This obligation is a matter of concern to shipowners as, combined with the low winter traf®c, it increases the cost of service disproportionately. This is the main reason for extended immobilisation, at the start of which almost all crew members are dismissed. 12.2.1.9

`Thin lines'

`Thin lines' are lines for which there is low demand. Therefore, independent operators show no interest in them.11 Consequently, the Greek state, in order to convey passengers, cargo, mail and tourists to under-subscribed designations, charters passenger or cargo ships to operate on a ®xed schedule. The entire cost of these services is borne by the state. The decision to charter ships to serve on thin lines is reached after discussions between the mercantile marine minister and a consultative committee. Thin line contracts are put out to public tenders and vessels are subject to various minimum requirements in respect of speed, tonnage, passenger and cargo capacity, age and so on (this does not happen with ships operating on the main lines). The duration of the voyage on some of these thin lines is extremely long due to the large number of ports of call, and many island inhabitants are dissatis®ed with the service. In our opinion, it would be better to replace these long routes with regular connections between small islands and the closest commercial island, from where the port of Piraeus and other main ports can be reached directly. 12.2.1.10

Freight rates

As provided for in the Code of Public Marine Law (Article 178), the rates charged by main and secondary lines are determined by the Mercantile Marine Ministry after consultation with the consultative committee. On the local lines, freight rates are determined by the local port authority and approved by the Mercantile Marine Ministry. It should be noted that the law forbids deviation from the ®xed rates. This serves to prevent the offering of discounts to boost competitiveness, a practice that was widespread before the Second World War, when discounts could reach as much as 70 per cent. However, of®cial discounts are granted to certain categories of cargo and passenger, including handicapped people, public servants, military personnel, professional groups, reporters, and members of parliament. These discounts are part of the state's social policy and have yet to be seriously analysed in terms of their social and economic impact. The chartering rates for passenger ships are set freely, and if the vessel

George P. Vlachos and Maria Lekakou 221

concerned is scheduled, the Ministry ensures that the schedule is adhered to. 12.2.1.11

Entry approval

Shipowners who wish to enter a vessel into coastal trade are required to apply to the Mercantile Marine Ministry for a license to operate on the main or secondary scheduled lines. They have to provide full details of the vessel in question (for example, type of ship, tonnage, speed, year of construction, passenger or vehicle capacity) and the predicted sailing time The consultative committee reviews each application but the ®nal decision is taken by the Mercantile Marine Minister. Critics of this process focus on: (1) the lack of integrated planning by the Ministry in respect of the economic and social needs of the country, especially the islands; (2) the composition of the consultative committee (the majority are appointed by the state and passengers and shipowners are not represented); and (3) the degree of power enjoyed the Mercantile Marine Ministry. As well as the provisions laid down in the Code of Public Marine Law, there are a number of presidential decrees that regulate important aspects of passenger/ferry operation, such as:

. Prohibitions on sailing due to bad weather conditions (PD 852/76); . Unfair competition between newcomers and established vessels (PD 859/78);

. Manning legislation (PD 177/74, PD 581/77, PD 230/84). 12.3 The provisions of the 1992 EU regulation that relate directly to Greek passenger/ferry lines Council Regulation 3577/92, which grants member-state nationals the freedom to provide maritime service within the EU (maritime cabotage), was issued in December 1992 as part of the second phase of the move towards a common maritime policy. Other measures concerned:

. The establishment of an EU shipping register (EUROS); . The improvement of port control within the EU; . A common de®nition of an EU shipowner. Cabotage had not been included in the ®rst regulations, issued in 1986, but subsequently became a topic of pressure and dispute. With the

222 Reform of the Institutional Framework for Coastal Shipping

intervention of the European Parliament and the Economic and Social Commission, agreement was eventually reached on the liberalisation of European cabotage and Article 1 of Regulation 3577/92 came into force on 1st January 1993. Articles 2, 3, 4 and 6 of the regulation are also very important to the future of Greek coastal shipping. Article 2 makes special reference to island cabotage, public service contracts and public service obligations. Article 3 refers to manning issues and to vessels servicing islands (paragraph 2) ± the latter are to be regulated by the host state. Article 4 refers to questions of public service. It requires Greece to ensure provision of year-round service to the islands for reasons of social justice and cohesion. Such public service contracts should be offered on a non-discriminatory basis into all EU shipowners. Finally, Article 6 (paragraph 3) grants Greek regular passenger and ferry lines exemption from the provisions until 1 January 2004, for reasons of socioeconomic cohesion. Paragraph 1 of Article 6 is also important because it categorises maritime services into cruise services, transport of strategic goods, services by ships of less than 650 tons and regular passenger and ferry services. This is very important for island transport, since it establishes a difference between cruise services and regular passenger/ferry services.

12.4

Final proposals

12.4.1 Towards a new legal framework We believe that the easing of the inequalities that exist in some areas of the Aegean archipelago and the implementation of the EU regulation on cabotage demand the reform of the current legal framework for Greek coastal passenger shipping. The existing framework determines most aspects of these services and changes need to be made to all of them. 12.4.2 Cabotage Restrictions Restricted entry to coastal services, which under Greek law is reserved for Greek-registered ships, is due to be terminated in 2004. As discussed above, a period of grace was extended to Greece for reasons of economic and social cohesion. The focus is now on suitable legislation to ensure that domestic maritime passenger services will continue to function without major disruptions.

George P. Vlachos and Maria Lekakou 223

12.4.3

Age restrictions

The existing legislation sets the upper age limit for the entry of a vessel into coastal passenger transport at 20 years. We believe that the age of a vessel does not directly correlate with its safety, nor with the quality of the service offered. On the contrary, what are needed are precise speci®cations that relate to each particular service. Moreover, this age restriction has an economic impact in the international market for second-hand passenger vessels; the price of 18±19 year-old vessels rises because of the strong demand by Greek buyers. Therefore, the abolition of the age limit would probably result in lower second-hand prices. The compulsory withdrawals of vessels aged 35 years was enforced in 1970 after a major ferry accident in which 240 people lost their lives.12 However, that sinking was not the result of structural or mechanical problems such as metal fatigue or mechanical failure, but was due to the movement of an improperly secured lorry in very bad weather conditions. Many researchers have shown age is positively correlated with accidents,13 but it should be remembered that ships of all ages incorporate numerous technological changes and safety provisions. It is necessary to consider all accidents, and not just those which result in total loss.14 Such a study for Greek coastal passenger shipping does not exist. The 35-year limit was imposed in an effort to regain public con®dence after the sinking of the Iraklio, and possibly to encourage ¯eet renewal.15 Other countries do not have a similar age limit for withdrawal (for example, Italy, France, Germany and Cyprus) and this is re¯ected in recent developments in the EU with regard to safety issues. In conclusion, we suggest that high safety standards and quality of service call for the imposition of regulations and controls that are not only related to the age of a vessel. 12.4.4

Annual itineraries

Coastal passenger shipping companies have to provide a service throughout the year but their itineraries vary according to the season (frequent services during the summer season, far fewer in the winter). Approval of these seasonal itineraries is given at the beginning of each season by the Mercantile Marine Ministry. Itinerary approval for the summer season is in¯uenced by the service company and provides services during the winter season aided by Ministry staff's experience of services to the islands. Complaints about the itineraries are routinely received from the inhabitants of islands. Most of the complaints are about the low frequency of the services, particularly in the case of sparsely populated islands.

224 Reform of the Institutional Framework for Coastal Shipping

We believe that the current procedures cause problems both for the shipping companies and for the island inhabitants. As far as the companies are concerned, annual rather than seasonal itinerary approval would enable better planning and more ef®cient promotion of the services offered to foreign tourists. With regard to island inhabitants, there are problems with the current network of shipping services as well as with the itineraries (we shall expand on this in the next subsection). Finally, we believe that the unwillingness of shipping companies to engage in winter operations offering limited revenues could be overcome by the introduction of a pooling arrangement, whereby the smaller vessels in the pool would operate during the off-peak season and the larger vessels during the more lucrative summer months. This would serve to reduce expenses considerably. 12.4.5 The shipping network and ports of call The Greek state plays a large role in determining routes and ports of call in that it is responsible for network design. The current network, particularly in the case of the main routes, is centred on Piraeus. Another signi®cant feature is poor inter-island transport: islanders are often able to travel to Piraeus with ease but not to the administrative centre of their region or to nearby islands.16 We believe that the network will change in the future and will come to resemble the `hub-and spoke' arrangement adopted by the airlines after deregulation.17 This belief is strengthened by the fact that the EC regulation also provides for the subsidisation of routes that serve the public interest but are avoided by shipowners as they are not pro®table. We predict that private operators will concentrate on the routes that link Piraeus with the bigger or tourist-oriented islands, while the public interest lines will have these islands as their starting point in the provision of inter-island transport. Proper coordination of such a system will be very important if the standard of services is to match the island inhabitants' needs. This type of network would have a positive effect on operational costs in that vessels would have fewer ports of call ± the greater the number of ports of call, the higher the consumption of the better-quality fuel that port manoeuvres require.18 The design of the new network should also take into account the current congestion at the port of Piraeus on certain days and at particular hours of the day, causing an increase in atmospheric pollution.19 The choice of an alternative port within Attica, such as Ra®na or Lavrion, is restricted by a serious lack of infrastructure and poor access.20

George P. Vlachos and Maria Lekakou 225

However, it is possible that when the restriction on foreign operators is abolished, companies will appear that will base themselves at alternative ports and create new routes in an effort to gain a share of the coastal shipping market. 12.4.6

Crew numbers

The criteria determining the manning level of coastal vessels, as regulated by the Mercantile Marine Ministry, are in great need of review. For instance, the number of cooks per vessel is currently based not on passengers number, but on the gross tonnage of the vessel. This and similarly inappropriate criteria should be reconsidered by the Ministry. 12.4.7

Certi®cate of approval for entry into the industry

The reform of the coastal shipping services could lead to abolition of the main institutional barrier to entry into the market: the certi®cate of approval. However, we do not believe that entry should be entirely free. Rather, it should be limited by circumstances and requirements. A prime constraint is port capacity, so a comprehensive plan should be drawn up that takes into account the numerous factors (quantitative and qualitative) involved in the optimal use of port infrastructure. A second constraint is imposed by safety and quality factors. All means of passenger transport are subject to very high international safety standards, and qualities such as speed, punctuality and reliability are essential. The compliance of companies with de®ned conditions should be controlled by the state, given the importance of maritime passenger transport for the cohesion of Greece. The development of transport provision to and from the Greek Islands requires considerable knowledge that is currently not available. Hence, there is a need to conduct extensive research using reliable data that are relevant to all aspects of passenger ferry operations (demand, supply, ports) and consider all the islands. What is really needed is an integrated information system that also covers passenger and vehicle reservations and provides information to the public,21 leading to improved quality of services. Such a system would also ®nancially bene®t the state as it would allow the collection of revenues that might otherwise be evaded. This proved to be the case after the deregulation of the airlines, where computerisation played a major role.22 To conclude, we believe that the eventual removal of barriers to entry should be accompanied by a coordinated state policy, and that the state regulator should be replaced by a state coordinator.

226 Reform of the Institutional Framework for Coastal Shipping

12.4.8 Compliance with rules on safety, navigation and environmental protection No discussion is required on this point as the rules imposed by the Mercantile Marine Ministry comply with those laid down in the international treaties (IMO and SOLAS), and are sometimes even more strict. The only contentious point is whether adherence to the rules is adequately monitored. 12.4.9 Fares At the moment all fares are determined by the State. The principle consideration when determining fares is distance.23 However, relating fares only to distance, and therefore to costs, overlooks important factors such as the level of demand, seasonality, and the different facilities and quality of service offered by various vessels. Another element that should be taken into account is the development policy for each island. We propose the adoption of price differentiation based on the characteristics of the service offered (time of year, day of the week, hour of the day, the vessel's facilities and so on), the island development policies (especially in the case of public interest routes) and the cost of providing the service. The need for an updated and rational pricing system is obvious, but we believe that the structure of the consultative committee should also be changed. It should consist of representatives of the state, the passengers and the carriers, plus specialists on passenger shipping, and it should be an independent body free from the state control.24

12.5

Conclusions

While Greece is not obliged to reform its institutional framework for coastal shipping, it should begin to make preparations for a certain amount of competition from EU shipowners.25 A presidential decree to open up the regular passenger/ferry lines to shipowners from other EU countries would be suf®cient to ensure compliance with EU Regulation 3577/92. However, requests for reform of the framework have come not only from abroad but also from within the country, speci®cally from users (passengers, shipowners, island authorities) and the shipowners' union. Recent research and articles on the question of coastal transport have also concluded that reform is needed.26 We believe that the time has come to review the operational framework of the coastal lines in order to ensure high-quality service, and regular and reliable inter-island and island±mainland connections. All this should

George P. Vlachos and Maria Lekakou 227

take place in a competitive environment; in our opinion, unrestricted competition is impossible where Greek island transport is concerned because of the high seasonality, sunk costs, high ®xed costs, limited port infrastructure, the all-year manning requirement and excess capacity.27 Questions that remain to be answered in the `new era of coastal shipping' are: (1) will certi®cates of approval continue to be required after the year 2004; and (2) will EU (non-Greek) shipowners have to comply with the rules and regulations described in Section 12.2? The answer to the second question is straightforward ± in order to promote fair competition, there should be equal terms for Greek and other EU vessels. The answer to the ®rst question is more complex, because it involves issues related to the reform of the institutional framework. To this end, we would recommend that an in-depth survey be conducted. The main topics of this survey would be the impact of the liberalisation of Greek maritime cabotage, the adoption of measures appropriate to a competitive environment and measures to ensure social and economic cohesion. Core considerations are the size of the ¯eet needed to maintain year-round operations, the institutional and infrastructural aspects of ports and clearly de®ned procedures in respect of public services. We use the word survey rather than research because we believe that reform of the framework should involve the Greek people as a whole ± not just government of®cials and specialists ± in order to meet the needs of Greece and the Greeks in the twenty-®rst century. Notes and References 1. G. Vlachos, `Marine Merchande HeÂlleÂnique: DeÂveloppement, probleÁmes structurels et perspectives possibles' (Paris: Thesis, 1980). 2. M. Lekakou, `The Economic Aspects of the Coastal Shipping Issue. Pricing Policy', PhD thesis, University of Piraeus, 1994. 3. Regular passenger ferry lines are lines where ships offer transfer services between ports on a regular and frequent basis. 4. See Part 4 of the Code of Public Marine Law, `Special Marine Issues', especially chapter A, `Coastal Shipping'. 5. A passenger ship is a ship that can carry more than 12 passengers. Nonpassenger ships are cargo ships or naval ships. 6. G. Vlachos and K. Zois, `Assessment of Greek Maritime Policy for the post-war period up to present', Archives of Economic History (Athens, 1997). 7. The Safety Certi®cate is a mandatory document issued by the Mercantile Marine Ministry and must be presented when a vessel is inspected. 8. (1) Main lines are based in Piraeus and connect at least two ports in different provinces; (2) secondary lines are based in ports other than Piraeus and connect at least two ports in different provinces; (3) local lines connect at least two ports in the same provinces; and (4) the (special) line of Argosaronikos sails between Piraeus and islands in the prefecture of Piraeus.

228 Reform of the Institutional Framework for Coastal Shipping 9. M. Lekakou and S. G. Sturmey, `Ferries as part of the transport infrastructure', Essays in Honour of Ameritus Professor B. N. Metaxas (Piraeus: University of Piraeus, 1997). 10. A shipowner is any national pursuing shipping activities (PD 684/76). 11. The continued existence of `thin lines' has special economic signi®cance and shows the extent to which the industry is subsidised by the state. 12. The Iraklio sank in December 1966 while travelling from Piraeus to Crete. 13. G. Vlachos, `Accidents marins: Nouveau cadre institutional et propositions d'ameÂliorations du Facteur Humain', in Revue du Groupe (GreÂnoble: Universite GreÂnoble II, Faculte des Sciences EÂconomiques, forthcoming). 14. K. Gizakis, A critique on the content of variables used in the analysis of accidents in world shipping (University of Piraeus, 1997; volume issued in the honour of Professor A. Stavropoulos). 15. G. P. Vlachos and K. Gizakis, `Casualties to ferry, ro/ro and passenger ships and the rules for safety', Cyprus Journal of Science and Technology (Nicosia, Cyprus: Frederick Research Center/Frederick Institute of Technology, 1997). 16. H. N. Psaraftis and A. D. Papanikolaou, `Impact of new technologies on Shortsea Shipping in Greece', European Shortsea Shipping, (London: Lloyd's of London Press, 1993), pp.79±92. 17. OECD, DereÂglementation et concurrence dans le transport aeÂrien (Paris: OECD, 1996), pp. 72±3. 18. CEPE `The Financial Status of the Greek Passenger Coastal Shipping Enterprises', unpublished study, Athens, 1992. 19. I. Egglezou, V. Tselentis, E. Tzannatos and G. Amanatidis, `Estimation of maritime induced atmospheric pollution in the Port of Piraeus' HELECO'93 (Athens, 1993). 20. V. Sambracos and M. Lekakou, `Possibilities of operating in coastal shipping from the port of Lavrion' (Lavrion: Technical Chamber of Greece, 1993). 21. A. J. Bos, and H. de Vos, `Information System for Improving Market Activities in the Shortsea Trade', European Shortsea Shipping (London: Lloyd's of London Press, 1993), pp. 95±102. 22. OECD, DeÂreglementation, op. cit., p. 73. 23. In regression analysis of the correlation between fare and distance, R2 came to 0.86, which shows how strongly fare and distance are correlated. 24. A. M. Goulielmos and M. Lekakou, `The Economic Structure of Greek Passenger Coastal Shipping', European Shortsea Shipping (London: Lloyd's of London Press, 1993), pp. 301±15. 25. We believe that competition by EU shipowners will be limited because of speci®c conditions we have described. Limited port infrastructure will prevent the massive entry of vessels from the Baltic and the North Sea. Therefore, there will be limited commercial possibilities for Mediterranean shipowners. 26. H. N. Psaraftis, `Greek Coastal Shipping: Status, Prospects and Investment Opportunities, Final Report to ETBA' (Athens, 1993) (in Greek); S. G. Sturmey, G. Panagakos and H. N. Psaraftis, `Institutional and socio-economic issues in Greek ferry services', in Second European Research Roundtable Conference on Short Sea Shipping (Delft: Delft University Press, 1994); Ministry of the Aegean, Diagnostic Study of the transportation System in the Island Region of Aigaio (Athens: Ministry of the Aegean, 1995); Ministry of the Aegean, Reorganisation of Coastal Shipping (Athens: Ministry of the Aegean, 1995).

George P. Vlachos and Maria Lekakou 229 27. H. Heijveld, and R. Gray `The competitive environment of a service industry: the example of the UK ± Continent passenger sea ferry services', Maritime Policy and Management, vol. 23 (April±June 1996), pp. 157±67.

13

Regional Inequalities and the Allocation of EU Funding: the Case of Greece Dimitrios Giannias

13.1

Introduction

Regional indices of almost any sort attract attention. The contention is that the well-being of economic agents (in terms of either utility or pro®t) depends on various environmental factors, such as infrastructure, transportation systems, the climate, environmental quality, crime and public services, as well as more traditional pecuniary factors such as the price of inputs and outputs, the cost of living, the technology that is available and its cost, the conditions of the business environment and so on. These are all important location factors. One consequence of special information technology is the rapid increase in the volume and availability of data on social, economic and physical environments. Economic agents must attempt to make sense of these data in order to make the best possible decisions. Unfortunately, the rate at which usable information is produced from these data is increasing only slowly ± there are a lot of data but too little information. A common way of avoiding being swamped by data is to use indicators as a tool to produce information. Ott (1978) describes indicators in the following way: Ideally, an index or an indicator is a mean devised to reduce a large quantity of data down to its simplest form, retaining essential meaning for the questions that are being asked of the data. In short, an index is designed to simplify. In the process of simpli®cation, of course, some information is lost. Hopefully, if the index is designed properly, the lost information will not seriously distort the answer to the question. Certainly, no indicator is perfect, and the price that has to be paid for extracting information from the available data is the probable distortion

230

Dimitrios Giannias 231

of that data. In order for indices to be a useful tool, they must be designed with care so that they minimise information distortion and are best able to answer the questions that economic agents and researchers seek to answer. Our societies express in various ways their need to improve their knowledge about, and information on, all the aspects of our natural and non-natural environments that eventually determine the possibility of healthy development and the alleviation of regional inequalities. Therefore, it is important to bring reports and statistics to life. To this end, the development of regional indices can be extremely useful, just as indices of prices and output are widely and successfully used to summarise various aspects of our economies. A number of justi®cations for developing an index to represent aspects of a region's environment can be found in Hope and Parker (1990) and Hope et al. (1991, 1992). The locational decisions of economic agents depend on various aspects of their external environment. Within our framework, composite indices are assumed to provide the necessary information to economic agents. A common characteristic of all indices is that they take the form of a weighted average of a set of variables (see, for example, Blomquist et al., 1985, 1988; Giannias, 1996). Although the selection of regional indicators is best achieved by scienti®c and expert consensus, public opinion should be used when setting the weights for such indices (Hope et al., 1991). Experts do not necessarily have superior knowledge at the aggregate level, and the major priorities and interests should be socially determined (Gould et al., 1988). The objectives of this chapter are: (1) to develop a composite index that allows a comparative evaluation of all ten regions of Greece; and (2) to use it to obtain rankings for the regions under consideration. Investigation of the meaning of the composite index in terms of microeconomics and analysis of the decision-making process of consumers indicate that we should use the same set of weights for each region (unlike in Hope and Parker, 1995); this is shown in Section 13.2. Our regional index may be interpreted as a standard of living index or an index of the relative development of a region, since it takes into consideration variables that, taken as a set, can approximate the overall state of the external environment of a region (infrastructure, the availability of public services, including health and education, and so on). Integration can be said to have been attained if the main characteristics of life and development across regions have converged, and the regional index values are relatively close. In this chapter, we introduce a methodology that allows us to compare life across Greece, and also

232 Regional Inequalities and EU Funding

investigate whether EU funding is being directed to the regions that need it most, so that the objective of integration can eventually be achieved.

13.2 Theoretical framework for index development and interpretation In this section, we present a framework for the analysis of consumer behaviour with regard to locational and other choices. An explicitly regional index is introduced to the analysis. Our framework assumes that consumers within well-de®ned homogeneous regions have identical tastes and skills, are completely mobile within their region and could not be made better off by relocating. Movement between regions is not possible because of high moving costs or institutional and legal barriers. In our analysis, sites or regions are fully described by a bundle of variables: ali ; a2i ; . . . ; ani , where, aki is the kth variable of the site or region i, k ˆ 1, 2, . . ., N, and N is the number of variables. Consumers perceive in their own way the variables of a region or site. These specify the regional index value that the consumer assigns to them ± RI, which includes all aspects of the natural and non-natural environment of a region. Consumers do not assign the same regional index value to identical bundles of site-speci®c variables. More speci®cally, the regional index, RI, is assumed to be scalar, and the regional index, RIji, that consumer j assigns to a bundle of attributes, ali ; a2i ; . . . ; aNi is: RI ˆ fj …ai ; a2 ; . . . ; aNi † The infrastructure of a region, the availability of public services and so on are assumed: (1) to be approximated and described by the regional index (RI); and (2) to affect consumer preferences, which are described by a utility function. All consumers in a region are assumed to have identical tastes. Let the utility function of a consumer in region j be: Uj(.). Individuals are assumed to consume the numeraire good, X, which is a composite good with a price that is equal to one. A consumer supplies one unit of labour and receives income I in return. This income is assumed to be a function of the regional index of the region. For example, for a consumer in region j, Ij ˆ Ij(RIjj), which is spent on housing and the numeraire good. The rental price of a house in region j is a function of the vector of housing characteristics, h, and the regional index; that is, the rental price of a house is speci®ed by a function of the following form: Pj = Pj(h,RIjj). It is assumed that Pj(h,RIjj) ˆ Rj(RIjj) h', where h' is the transpose of h and Rj(.) is the vector of implicit prices for each housing

Dimitrios Giannias 233

characteristic in region j. An equilibrium must be characterised by equal utility for identical consumers within a region. The consumer's income, Ij(RIjj), may depend in equilibrium on the regional index value that the consumer places on the bundle of regional variables that he or she faces, because a high RI value indicates a better external environment, which has a positive effect on productivity and thus on the consumer's income. Moreover, rents, Pj(h,RIjj), may depend in equilibrium on the characteristics of the region and the value of its regional index. The latter formulation is equivalent to assuming that there may exist price differentiation in the rental housing market since different consumers may assign a different regional index to a region, which implies (or better, assumes) that the use consumers get from their houses depends on how they perceive and appreciate the overall conditions of the external environment of the region in question (in our formulation this is assumed to be given by the consumer's regional index value). The above relationship among consumer income, rents and regional characteristics has been empirically veri®ed by Bellante (1979), Blomquist et al. (1985, 1988), Eberts and Stone (1986) and Johnson (1983). A utility maximising consumer in region j solves the following optimisation problem: max Uj …h; X; RIjj † with respect to h,X,RIjj, subject to Ij …RIjj ˆ Rj …RIjj †h0 ‡ X where Ij(.) and Pj(.) are the equilibrium income and rental hedonic equations, respectively. Let RIjj , h*, and X* be the solutions to the above utility maximisation problem, specifying, respectively, the site within the region where the consumer will be located, RIjj , the kind of house he or she will live in, h*, and how much of the numeraire good he or she will be able to consume, X*. As a result of this, the income of the consumer will be: Ij* ˆ Ij(RIjj ), and the rent on the house will be: Pj* = Pj(h*,RIjj ) ˆ Rj* h*', where Rj* ˆ Rj(RIjj ). Equivalently, the problem can be stated in terms of an indirect utility function, Vj(.), where: Vj …RIjj † ˆ max U

JI…h; X; RIjj †

with respect to h, X, subject to Ij …RIjj † ˆ Rj …RIjj †h0 ‡ X

234 Regional Inequalities and EU Funding Utililty

Vj (RI ) Vj ( RIj 2 ) = vj 2

Vj (RIj 1 ) = vj 1

RI RIj 1

RIj 2

Figure 13.1 The relationship between utility and regional index

Equilibrium for a consumer in region j requires that the consumer's utility is the same at all sites in region j; that is, Vj(RIjj ) ˆ vjj for all j, where vjj is a constant for all sites in region j. This equilibrium condition implies that individuals in sites with a better regional index pay for it through a reduction in real income in the form of higher rent and lower wage income. The model described above is illustrated in Figure 13.1. The upward sloping curve, labelled Vj(RI), shows combinations of regional index values and the maximum utility that an individual in region j would enjoy with different vectors of variables …al1 ; a21 ; . . . ; aN1 †. For example, if our representative consumer were in region 1, whose characteristics are a12 ; a22 ; . . . ; aN2 †, the regional index value would be RIj1 and utility would be vj1. If he or she were located in region 2, whose characteristics are (a12, a22, . . ., aN2), the regional index value would be RIj2 and utility would be vj2, where vji is the maximum utility that a consumer in region j can enjoy in all locations of region i in equilibrium for all j and i, that is, Vj(RIji) ˆ vji. Figure 13.2 shows the quality of life ± indirect utility curves of three different consumers; that is, for j ˆ 1, 2, 3. To compare life across regions, we must compare the maximum utility that a consumer can enjoy in the regions under consideration. This can be done by looking either at the vjj values for all j, or at the vji values for

Dimitrios Giannias 235 Utililty

V3(RI ) V3(RI33) = v33 V2(RI )

V1(RI )

V1(RI11) = v11

V2(RI22) = v22

RI RI22

RI33

RI11

Figure 13.2 The relationship between utility and regional indexes

all regions i and a given j, where vjj ˆ Vj(RIjj) and vji ˆ Vj(RIji). The problem, however, is that the vjj and vji values are not available or are dif®cult to obtain. Moreover, the series of the vjj values shows the maximum utility that a consumer enjoys in his or her own region (for example, an Athenian in Athens, a Cretan in Crete and so on) and should not be used to compare life in different regions, because the utilities of different consumers cannot be compared. Therefore, only a vji based ranking is conceptually correct.1 For our comparison, the vji values are used for all i and a given j, and we compare the maximum utility that the consumer we have chosen (the consumer of region j) would enjoy if she or he were located in region i, for all regions i we want to compare. For example, if j = Athens, then we compare the maximum utility of an Athenian consumer in Athens, Crete, Peloponnisos and so on.

236 Regional Inequalities and EU Funding

It can be seen from Figure 13.1 that there is a monotonic relationship between vji and RIji. This implies that a ranking based on vji and one based on RIji will be identical. That is, both RIj2 > RIj1 and vj2 < vj1 imply that region 2 is preferred to region 1, or R(2) < R(1), where R(i) is the ranking of region i, i ˆ 1, 2, R(2) ˆ 1, R(1) ˆ 2; given that (1) the greater the maximum utility, the higher the position of a region on the relevant ranking (the lower the R(i) value), and (2) the higher the RI value, the higher the position of a region on the relevant ranking (the lower the R(i) value). Therefore, we can obtain a ranking of regions i based on the vji utility levels and the preferences of a consumer in region j by looking at his or her regional indices across regions, since the two rankings will be identical. The above implies that we can obtain a ranking of regions i based on the vji utility levels and the utility structure of a consumer in region j, if we are able to compute the RIji values. To apply the above theory, quality of life can be de®ned as follows: N RIji ˆ N kˆ1 …wkj aki †=kˆ1 …wkj † for i ˆ 1; 2; 3; . . . ; m

where aki is the kth variable of region i, wkj is the weight for the variable k of individual j, N is the number of variables considered and m is the number of regions being examined. The weights wkj are not necessarily the same across regions since individuals may put a different value on the various regional variables and perceive them in a different way. That is, the regional index of region i will depend on whose weights are used to compute it. For example, in the above formula the weights of a consumer in region j are used, which implies that RIji is the regional index value that the consumer in region j would assign to region i if she or he moved there. In general, the weights can take any value. For example, they can all be equal to 1/N or be assigned a theoretical one using principal component or survey results. The above analysis indicates that, in order to obtain a ranking of regions based on the maximum utility available to residents, we must look at the regional index values across regions of a particular consumer and take a ranking based on these regional indices, since the two rankings will be identical. This requires that, for each region, we substitute in the above formula the aki values of the region and compute the regional index value using the same weights, wkj of a consumer in region j.

Dimitrios Giannias 237

13.3 Computation of regional indices and the allocation of EU funding The regional index speci®ed in the above section can be computed to compare the 10 major regions of Greece; namely the Athens region, Sterea Ellada and Evia, Peloponnisos, the Ionian Islands, Epirus, Thessalia, Macedonia, Thrace, the Aegean Islands and Crete. To compute such indices, we usually consider a variety of variables. The variables that were available for all regions for 1995 (Epilogi, 1996 issue) and were eventually considered are the following: 1. Passenger cars per capita (weight: 0.89); 2. Telephones per capita (weight: 0.61); 3. Number of students in elementary and high school per capita (weight: 0.54); 4. Number of doctors per capita (weight: 0.77); 5. GNP per capita (weight: 0.91); 6. Electricity consumption per capita (weight: 0.77); 7. Unemployment (weight: 0.82); 8. Births per capita (weight: 0.22); 9. Marriages per capita (weight: 0.31); 10. Arable land per capita (weight: 0.17). All these variables are scaled from 0±100 and are taken into account when computing our regional index for each of the regions in question. The index is weighted and the weight of each variable is shown in parenthesis in the above list. The weights are based on an opinion survey conducted in Greece in the spring of 1996. For this survey, we asked 70 Greek experts to evaluate the information provided by each one of the above variables relevant to the development of and quality of life in a region. The scaling of the variables is such that the value 100 is reserved for the region with the `best' value (depending on the variable considered) and 0 denotes the `worst'. All the variables considered give an indication of some aspect(s) of the external environment of a region, and hence some indication of the standard of living there and the stage of development of the region compared with that of others. The contention is that, if the value of the index is relatively high, then the external environment in the region is relatively good, life in it is relatively better and its development is relatively more advanced. The results and rankings based on the regional index values, are presented in Table 13.1.

238 Regional Inequalities and EU Funding Table 13.1 Results and ranking of regional indexes (RI) Region Athens Sterea Ellada and Evia Peloponnisos Ionian Islands Epirus Thessalia Macedonia Thrace Aigean Islands Crete

RI

RI based ranking

80.20 32.32 34.05 25.21 23.33 40.90 44.46 26.30 29.17 44.96

1 6 5 9 10 4 3 8 7 2

Source: Epilogi (1996).

The RI values in Table 13.1 also provide an indication of regional inequalities. The Athens region has an RI value of just over 80, while the second highest RI value (that for Crete) is less than 45. Epirus is the only region with an RI value of below 25. So, according to the variables considered and our index, we can see that the Athens region is quite different from (unequal to) all the others, which can be classi®ed into four homogeneous groups of regions according to their RI values:

. . . .

Group A: Crete, Macedonia and Thessalia; Group B: Sterea Ellada and Evia, and Peloponnisos; Group C: the Aegean Islands, Thrace and the Ionian Islands; Group D: Epirus.

Hence, it is clear that convergence has not been attained, since inequality among the regions (as described/summarised by the RI values) is rather signi®cant. To investigate the structure of this inequality, we compute the coef®cient of variation, CV, for each of the ten variables included in our analysis, where: CV…Xi † ˆ 100‰Xi

m…Xi †Š2 =m…Xi †Š1=2

where m(Xi) is the mean of the ith variable, and i = 1, 2, 3, . . ., 10. The values of the coef®cients obtained are CV(X1) = 85, CV(X2) = 58, CV(X3) = 42, CV(X4) = 60, CV(X5) = 45, CV(X6) = 77, CV(X7) = 99, CV(X8) = 63, CV(X9) = 79, CV(X10) = 95. According to these values, the variables can be grouped in the following way: 1. Passenger cars per capita and number of doctors per capita: coef®cient of variation value 95±100;

Dimitrios Giannias 239

2. Births per capita, electricity consumption per capita and number of students in elementary and high school per capita: coef®cient of variation value 75±85; 3. Arable per capita, unemployment per capita and telephones per capita: coef®cient of variation value 55±65; 4. Marriages per capita and GNP per capita: coef®cient of variation value 40±45. The inequalities among the variables in groups 1 and 2 are rather large, while those in groups 3 and 4 are quite moderate. Thus, the regional index allows us to draw conclusions about average inequality among the ten regions of Greece.

13.4

The EU funding

The per capita funds allocated to the 10 regions of Greece for the 1994±9 period from the EU structural funds are shown in Table 13.2. Given that integration is the EU's main concern nowadays, it is fair to ask whether the criterion that funding should be directed to the regions that need it most is being satis®ed. This criterion is justi®ed in terms of ef®ciency and fairness, and would boost economic convergence within Greece. Table 13.2 gives a ranking based on the per capita EU funding received by each of the Greek regions. The criterion will be met if the sum of the regional index and the per-capita-based rankings equal N + 1 for each region, where N is the number of regions. However, this condition is

Table 13.2 Results and ranking of per capita EU funding and regional indexes Region Athens Sterea Ellada and Evia Peloponnisos Ionian Islands Epirus Thessalia Macedonia Thrace Aegean Islands Crete

EUF* 91.49 146.75 120.39 188.05 303.31 226.70 147.55 597.71 460.25 239.33

* EU per capita funding (ecu per person). Source: Epilogi (1996).

EUF-based ranking 10 8 9 6 3 5 7 1 2 4

Sum of RI and EUF-based rankings 11 14 14 15 13 9 10 9 9 6

240 Regional Inequalities and EU Funding

satis®ed only for the Athens region. If the sum of the two rankings is less than N ‡ 1 for a region, the region is receiving relatively more than it deserves. This is case with regard to Thrace, Thessalia, Macedonia, the Aegean Islands and Crete. On the other hand, if the sum of the two rankings is greater than N ‡ 1, the region receives relatively less than it deserves, as is the case with the Ionian Islands, the Athens region, Sterea Ellada and Evia, Peloponnisos and Epirus. Table 13.3 Results and ranking of actual and planned spending, 1994±95, ecus per capita Region Athens Sterea Ellada and Evia Peloponnisos Ionian Islands Epirus Thessalia Macedonia Thrace Aegean Islands Crete

Ecus per capita 0.65 0.46 0.31 1.00 0.72 0.65 0.63 0.58 0.52 0.72

The above indicates that the allocation of EU funds may not support economic convergence in Greece, since some regions seem to receive more funding than they deserve and some less. To ascertain the overall situation, we proceed in the following way. If the available funding and its allocation support convergence, the regional index (RI) and the EU per capita funding (EUF) values for each region should be negatively correlated. This correlation has been computed and found to be equal to ±0.5053.

13.5

Conclusions

This chapter has analysed the decision-making process of consumers and concludes that a regional index should be weighted by a set of weights that hold constant across regions. This result of our theoretical analysis was subsequently used to compare living standards across Greece and identify inequalities among the regions. For this comparison, data from the annual issue of Epilogi (1996) were used to obtain information on variables that affect some aspect of life in the 10 regions of Greece. Although the variables included in our index provide an adequate picture

Dimitrios Giannias 241

of important characteristics of the regions considered, information on other variables is needed to estimate global regional indices. The analysis indicates that, according to the adapted regional index, the Athens region is at the top of the ranking, followed by Crete, Macedonia, Thessalia, Peloponnisos, Sterea Ellada and Evia, the Aegean Islands, Thrace, the Ionian Islands and Epirus. In addition to the above, a comparison of the regional index values with per capita EU funding indicates that the allocation of funds supports the convergence process. Finally, Table 13.3 shows that the ratio of actual versus planned spending of EU funds was relatively low in all cases for the 1994±95 period, which indicates poor project management capabilities on the part of the local (regional) bodies that handled these funds. The correlation between the actual planned spending ratio and the regional index is relatively low and negative (±0.0277), indicating that the better developed regions did not adhere to their time schedules for project implementation. Moreover, the correlation between the actual planned spending ratio and EU funding per capita is negative (±0.0134), indicating that those regions which receive a relatively larger proportion of EU funding should devote some of their funds to improving the ef®ciency of their project management. These preliminary results of the correlation analysis should be extended to identify the nature of the relationship among the population variables, as well as possible distortions that may be incorporated into the results due to the absence of other important variables. Notes 1. A ranking based on the vjj values can be obtained if consumer preferences are identical in all regions and we know the positions of the V(RI) curves. Suppose that consumer preferences are identical and that the V(RI) curves are the ones given in Figure 13.2; note that the curves are not the same because the price of good X may vary across regions. For example, if the information in Figure 13.2 were available, we could conclude that region 3 is preferred to 1 and that region 1 is preferred to 2; that is, R(3) < R(1) < R(2), since v33 > v11 > v22, where R(i) is the ranking of region i, i ˆ 1, 2, 3, R(3) ˆ 1, R(1) ˆ 2, R(2) ˆ 3. However, as can be seen from the example, there is not a monotonic relationship between RIjj and vjj. This implies that a ranking based on RIji, which is observable, will not necessarily be identical to one based on vji , which is unobservable, since RI11 > RI33 > RI22, which implies that, according to the RI criterion, the ranking should be R(1) < R(3) < R(2), where R(1) ˆ 1, R(3) ˆ 2, R(2) ˆ 3.

References Bellante G. (1979) `The North±South Differential and the Migration of Heterogeneous Labour', American Economic Review, vol. 69, no. 1 (March). Blomquist, M., G. Berger and J. Hohen (1988) `New Estimates of Quality of Life in Urban Areas', American Economic Review, (March), pp. 89±107.

242 Regional Inequalities and EU Funding Blomquist, M., W. Berger and Waldner (1985) `Quality of Life: Ranking Urban Areas Using a Wage and Rent Based Index', University of Kentucky, Department of Economics, Working Papers, no. E-85 (Kentucky: University of Kentucky). Eberts, G. and J. Stone (1986) `Metropolitan Wage Differentials: Can Cleveland Still Compete?' (Cleveland: Federal Reserve Bank of Cleveland). Epilogi (1996) Annual issue. European Omnibus Survey (1988) `Les EuropeÂens et l' environnement en 1988' (Brussels: European Omnibus Survey). Giannias, D. (1996) `Regional amenity and productivity classi®cations within the EU', European Environment 96 Conference Proceedings. Gould, G., G. Gardner, D. DeLuca, A. Tiemann, L. Doob and J. Stolwijk (1988) Perceptions of technological risks and bene®ts (London: Russell Sage Foundation). Hope, G. and J. Parker (1990) `Environmental indices for all ± the need for a monthly index', Energy Policy, May, pp. 312±9. Hope, G., and J. Parker (1995) `Environmental indices for France, Italy and the UK', European Environment, vol. 5, no. 1. Hope G., J. Parker and S. Peake (1991) `A pilot index for the UK ± results of the last decade', Statistical Journal of the UN Economic Commission for Europe, vol. 8, no. 1, pp. 85±107. Hope, G., J. Parker and S. Peake (1992) `A pilot environmental index for the UK in the 80s', Energy Policy, April, pp. 335±43. Johnson, J., (1983) `Inter-metropolitan Wage Differentials in the United States', in Jack E. Triplett (ed.), The Measurement of Labour Cost (Chicago: University of Chicago Press). Ott, G. (1978) Environmental indices: Theory and practice (Ann Arbor, MI: Ann Arbor Science).

Part 4 Challenges for an Integrated Europe

This page intentionally left blank

14

Economic Integration and the Future of the Welfare State in the European Union Eleni Paliginis

14.1

Introduction

This chapter discusses the development of the welfare state in the EU member states, and examines the effects that globalisation and changing socioeconomic patterns have produced since the 1980s. It distinguishes between the different social regimes and economic structures within the EU, and examines the contradictions that exist in the establishment of EU social policies. It is argued that although in the short term it is dif®cult to adopt a single model of social protection, the creation of a cohesive European Union should involve at least a minimum social safety net for all citizens of the EU. The EU has undergone a fundamental transformation in recent years, and since the 1980s there has been a rapid stream of measures aimed at deepening integration. The Single European Market (SEM) programme provided the framework for economic integration, while European Monetary Union (EMU) is expected to enhance it. Genuine integration, however, necessitates a serious examination of the role and extent of welfare provisions in the EU as a whole. The welfare state is deeply embedded in the core EU countries and welfare provisions have farreaching social and economic consequences. The welfare state was born in the core countries of Europe and re¯ected the sociopolitical structures of the post-war period. Despite signi®cant differences in the structure and ®nancing of the various welfare states, partly re¯ecting historical differences and local idiosyncrasies, in all the core countries there is a commitment to provide social protection to all individuals throughout their lifetime. In recent years, enormous pressures have been brought to bear on the welfare state as a result of the ageing

245

246 Economic Integration and the Welfare State

population, changes in household structures and globalisation, and because of the pressure on governments to meet the Maastricht criteria. Social protection in the peripheral countries of the EU has risen substantially in recent years, but is still limited by the economic structures of these economies. The extended family network traditionally was, and to some extent still is, the main source of welfare provision in these countries. A cohesive EU requires the development of a sense of citizenship; the feeling that all people in the EU are a part of this developing new community. This chapter examines the development of the welfare state in the EU countries, discusses the changes that resulted from the transition from Fordism to post-Fordism, explores the problems that have arisen out of economic but not social integration in the EU and suggests ways of creating an EU community.

14.2

The development of the welfare state

Although there was some welfare provision before the Second World War, the full development of the welfare state was a post-war achievement, mainly in the developed European economies. It was the result of a powerful working-class movement and a structure of production that enabled the state to fund it; it was the product of social democracy and Fordism and operated at the national level. At the time, the goals of equality, ef®ciency and economic growth were seen as both desirable and possible. The social structures upon which the welfare state was based were characterised by full employment of the male population, a long working life and the existence of the nuclear family. Fordism created full employment for semi-skilled ± mainly male ± workers. High levels of productivity allowed a rapid increase in wages and, for the ®rst time, enabled the working class to be consumers of the products of their labour. The norm was for the male breadwinner to work full time to support himself and his family. Workers would join the labour force in their mid-teens and remain employed until they retired in their mid-sixties. The parallel development of the welfare state supported and supplemented their incomes through cheaper housing and the provision of health care and education, and at the same time removed the ®nancial insecurity that had accompanied unemployment, old age and ill-health. The role of the welfare state was to: modify the play of market forces in at least three directions: ®rst by guaranteeing individuals and families a minimum income irrespective of

Eleni Paliginis 247

the market value of their property; second by narrowing the extent of insecurity . . .; and third by ensuring that all citizens without distinction of status or class are offered the best standards available in relation to a certain agreed range of social services (Briggs, 1961, p. 228). Full employment and a long working life meant that the welfare state was affordable in the Fordist period of the 1950s and 1960s. Furthermore, the continuing existence of the nuclear family, maintained ®nancially by the male worker and sustained by unpaid maternal contributions, minimised the demands on the welfare state. Keynesianism provided the regulatory framework for Fordism. Demand management allowed non-in¯ationary, demand-led growth to coincide with full employment, while institutional arrangements guaranteed that wage settlements would not undermine pro®tability. Although welfare provision expanded in all the core countries of Europe, there were differences in both the extent of provision and the form of funding. Historical, social and religious differences were important determinants of these differences, as were `the degree of class con¯ict and, especially, the strength and form of working class struggle, and the ability of the capitalist state to formulate and implement policies to secure the long-term reproduction of capitalist social relations' (Gough, 1979, p. 65). By 1960 the unweighted average social spending in the OECD countries was 14 per cent of GDP. Within OECD Europe, the highest level was in Germany, with just over 20 per cent of GDP, and the lowest was in Greece, with 8 per cent (data for Spain and Portugal are not available) (Hills, 1993). While the Fordist economic boom created the revenue needed for the expansion of the welfare state, the long period of full employment and the subsequent rise in wages, which eventually exceeded the productivity rate, contributed to stag¯ation and the crisis of the Fordist system. Furthermore, the expansion of the welfare state undermined some of the conditions that sustained Fordist accumulation, as it strengthened organised labour and institutionalised a rigid social wage (Jessop, 1991, p. 89). The oil crises of the 1970s deepened and strengthened the internal contradictions of Fordism.

14.3

The crisis of the welfare state in the 1980s

The 1980s was a period of profound economic, social and ideological change in the core European countries. Globalisation, changes in social

248 Economic Integration and the Welfare State

patterns and the restructuring of the EU were the most important events, and all had implications for the funding of the welfare state. 14.3.1 Economic changes Globalisation, involving the internationalisation of production, and technological progress had a serious effect on the structure of production and the labour market. The relocation of some aspects of production to less developed countries, together with increasing computerisation and robotisation, led to de-industrialisation and a restructuring of the core economies. De-industrialisation caused high levels of unemployment (Figure 14.1) with serious regional implications. Unemployment particularly affected the young, older workers and the unskilled. These trends were reinforced by the new ¯exibilisation and casualisation of the labour market. Part-time employment became widespread. Fixed-period contracts replaced permanent ones, causing increased uncertainty in the labour market. The new employment structures led to the so-called 30/30/40 economy, where the ®rst 30 per cent represents the of®cially or unof®cially unemployed, the second 30 per cent the marginalised and insecure, and the remaining 40 per cent a privileged group with `secure' full-time employment. The substantial increase in part-time employment between 1973 and 1993 mainly affected the core EU countries (Figure 14.2). It was driven by both demand and supply conditions in the labour market and had a gender bias (Figure 14.3). Companies took advantage of the fact that parttime workers only were entitled to limited social protection, while

25 20 15 10 5 0

B

DK

D

E

F 1970

Source: Eurostat.

GR 1980

I

IRL

L

NL

P

1993

Figure 14.1 Rates of unemployment in the EU, 1970±93 (per cent)

UK

Eleni Paliginis 249

35 30 25 20 15 10 5 0

B

DK

D

GR

E

1973

F

IRL 1983

I

L

NL

P UK

1993

Source: Employment Outlook (1995).

Figure 14.2 Part-time employment as a percentage of total employment

women, who often had to cope with domestic pressures as well, opted for the ¯exibility offered by this part-time market (Figure 14.3). The internationalisation of production weakened the ability of the state to in¯uence the economy through monetary and ®scal policies. Furthermore, there was a strengthening of the EU through the creation of the Single European Market (SEM) and a partial transfer of power from the nation-state to the EU.

70 60 50 40 30 20 10 0

B

DK

D

E

F Men

Source: Employment Outlook (1995).

GR

IRL

I

L

NL

P

UK

Women

Figure 14.3 Male and female part-time employment as a percentage of total employment, 1993

250 Economic Integration and the Welfare State 2500 2000 1500 1000 500 0

B

DK

D

E

F

GR IRL 1981

I

L

NL

P

UK

1991

Source: Eurostat yearbook (1995).

Figure 14.4 Single women with children

14.3.2 Social Changes Changes in the labour market were supplemented by changes in the family structure. By now the two pillars of the postwar social structure ± full employment and the nuclear family ± were breaking down, intensifying the demands made on the welfare state. The number of single-parent families rose, increasing the dependence of women on the welfare state either to maintain themselves and their children, or for the provision of good quality child care (Figure 14.4). Improvements in the standard of living and medical care led to increased life expectancy. At the same time, there was a decline in female fertility, and taken together these two factors resulted in a rise in the average age of the population (the ageing population). There is evidence that when female employment is linked to inadequate social provisions there is a lowering of female fertility. Evidence from Italy and Spain (low fertility/few social provisions) and the Scandinavian countries (high fertility/high-quality welfare provisions) strengthens this view. The ®nancial implications are obvious: a signi®cant increase in dependency on the welfare state. This period also saw an increase in income inequality and poverty in a number of EU countries. The Gini coef®cient, which measures income inequality, rose signi®cantly. The fastest rise was in the UK, where the Gini coef®cient rose from 29 per cent in 1986 to more than 33 per cent in 1990. Similarly rising levels of inequality were experienced in the Netherlands, Sweden, Norway, Germany, France and Belgium. Countries that experienced a fall in inequality were mainly the peripheral countries

Eleni Paliginis 251

of the EU: Portugal, Spain, Italy and Ireland (Hills, 1995). Furthermore, measuring relative poverty at 50 per cent of the equivalent mean national expenditure, the UK, which was the ®rst EU country to adopt neoliberal policies, experienced an increase in relative poverty from 14.2 per cent to 18.2 per cent within ®ve years (Eurostat, 1991). Relative poverty was also a problem in the peripheral countries of the EU: Portugal, Greece, Spain and Ireland. In the core countries, household changes meant that poverty acquired a gender dimension as it mainly affected women and children. 14.3.3

Ideological changes

In the 1980s, the new right questioned the role of the welfare state with regard to its contribution to economic ef®ciency and growth as well as to its effect on personal liberty. According to the new right, the post-war welfare state necessitated the imposition of higher levels of taxation, which adversely affected the rate of pro®t and lowered the level of investment. In addition, the commitment to full employment was considered by the new right as having distortionary effects on the economy. This was based on their main premise that any tinkering with the markets would have only short-term bene®cial effects and, in the long term, would lead to high levels of in¯ation and unemployment. At the individual level, the main criticisms were that the welfare state reduced choice and limited personal freedom. Citizens were not in command of their own incomes and were deprived of choices. Furthermore, the existence of the welfare state sti¯ed initiative and created a dependency culture. As a result, unemployment in Europe was higher than in the United States and Japan, while the growth rate was considerably lower. During this period, as a result of the internationalisation of production, the nation state in the advanced economies underwent two types of transformation. First, the Keynesian welfare state gave way to the Schumpeterian `workfare state' whose objective was to `promote process, organisational, market innovations in open economies in order to strengthen as far as possible the structural competitiveness of the national economy by intervening on the supply side; and to subordinate social policy to the needs of labour market ¯exibility' (Jessop, 1994). Second, the globalisation of production undermined the ability of the state to follow traditional macroeconomic policies and led to the `hollowing out' of the nation state (ibid., 1994). This relative loss of autonomy led to a need for supranational cooperation and regulation. Within the EU, this took the form of an increase in the relative power of the EU, with the transfer of some aspects of sovereignty from the national governments to the EU.

252 Economic Integration and the Welfare State 35 30 25 20 15 10 5 0

B

DK

D

GR

E 1970

F

IRL

1980

I

L

NL

P

UK

1993

Source: Eurostat.

Figure 14.5 Total social expenditure as a percentage of GDP, 1970±93

14.3.4 The funding of the welfare state The stagnation of the 1970s led to a decline in tax revenues, while unemployment contributed to an increase in social expenditure. At ®rst, governments tried to maintain full employment by creating jobs, assisting industries and ®nancing unemployment. For a number of countries, this was their ®rst experience of running substantial budget de®cits in an attempt to resolve the crisis. There was a dramatic increase in social expenditure between 1970 and 1980 in the EU12, from 19 per cent to 24 per cent of GDP (Figures 14.5 and 14.6). Budget de®cits had to be covered by increases in taxation and/or internal or external borrowing. Despite criticisms of the welfare state, expenditure on social protection rose further in the 1980s, and by 1993 it represented 28 per cent of GDP, a rise of 4 percentage points since 1980. However, this rise did not signify a net increase in provision as it was affected by the recession of the 1980s, which depressed the level of GDP and increased unemployment. Excluding unemployment bene®t, EU social expenditure rose by 2 per cent between 1980 and 1993. This was hardly enough to cope with the consequences of the changing household structure and the ageing population (European Commission, 1995). During the recession of the 1980s, unemployment in the EU became higher than elsewhere in the advanced industrial world, rising to 10.8 per cent compared with 7.5 per cent in the United States. An overpaid, in¯exible workforce and a high degree of social protection in the EU was considered responsible for the high levels of unemployment and the uncompetitiveness of the EU economy. In the early 1990s, social protection amounted to 19 per cent of GDP in Canada and New Zealand,

Eleni Paliginis 253 12 10 8 6 4 2 0

B

DK

D

GR

E

1980–85

F

IRL 1988–90

I

L

NL

P

UK EU12

1990–93

Source: Eurostat.

Figure 14.6 Total growth of social expenditure, annual percentage change

15 per cent in the United States, 13 per cent in Australia and less than 12 per cent in Japan. The welfare state was widely perceived as contributing to the noncompetitiveness of the EU economy on two accounts: it provided a disincentive to work and rigidities in the labour market; and its ®nancing through taxation or debt was detrimental to growth. These criticisms were levelled against both the Swedish model (Lindbeck, 1994) and the European one (Dreze and Malinvaud, 1994). However, it is unreasonable to consider the cost of the welfare state without referring to its bene®ts. Social expenditure, through its impact on human capital, health and security, has a signi®cant effect on both the general welfare of citizens and the rate of growth. Pressure on governments to control social expenditure increased in the 1990s and led to the introduction of plans to increase the retirement age (Belgium, Germany and Greece), reduce pensions and reform unemployment bene®ts (Denmark, Germany, Ireland and Portugal) and reduce expenditure on health (France). In Germany alone, it is planned to remove a total of DM50 billion from the federal, state and local authority budgets. While globalisation and capital mobility increased the rate of pro®t accruing to capital, the effect on labour was declining employment and social provisions at the national level. Governments found it increasingly dif®cult to deal with these problems and international competition led to a downward spiral of social provision. These changes created the necessity for intervention and regulation at the supranational (EU) level.

254 Economic Integration and the Welfare State

14.4

Different forms of social protection in the EU

In addition to the socioeconomic transformations that took place in the core EU countries in the 1980s, the incorporation of the Mediterranean countries into the EU at that time led to greater divergence with respect to economic structures and welfare regimes. 14.4.1 Social Regimes Social protection in the EU countries has different patterns as a result of different political and historical in¯uences. Following the work of G. Esping-Andersen (1990), the EU countries can be grouped according to the form of social protection offered:

. The Scandinavian countries (Denmark, Finland and Sweden): coverage

is universal and mainly ®nanced by taxation. Unemployment insurance is managed by the trade unions, but heavily subsidised by the government; . UK and Ireland: social protection is universal but there is extensive use of means testing. The bene®ts are less generous than in the Scandinavian countries; . Germany, France, the Benelux countries and Austria: based on the Bismarckian model, the insurance principle determines bene®ts. Bene®ts are mainly ®nanced from contributions, which vary considerably between occupations; . The Southern European countries: social protection in these countries is rudimentary. Occupational funds play an important part in income maintenance ± occupational pension schemes are generous, but unemployment bene®ts are very low. Generally, these countries lack an integrated national income safety net. The family is still very strong and plays an important role in supporting the unemployed, the sick and the old. 14.4.2 The EU Periphery The peripheral countries have experienced a different economic trajectory from that of the core. While the core countries reaped the rewards of Fordist expansion and built the welfare state upon its success, the peripheral countries were dominated by an unproductive agricultural sector and a small and fragmented domestic capital base. Their failure to become integrated into the Fordist model deprived them of the social and ®nancial basis for a welfare state.

Eleni Paliginis 255

The integration of the peripheral countries into the EU had mixed effects. By 1993, per capita GDP was still lagging behind the EU average. In Greece it represented 49 per cent of the EU12 average, in Portugal 60 per cent, in Spain 76 per cent and in Ireland 78 per cent. The level of unemployment in Ireland (18.4 per cent) and Spain (21.5 per cent) was considerably higher than the EU average (10.2 per cent), while the lower levels in Greece (9.3 per cent) and Portugal (5 per cent) were partly the result of hidden underemployment. Youth unemployment (the 15±24 age group), while a serious problem in a number of countries, was particularly high in the periphery (Spain 43 per cent, Greece 29 per cent, Ireland 25 per cent). Neoclassicists see the free mobility of labour as the means of resolving the problem of regional unemployment. Wage differentials will entice the movement of capital and labour, thus resolving unemployment. The EU has embraced the principle of free movement, and seeks to facilitate it in the Social Chapter through the standardisation and harmonisation of professional quali®cations, by making social welfare transferable, and so on. But there are inherent problems with labour mobility. While labour migration is desirable for the most pro®table operations of capital, there is no clear proof that it increases workers' welfare. For professionals, migration is relatively easy (in the professional world the language problem is not always great, cultural differences are not very acute and important pecuniary and non-pecuniary advantages are often attached to migration), but for the working class the case may be different. Among unskilled or semi-skilled workers mobility can be a harsh economic necessity, but language dif®culties and cultural divergence make migration an undesirable prospect. Faini and Venturini (1994) have shown that there is `a home bias' effect; that is, there is a strong link between income growth and migration. As income grows, the frequency of migration increases as the ®nancial constraint is lifted, but as income rises further the home-bias effect becomes stronger and migration falls. Recent patterns of migration reinforce these ®ndings. In the 1960s and 1970s, labour movement from the periphery to the core countries of the EU was the norm, but migration contracted rapidly in the 1980s despite fears that the lifting of cross-border restrictions as a result of EU membership would create a wave of migration to the core countries. This was partly the result of changed market conditions in the core countries, but also because rising living standards in the periphery reduced the need and willingness of workers to migrate. Faini and Venturini (ibid.) point to the personal cost of migration. Free mobility of labour is desirable but it should re¯ect free choice. Workers' welfare should be increased by

256 Economic Integration and the Welfare State

improving their conditions at home, rather than forcing them to seek employment by migrating within the EU. If future labour mobility is not as high as predicted because of an increase in workers' preference to stay at home, alternative forms of assistance will be needed. Economic development of the south is the only means by which a real increase in welfare could be achieved. An increase in expenditure on education and training would signi®cantly assist the process of development. However, this would be a slow process, and in the meantime the differential in welfare provision would continue and the gap might even widen. The creation of a European citizenship, with common economic and social rights, will require the harmonisation of social policies, but funding the latter will be dif®cult for some countries. An increase in social expenditure in the peripheral countries, although badly needed, would be handicapped by serious internal constraints. Social expenditure in these countries increased faster than in the core in the 1980s, but as a percentage of GDP they are still lagging behind (Figure 14.6). A further increase in social provisions in the southern European countries would impose a serious strain on government ®nances and divert funds away from economic restructuring and higher economic convergence. Equally, even if these provisions were funded by the private sector, this would make the latter even more uncompetitive in the global market. In 1992 employers' social fund contributions amounted to almost 50 per cent of total contributions in Greece and Spain, and 40 per cent in Portugal, but they were 25 per cent in the UK, 20 per cent in Netherlands and less than 10 per cent in Denmark.

14.5

EU social policies

Structural differences, together with differences in social regimes, represent a serious threat to the economic integration of the EU. They hamper the free movement of labour and create real obstacles to the creation of a cohesive community. Although economic integration measures such as the SEM have been fully supported by the member states, the prospect of EU-wide social policies has elicited mixed reactions, with ideological differences seriously in¯uencing behaviour. The adoption of an EU model will be dif®cult, as any such scheme will have to involve either a `levelling up', which will make the southern economies uncompetitive, or a `levelling down', which will be unacceptable to the citizens of the core countries. Nonetheless, a minimum EU-wide safety net is necessary as a ®rst step towards the creation of a cohesive union.

Eleni Paliginis 257

The main questions to arise are: (1) at the state level, how can the welfare state respond to the socioeconomic changes; and (2) under all the constraints raised earlier, how can the EU address the issue of social protection? In the core countries, there is a widespread view that the welfare state, which was established to address the needs of the population in the immediate post-war period, does not fully re¯ect the values and needs of the post-industrial societies of the 1990s. A move away from passive income maintenance towards labour supply management is seen by some as the way forward. This would involve improved education and training. Such policies could be in line with the policies of neo-industrialism (Streeck, 1992). According to the latter, advanced industrial economies could remain competitive at the global level by competing on quality rather than price. This would involve a highly skilled labour force and the use of new technology. Thus, a high-wage policy would be compatible with effective competition in world markets, but would ®rst require appropriate institutional arrangements to be put in place. Although conditions in the developed economies could be suitable for these policies, the strategy may not be very relevant to the periphery in the short term. The prerequisites for high-quality, high-technology products and high-wage labour may not be present at the moment. This policy could therefore accentuate the existing dualistic markets in the core and the periphery. At the EU level, the process of globalisation and the welfare state have been addressed at two different levels. At the economic level the EU has attempted to stimulate growth and enhance the competitiveness of EU (mainly core) countries through the Single European Market programme, while at the social level it has dealt with social issues and peripherality through the structural funds and the introduction of the Social Chapter. 14.5.1

Structural changes

`Eurosclerosis' (the non-competitiveness of the European economies) was addressed by restructuring and increasing the competitiveness of European capital through the creation of the Single European Market and setting a framework for European Monetary Union (EMU). The Single European Market ± which was initiated in 1986 but only came into operation in 1992 ± was the EU's attempt to liberalise, restructure and increase the ef®ciency and competitiveness of capital relative to the United States and Japan. It involved the free movement of capital, commodities and labour. The restructuring of capital involved EU support of strategic industries, such as electronics and information technology, through the partial funding of R&D.

258 Economic Integration and the Welfare State

This would enable winners to emerge, and larger markets would allow them to take advantage of economies of scale. From this process, a stronger and more competitive European capital was expected to emerge. Whatever the long-term outcomes of this strategy, in the short term an adverse effect on employment was anticipated in sectors and economies where there was a concentration of ineffective industries. European Monetary Union was also expected to have a negative initial effect on the level of employment. The strict criteria set for convergence forced governments to sacri®ce employment and social protection in an effort to reduce government de®cits. 14.5.2 Social changes To supplement its economic reforms the EU reformed the structural funds and introduced the Social Charter (for details, see, EC, 1995). The structural funds, which include the Regional, Social and Agricultural Guidance Funds, were reformed and enhanced in the late 1980s. Their purpose is to assist the development of the peripheral countries, deal with the economic and social consequences of de-industrialisation and address social issues in the EU. In 1993 they represented almost 25 per cent of the EU budget (15 982 million ecus), but only 7 per cent of the EU budget was earmarked for the Social Fund ± a very small percentage of an already limited budget. The EU will need to divert more funds into this area to deal with the increasing problems. The idea of the Social Charter arose from the need to guarantee workers' social rights and control unfair competition between countries. The Charter of the Fundamental Rights of Workers ± a political declaration without the force of the law ± outlines the social and economic rights of European citizens.1 Although most of these rights were non-controversial, disputes arose in relation to (1) atypical forms of employment such as part-time employment and (2) the maximum duration of the working week and holiday entitlement. Furthermore, the principle of an `equitable wage' was interpreted as a minimum wage. Although this already existed in a number of EU countries, the Conservative UK government was vehemently against it. The UK's objection to the charter meant that it could not form part of the Maastricht agreement and it instead became a protocol, signed by the other 11 countries. The white paper on European social policy (EC, 1994) set the terms for a European social policy for 1995±9. In a number of ways, this policy was more progressive than some national schemes. It was based on the principles of equity, participation and consensus, and hence was not fully compatible with the free operation of market forces ± it provided a broader

Eleni Paliginis 259

vision of the interaction of the social and economic spheres, and signi®ed a move away from passive social policies to support the unemployed towards an active one offering training and reskilling. In the white paper there was an emphasis on diversity and a reassurance that the harmonisation of social policies was not the objective. However, unless some harmonisation takes place the impetus towards a real European community will be lost. According to the white paper, `The objective in the coming period must be to preserve and develop the European social model as a move towards the 21st century, in order to give the people of Europe the unique blend of economic well-being, social cohesiveness and high overall quality of life which was achieved in the post-war period' (ibid., p. 7). This could only be achieved by increased EU assistance in the social domain. A social safety net, concentrating on education and the reskilling of the workforce, would provide jobs and dignity to the unemployed and guarantee an education compatible with the needs of the new millennium. In fact, however, the EU's intervention in the social domain has been rather limited. States that are vying to increase their own competitiveness support the existence of the SEM and the principle of subsidiarity, but some prefer more limited EU intervention in the social ®eld. Nonetheless the only way forward is to establish at least minimum welfare provisions within the EU. Global competition could lead to a downward spiral of social provisions and present a real obstacle to the development of the EU periphery. As a starting point, an EU social welfare net is necessary.

14.6

Conclusions

The development of the welfare state in the postwar period was mainly the result of solidaristic attitudes within the nation state. It was an attempt to introduce social cohesion and provide a counterweight to socialism and fascism. It was a corrective mechanism for market forces. More recently, in the core countries increasing unemployment ± as a result of technological changes and globalisation, changes in the structure of the nuclear family and an increasingly ageing population ± has resulted in signi®cant ®nancial pressure on the welfare system. There is a real fear that competition between the EU countries for markets could lead to a downward spiral in social expenditure. In the peripheral countries there is a de®nite need for the development of a modern welfare state, but neither governments nor the private sector are able to fund it. The EU has interfered substantially in the economic ®eld ± it established the SEM and is working towards full monetary union ± but it has left social

260 Economic Integration and the Welfare State

welfare mainly to the national governments. The principle of subsidiarity has prevailed in this domain. This differential treatment has created a clear anomaly, as it goes against the desire for a true European community and has even created problems for the smooth operation of the SEM. Commitment to a European social welfare net would be an important step towards the creation of a more cohesive EU. Note 1. The main provisions included the right to free movement, fair remuneration, improved social protection, freedom of association, vocational training, equal rights for men and women, an adequate pension and for the disabled to receive training and to be integrated.

References Arestis, P. and E. Paliginis (1995) `Divergence and Peripheral Fordism in the European Union', Review for Social Economics, vol. L I I I , no. 2. Briggs G. (1961) `The Welfare State in Historical Perspective', Archives EuropeÂennes de Sociology, vol. 2. Dreze, J. and E. Malinvaud (1994) `Growth and Employment', European Economy, no. 1. Esping-Andersen, G. (1990) The Three Worlds of Welfare Capitalism (Cambridge: Polity). European Commission (1995) Social Protection in Europe (Luxembourg: European Commission). Eurostat (1991) A Social Portrait of Europe (Luxembourg: Eurostat). Gough, I. (1979) The Political Economy of the Welfare State (London: Macmillan ± now Palgrave). Faini, R. and A. Venturini (1994) `Migration and Growth: The Experience of Southern Europe', CEPR Discussion Paper no. 964 (London: CEPR). Hills, J. (1993) The Future of the Welfare State. A Guide to the Debate (York: Joseph Rowntree Foundation). Hills, J. (1995) Inquiry into Income and Wealth, vol. 2 (York: Joseph Rowntree Foundation). Jessop, B. (1991) `The Welfare State in the Transition from Fordism to PostFordism', in B. Jessop et al. (eds), The Politics of Flexibility. Restructuring State and Industry in Britain, Germany and Scandinavia (Aldershot: Edward Elgar). Jessop, B. (1994) `Post-Fordism and the State', in J. Amin (ed.), Post-Fordism. A Reader (Oxford: Blackwell). Lindbeck, A. (1994) Turning Sweden around (Cambridge, MASS: MIT Press). OECD (1995) Employment Outlook (Paris: OECD). Streeck, W. (1992) Social Institutions and Economic Performance (London: Sage).

15

Technological Change, Productivity and Economic Integration of the European Union George M. Korres

15.1

Introduction

Many studies have suggested that there is a close correlation between technological development and productivity (see, for example, Abramovitz, 1986; Fagerberg, 1987, 1988, 1994), and economists have analysed different possible views of why productivity growth has declined. These alternative explanations can be grouped into the following categories: (1) the capital factor; for instance, investment may have been insuf®cient to sustain the level of productivity growth; (2) the technology factor; for instance a decline in innovation might have affected productivity growth; (3) the increased price of raw materials and energy; (4) government regulations and demand policies that affect the productivity level; (5) the skills and experience of the labour force may have deteriorated, or workers may not work as hard as they used to; (6) the products and services produced by the economy have become more diverse; and (7) productivity levels differ greatly across industries. This chapter attempts to measure the relationship between technology and productivity, or more precisely, to investigate the correlation between technological development and the decline in productivity growth. We shall empirically test the technological and catching-up models using data for the EU member states.

15.2

Technology, productivity and economic growth

Abramovitz (1986), Kendrick (1991) and Schmookler (1966) have studied the interaction between technological change and productivity. In these studies, factor prices were used to weight the various inputs in order to obtain a measure of total input growth. The approach developed by

261

262 Technological Change, Productivity and Economic Integration

Abramovitz (1986), Denison (1962) and Solow (1957) involves the decomposition of output growth into its various sources, which can be de®ned as the growth accounting and residual method.1 Growth accounting tries to explain changes in real product and total factor productivity based mainly on a comparison between the growth of inputs (capital and labour) and the growth of output. One part of actual growth cannot be explained and has been classi®ed as `unexplained total factor productivity growth' (or the so-called residual). In particular, following the decomposition analysis by Solow (1957), many alternative factors can explain the path of economic growth. According to Solow's ®ndings, technology was responsible for 90 per cent of the increase in labour productivity in the United States in the twentieth century. The unexplained decline in productivity growth can thus be regarded as resulting from a collapse in technological activities. This may have happened because the availability of technological opportunities has been temporarily or permanently reduced. Furthermore, technological gap theories (Abramovitz, 1986; Fagerberg, 1987, 1988, 1994) relate the technological level and innovation activities to the level of economic growth. According to these theories, countries where more innovation activities take place tend to have a higher level of value added per worker (or a higher per capita GDP). Following the technological-gap argument, it would be expected that the more technologically advanced countries would also be the most economically advanced (in terms of innovation activities and per capita GDP). The level of technology in a country cannot be measured directly, but an approximation measure can be used to obtain an overall picture of the set of techniques invented or diffused by that country. We shall use real per capita GDP as an approximate productivity measure. The most representative measures for technological inputs and outputs are patent activities and research expenditure. Catching-up theory (Abramovitz, 1986; Fagerberg, 1987) starts with the investigation of growth performance. The main idea is that large differences in productivity among countries tend to be due to unexpected events (for instance, wars). According to these studies, the only possible way for technologically weak countries to converge or catch up with the advanced countries is to copy their more productive technologies. The outcome of the international innovation and diffusion process is uncertain; the process may generate a pattern where some countries follow diverging trends or one where countries converge towards a common trend. In this literature, economic development is analysed as a disequilibrium process charac-

George M. Korres 263 Table 15.1 Patterns of growth

OECD EU USA Japan Category A Category B

GDP growth (1960±90)

GDP per head (1960±90)

BERD (1975±85)

GERD (1975±85)

3.60 3.68 (3.4) 3.20 6.30 3.17 4.20

2.70 2.80 (3.10) 2.00 5.30 2.67 3.54

6.00 4.50

5.00 4.00

5.90 9.80 4.97 5.44

5.00 8.00 4.31 6.42

Notes: the numbers in brackets are for the 12 member states. GDP = gross domestic product, BERD = business expenditure on research and development; GERD = government expenditure on research and development. Source: The table is derived from calculations using various statistical data from the OECD.

terised by two con¯icting forces: (1) innovation, which tends to increase economic and technological differences between countries; and (2) diffusion (or imitation), which tends to reduce them. Technological gap theories are an application of Schumpeter's dynamic theory. Table 15.1 illustrates the pattern of growth using different technological and economic measures (in our case GDP growth, per capita GDP, growth of business expenditure on research and development, and government expenditure on research and development). We can classify the EU member states into two categories: the more advanced countries (Belgium, Luxembourg, Denmark, Germany, France, Italy, the Netherlands and the UK), and the less favoured countries (Greece, Ireland, Portugal and Spain). According to various research measures (such as GERD and BERD), it is obvious that there is a large gap between the ®rst and second groups of member states ± although the latter have increased their technological investment and research activities, they still lag far behind the former. Table 15.2 presents a clearer picture of the technological differences between the member states. The ®rst two columns indicate that although the less advanced member states have increased their research expenditure and related investment, the technological gap between the less and more advanced member states is still widening. Patent statistics measure innovation activities, while the R&D data measure both innovation and imitating activities. From the patent index, it seems that the (absolute) technological difference is larger than is indicated by research expenditure.2

264

Table 15.2 GERD, GFCF (percentage growth rates), RDI and external patent applications GERD/GDP (1981±90) USA Japan Germany France Italy UK Spain Netherlands Belgium/Luxembourg Denmark Greece Portugal Ireland OECD

3.2 4.9 2.3 3.4 6.9 ±1.0 7.4 1.4 2.9 3.3 13.0 5.3 2.0 3.3

GERD/GFCF (1981±90) 3.9 7.6 4.8 6.8 10.0 ±2.4 10.8 1.6 9.3 ±0.9 17.5 10.1 12.1 5.2

GFCF/GDP (1981±90) ±0.7 ±2.5 ±2.5 ±3.2 ±2.8 1.4 ±3.1 ±0.3 ±5.9 4.2 ±3.8 ±4.4 ±9.0 ±1.8

RDI (Year)

External patent (1989)

1.33 (1988) 1.95 (1988) 1.75 (1990) 0.96 (1988) ± 1.06 (1988) 0.30 (1987) ± 1.13 (1988) 0.68 (1987) 0.08 (1988) 0.11 (1986) 0.40 (1986) ±

149 707 74 363 93 974 36 773 16 596 37 553 1 785 13 496 5 001 4 480 169 93 585 ±

Notes: The research and development intensity index (RDI) is de®ned as follows: (BERD/GDP)* 100, where BERD is business expenditure on R&D. The numbers in brackets in column 3 are the most recent years for which relevant data are available. GERD/GDP = the ratio of gross expenditure on research and development to gross domestic product. GERD/GFCF = the ratio of gross expenditure on research and development to gross ®xed capital formation. RDI = research and development intensity (the number of external patent applications), GFCF/GDP = the ratio of gross ®xed capital formation to gross domestic product. Source: OECD statistical data.

George M. Korres 265 Table 15.3 Productivity: average percentage changes at annual rate

USA 1960±73 1973±9 1979±90 Japan 1960±73 1973±9 1979±90 Canada 1960±73 1973±9 1979±90 EU 1960±73 1973±9 1979±90 OECD 1960±73 1973±9 1979±90

TFP

LB

CB

LB agr.

LB ind.

LB man. LB serv.

1.6 ±0.4 0.4

2.2 0.0 0.8

0.2 ±1.1 ±0.4

4.1 1.1

2.0 ±0.1

3.4 0.9

1.5 0.0

6.0 1.5 2.0

8.6 3.0 3.2

±2.5 ±3.1 ±1.7

6.3 0.9 2.3

8.9 3.5 4.5

10.3 5.0 5.8

6.1 2.2 1.5

2.0 0.7 0.3

2.8 1.5 1.5

0.5 ±0.7 ±2.0

5.8 2.5 4.3

3.6 ±0.3 2.2

4.4 1.3 2.3

2.6 1.0 0.9

3.3 1.4 1.2

5.0 2.6 2.1

±0.4 ±1.4 ±0.7

6.1 4.0 5.0

4.7 2.6 2.5

5.5 3.1 2.6

3.1 1.8 0.6

2.9 0.6 0.9

4.1 1.4 1.6

±0.4 ±1.5 ±0.8

4.9 3.2 3.5

3.9 1.8 2.7

4.9 2.6 3.3

2.6 1.1 0.8

Notes: TFP = total factor productivity (a weighted average of the growth in labour and capital productivity). CP = capital productivity (the ratio of output to capital). LB = labour productivity (output per employed person). The indexes of agriculture, industry, manufacturing and the service sector are derived in a similar way. Source: Various OECD publications.

Table 15.3 shows the average annual percentage growth rates of productivity.3 Using these measures, it is obvious that productivity growth rates have declined rapidly since the 1960s. We can easily investigate the relationship between these two approximate measures using cross-section data on average growth rates in the period 1973±90 for the EU member states. The results are presented in Table 15.4. Whatever the form of the independent variable, a positive relationship exists between productivity and gross expenditure on R; this can be interpreted as due to the poor reliability of gross research expenditure data as an explanatory variable of innovation activities. As expected, the best results are obtained for the logarithmic models, which imply a steeper curve. Patenting data re¯ect the innovation process, while both the research indexes re¯ect the imitation and the innovation process. The research and development data re¯ect imitation, innovation and diffusion activities. The relation between productivity (as

Relation between productivity and patents: GDPPC = 5547.23 + 529.695EXPA t = (7.455) (4.544). R2 = 0.28 (adj. df 0.22). DW = 2.05 Rho (autocorrelation coef®cient) = ±0.0962. t = ±0.344. SEs and variance shown are heteroskedastic consistent estimates The logarithm models: LGDPPC = 8.068 + 0.564LEXPA t = (21.099) (2.336). R2 = 0.23 (adj. df 0.16). DW = 1.69 Rho (autocorrelation coef®cient) = 0.705. t = 0.223. SEs and variance shown are heteroskedastic consistent estimates LLGDPPC = 2.160 + 0.783LLEXPA t = (128.747) (2.868). R2 = 0.31 (adj. df 0.24). DW = 1.81 Rho (autocorrelation coef®cient) = ±0.032. t = ±0.101. SEs and variance shown are heteroskedastic consistent estimates The relation between productivity and gross expenditure on research and development: GDPPC = 9584.54 ± 366.10GERD t = (5.738) (±1.324). R2 = 0.76 (adj. df 0.52). DW = 1.644 Rho (autocorrelation coef®cient) = 0.131. t = 0.475. SEs and variance shown are heteroskedastic consistent estimates The logarithm models: LGDPPC = 9.424 ± 0.384LGERD t = (25.721) (±1.529). R2 = 0.091 (adj. df 0.02). DW = 1.24 Rho (autocorrelation coef®cient) = 0.347. t = 1.352. SEs and variance shown are heteroskedastic consistent estimates LLGDPPC = 2.200 ± 0.0647LLGERD t = (141.439) (±1.586). R2 = 0.087 (adj. df 0.017). DW = 1.177 Rho (autocorrelation coef®cient) = 0.385. t = 1.525. SEs and variance shown are heteroskedastic consistent estimates Notes: GDPPC = GDP per capita average for the period 1973±90, absolute values in constant (1985) prices (US$000) for per capita GDP. EXPA = average annual growth rates for the period 1973±90 for external patent applications. GERD = average annual growth rates for the period for gross expenditure on research and development. LGDP, LPROD, LEXPA, LGERD, LEXP, LINV, LTRD, LLGERD and LLGDPCP are the above variables in logarithmic and in loglogarithmic form.

266

Table 15.4 Relationship between productivity and innovation for twelve EU member states, 1973±90

George M. Korres 267

80

GDP PER CAPITA EXT. PATENTS

60

40

20

UK.

SW.

SP.

PROT.

NETH.

ITALY

IRE.

GR.

GER.

FR.

FIN.

DEN.

BELG./LUX.

AUS.

0

Figure 15.1 Per capita GDP and patenting in EU member states

measured by per capita GDP) and innovation activities should be expected to be log linear, rather than linear, and steeper for the patent data than for the index based on research data.4 The correlation between productivity and patenting is much closer than between productivity and research expenditure. Figure 15.1 shows per capita GDP and the external patent applications of the EU member states for the period 1973±90. When conducting an econometric analysis of the technological gap models, it is important to include the most relevant variables. For the level of productivity, as a proxy we can use real GDP per capita (GDPPC). For the national technological level we can use some approximate measures; for instance; we can again use the traditional variables of technological input and technological output (GERD and EXPA). Following the model of Fagerberg (1987, 1988, 1994), we can test the basic technological gap model (with and without these variables), re¯ecting structural change, in order to determine the degree to which these variables have added something to the other explanatory variable of the model. We shall use external patent applications (EXPA) and gross expenditure on research and development (GERD) as proxies for the

268 Technological Change, Productivity and Economic Integration

growth of national technological activities, and GDP per capita (GDPPC) (in absolute values at constant prices) as a proxy for the total level of knowledge appropriated in the country (or productivity). Investment share (INV) has been chosen as an indicator of an improvement in the capacity for economic exploitation of innovation and diffusion; the share of investment may also be seen as the outcome of a process in which institutional factors take part (since differences in the size of investment share may re¯ect differences in the institutional system). For structural change, we use as an approximation changes in the share of exports and agriculture in GDP.5 Technological gap models, as developed here, can say little about how to boost the level of innovation activities or improve diffusion and innovation. We test the following versions of the models: GDP …or PROD† ˆ f ‰GDPPC; EXPA …or GERD†; INVŠ …the basic model† …15:1† GDP …or PROD† ˆ f ‰GDPPC; EXPA …or GERD†; INV; EXPŠ

…15:2†

GDP ˆ f ‰GDPPC; EXPA …or GERD†; INV; TRDŠ

…15:3†

Since annual observations are strongly affected by short-term ¯uctuations, average values of the variables for the period 1973±90 are calculated.6 The ®rst model may be regarded as a pure supply model, where economic growth is a function of the level of economic development GDPPC (GDP per capita with a negative expected sign), the growth of patenting activity (EXPA with a positive sign) and investment share (INV with a positive sign). However, it can be argued that this model overlooks differences in overall growth rates between periods due to other factors, especially differences in economic policies. Table 15.5 shows the model for twelve EU member states including, as additional variables, exports (as a share of GDP) and the terms of trade; this indicates that growth has been in¯uenced by changes in the terms of trade (terms of trade shock). The export variable also has the expected sign and the results support the hypothesis of structural change as a source of economic growth. The second model takes account of structural changes, using as a proxy the share of exports in GDP. The third model uses an additional variable that re¯ects changes in the macroeconomic conditions and suggests that growth rates are seriously affected by changes in the terms of trade. The models are tested for twelve EU member states. The basic model is tested for the variables of GDP, GDP per capita, external patent

Table 15.5 The basic model tested for twelve EU member states, 1973±90 The basic model including patents: GDP = 2.824 ± 0.002GDPPC + 0.10EXPA + 0.027INV t = (1.53) (±3.30) (2.30) (0.32). R2 = 0.52 (adj. df 0.39). DW = 1.52 Rho (autocorrelation coef®cient) = 0.385. t = 1.475. SEs and variance shown are heteroskedastic consistent estimates The logarithm model: LGDP = 1.499 ± 0.384LGDPPC + 0.155LEXPA + 0.806LINV t = (0.593) (±2.569) (0.930) (1.340). R2 = 0.56 (adj. df 0.42). DW = 1.36 Rho (autocorrelation coef®cient) = 0.297. t = 0.985. SEs and variance shown are heteroskedastic consistent estimates. The basic model including patents: PROD = 0.453 ± 0.00015GDPPC ± 0.0198EXPA + 0.174INV t = (±0.386) (±3.979) (±0.245) (3.012). R2 = 0.64 (adj. df 0.54). DW = 1.49 Rho (autocorrelation coef®cient) = 0.301. SEs and variance shown are heteroskedastic consistent estimates The logarithmic model: LPROD = ±0.566 ± 0.384LGDPPC ± 0.131LEXPA + 1.558LINV t = (±0.220) (±2.519) (±0.770) (2.541). R2 = 0.75 (adj. df 0.66). DW = 1.38 Rho (autocorrelation coef®cient) = 0.241, t = 0.786. SEs and variance shown are heteroskedastic consistent estimates The basic model including gross expenditure on research and development: GDP = 1.775 ± 0.00129GDPPC + 0.0142GERD + 0.0646INV t = (0.92) (±1.86) (0.21) (0.75). R2 = 0.40 (adj. df 0.24). DW = 2.30 Rho (autocorrelation coef®cient) = ±0.153. t = ±0.539. SEs and variance shown are heteroskedastic consistent estimates The logarithm model: LGDP = 0.619 ± 0.275LGDPPC + 0.00625LGERD + 0.837LINV t = (0.246) (±2.098) (0.0396) (1.408). R2 = 0.47 (adj. df 0.33). DW = 2.38 Rho (autocorrelation coef®cient) = ±0.228. t = ±0.815. SEs and variance shown are heteroskedastic consistent estimates 269

270

Table 15.5 The basic model tested for twelve EU member states, 1973±90 (continued ) The basic model including gross expenditure on research and development: PROD = 0.349 ± 0.00018GDPPC ± 0.0716GERD + 0.168INV t = (0.231) (±3.413) (0.933) (2.677) R2 = 0.66 (adj. df 0.57). DW = 1.43 Rho (autocorrelation coef®cient) = 0.301. SEs and variance shown are heteroskedastic consistent estimates The logarithmic model: LPROD = ±0.404 ± 0.421LGDPPC ± 0.0345LGERD + 1.568LINV t = (±0.130) (±2.585) (±0.176) (2.126). R2 = 0.61 (adj. df 0.50) DW = 1.79 Rho (autocorrelation coef®cient) = ±0.0131. t = -0.0402. SEs and variance shown are heteroskedastic consistent estimates Notes: The 12 EU states include the three prospective member states. GDP = annual average growth rates for real gross domestic product. PROD = annual average growth rates for product (de®ned as labour product GDP per person employed). GDPPC = average absolute values in constant (1985) prices (US$ 000) for GDP per capita. EXPA = annual average growth rates for external patent applications. GERD = annual average growth rates for gross expenditures on research and development. EXP = annual average growth rates for exports as a share of GDP. INV = annual average growth rates for investment as a share of GDP. TRD = annual average growth rates for the terms of trade. LGDP, LPROD, LEXPA, LGERD, LEXP, LINV and LTRD are the above variables in logarithmic form.

George M. Korres 271

applications and investment as a share of GDP. The results are presented in Table 15.6. The explanatory power (or the overall goodness of ®t of the estimated regression models) is not very high, but this is not surprising for crosssectional data. However, there is a problem with interdependence between the variables.7 For this reason, we shall focus on the relationship between productivity and innovation. Most of the variables have the expected signs. Some of the results presented in Table 15.6 are not statistically signi®cant. In addition, the results were not sensitive to the choice of innovation proxy; as before, better results are obtained for the loglinear models. The introduction of the terms of trade variable into the basic model led to a negative sign for the innovation variables (GERD and EXPA); this indicates that the economic slowdown after 1973 can be better explained by a terms of trade shock. However, some of the results are not statistically signi®cant and the explanatory power is not very high. Finally, we can divide the EU member states into two categories. The ®rst category of countries consists of the technologically advanced member states (Belgium, Luxembourg, Denmark, Germany, France, the Netherlands, Italy and the UK), and the second category comprises the less technologically advanced countries (Greece, Ireland, Portugal and Spain). The results for the ®rst category are shown in Table 15.7 and those for the second category in Table 15.8. In both cases, we used the same approach, ®rst testing the basic model and then introducing the terms of trade and export variables. It is worth noting that for the technologically advanced member states, the estimated coef®cients display the expected signs except for exports (EXPA) and gross expenditure on R&D (GERD). The results do not support the hypothesis that structural changes are independent causal factors of economic growth. These results can be seen as supporting the view that the in¯uence of a change in outward orientation on growth depends on the international macroeconomic conditions (since random shocks and crises and slow growth in world demand in the 1970s restrained the growth of outward-oriented countries). For the less technologically advanced countries, the estimated coef®cients display the opposite signs from those found for the advanced countries, the only exception being patents. According to these results, the coef®cient of investment (INV) has the wrong sign. In terms of data, it is not dif®cult to see why this happened. For instance, during the whole period under examination only the more advanced countries had a large capacity for innovation activities; they had already established a

Table 15.6 Model with exports and terms of trade for twelve EU member states, 1973±90

The logarithm model: LGDP = ±0.040 ± 0.381LGDPPC + 0.202LEXPA + 0.300LEXP + 0.930LINV t = (±0.0150) (±2.641) (1.227) (1.299) (1.583). R2 = 0.64 (adj. df 0.46) DW = 1.72 Rho (autocorrelation coef®cient) = 0.309, t = 1.03. SEs and variance shown are heteroskedastic consistent estimates Test of the basic model for the effects of changes in exports as a share of GDP: GDP = 0.0096 ± 0.00124GDPPC + 0.0928GERD + 0.0293EXP + 0.0769INV t = (0.00758) (±2.746) (1.669) (1.717) (1.215). R2 = -0.57 (adj. df 0.40). DW = 2.65 Rho (autocorrelation coef®cient) = ±0.340, t = ±1.222. SEs and variance shown are heteroskedastic consistent estimates The logarithm model: LGDP = ±1.028 ± 0.269LGDPPC + 0.153LGERD + 0.352LEXP + 0.879LINV t = (±0.418) (±2.253) (0.929) (1.810) (1.623). R2 = 0.60 (adj. df 0.45). DW = 2.61 Rho (autocor relation coef®cient) = ±0.366. t = ±1.33. SEs and variance shown are heteroskedastic consistent estimates Test of the basic model for the terms of trade hypothesis: GDP = 1.14 ± 0.00126GDPPC ± 0.027GERD + 0.115INV + 0.283TRD t = (1.055) (±2.922) (±0.493) (1.679) (3.893). R2 = 0.56 (adj. df 0.39). DW = 1.92 Rho (autocorrelation coef®cient) = 0.067. t = 0.224. SEs and variance shown are heteroskedastic consistent estimates PROD = ±0.797 ± 0.0014GDPPC ± 0.0360EXPA + 0.191INV + 0.0585TRD t = (±0.551) (±2.733) (±0.432) (2.824) (0.549). R2 = 0.64 (adj. df 0.50). DW = 1.52 Rho (autocorrelation coef®cient) = 0.277 t = (0.938), SEs and variance shown are heteroskedastic consistent estimates PROD = 0.225 ± 0.00181GDPPC ± 0.0797GERD + 0.173INV + 0.0554TRD t = (0.134) (±3.278) (±1.005) (2.397) (0.414). R2 = 0.67 (adj. df 0.53). DW = 1.43 Rho (autocorrelation coef®cient) = 0.294. t = 1.010. SEs and variance shown are heteroskedastic consistent estimates Notes: See notes for Table 15.5.

272

Test of the basic model for the effects of changes in exports as a share of GDP: GDP = 2.031 ± 0.00197GDPPC + 0.090EXPA + 0.0174EXP + 0.037INV t = (1.169) (±4.008) (2.051) (1.113) (0.514).R2 = 0.59 (adj. df 0.43). DW = 1.75 Rho (autocorrelation coef®cient) = ±0.258. SEs and variance shown are heteroskedastic consistent estimates

Table 15.7 Tested model for the eight advanced member states (category 1), 1973±90 The basic model including patents and changes in exports as a share of GDP: GDP = ±1.010 ± 0.00057GDPCP + 0.069EXPA ± 0.011EXP + 0.156INV t = (±0.999) (±2.516) (1.871) (±0.424) (3.903). R2 = 0.81 (adj. df 0.56). DW = 1.88 Rho (autocorrelation coef®cient) = ±0.987. t = ±51.667. SEs and variance shown are heteroskedastic consistent estimates The logarithm model: LGDP = ±3.113 ± 0.300LGDPCP + 0.196LEXPA + 0.098LEXP + 2.049LINV t = (±0.877) (±1.597) (0.942) (0.083) (2.499). R2 = 0.80 (adj. df 0.55). DW = 1.80 Rho (autocorrelation coef®cient) = ±0.983. t = ±40.16. SEs and variance shown are heteroskedastic consistent estimates The basic model including gross exports on R&D and changes in exports as a share of GDP: GDP = ±0.281 ± 0.00051GDPCP ± 0.0169GERD ± 0.0452EXP + 0.141INV t = (±0.353) (±4.472) (±1.050) (±2.090) 3.223). R2 = 0.77 (adj. df 0.47). DW = 2.77 Rho (autocorrelation coef®cient) = ±0.992. t = ±87.79. SEs and variance shown are heteroskedastic consistent estimates The logarithm model: LGDP = ±2.110 ± 0.258LGDPCP ± 0.049LGERD ± 0.083LEXP + 1.812LINV t = (±0.577) (±1.124) (0.455) (±0.616) (2.144). R2 = 0.76 (adj. df 0.45). DW = 2.69, Rho (autocorrelation coef®cient) = ±0.988. t = ±56.887. SEs and variance shown are heteroskedastic consistent estimates Notes: See notes for Table 15.5.

273

274

Table 15.8 Tested model for the four less advanced member states (category 2), (1973±90) The basic model including patents and changes in exports as a share of GDP: GDP = 2.140 ± 0.00029GDPCP ± 0.075EXPA + 0.0701EXP ± 0.056INV t = (3.584) (±5.161) (±0.231) (7.330) (±3.587). R2 = 0.86 (adj. df 0.58). DW = 1.62 Rho (autocorrelation coef®cient) = ±0.990. t = ±70.63. SEs and variance shown are heteroskedastic consistent estimates The basic model including gross exports on R&D and changes in exports as a share of GDP: GDP = 0.245 ± 0.00042GDPCP + 0.099GERD + 0.076EXP ± 0.0127INV t = (0.234) (±1.064) ( 2.204) (9.689) (±0.395). R2 = 0.90 (adj. df 0.72). DW = 1.93 Rho (autocorrelation coef®cient) = ±0.837, t = ±0.206. SEs and variance shown are heteroskedastic consistent estimates The logarithm model: LGDP = ±0.456 ± 0.040LGDPCP + 0.164LGERD + 0.739LEXP ± 0.321LINV t = (±0.143) (±0.241) (0.571) (2.976) (±0.372). R2 = 0.83 (adj. df 0.49). DW = 1.45 Rho (autocorrelation coef®cient) = 0.112. t = 0.162. SEs and variance shown are heteroskedastic consistent estimates Test of basic model for the terms of trade hypothesis: GDP =1.64 ± 0.00157GDPCP + 0.052EXPA + 0.083INV + 0.200TRD t = (1.12) (±5.080) (1.588) (1.080) (3.110). R2 = 0.57 (adj. df 0.41). DW = 1.68 Rho (autocorrelation coef®cient) = 0.33. t = 1.205. SEs and variance shown are heteroskedastic consistent estimates Notes: See notes for Table 15.5.

George M. Korres 275

technological infrastructure and could produce a large number of patents, while the second group were trying to establish and upgrade their technological infrastructure. The results show that the degree of explanation is very high (above 80 per cent) ± most of the variables are statistically signi®cant, while the standard errors and the variance shown are heteroskedastic consistent estimates.

15.3

Conclusions

Technological progress has become virtually synonymous with long-term economic growth. This raises a basic question about the capacity of both industrial and newly industrialised countries to translate their seemingly greater technological capacity into productivity and economic growth. Usually, there are dif®culties in estimating the relation between technological change and productivity. Technological change may have accelerated, but in some cases there is a failure to capture the effects of recent technological advances in productivity growth or a failure to account for quality changes in previously introduced technologies. The countries of Europe have a long cultural and scienti®c tradition and the major scienti®c discoveries and developments in technology are products of European civilisation. There is a close relationship between innovation and productivity levels. However, there are large technological disparities between the member states, which affects productivity performance, increases economic disparities and hinders economic integration. There are various explanations in the literature for the slow-down in productivity growth in the OECD countries. One source of the slow-down may be substantial changes in the industrial composition of output, employment, capital accumulation and resource utilisation. Another may be that technological opportunities have declined; or else new technologies have been developed but their application to production has been less successful. Technological factors act in a long-term way and should not be expected to explain medium-term variations in the growth of GDP and productivity. The technological gap models represent two con¯icting forces: innovation, which tends to increase productivity differences between countries; and diffusion, which tends to reduce them. In Schumpeterian theory, growth differences are seen as the combined result of these forces. We have applied an economic growth model based on Schumpeterian logic. This technological gap model provides a good explanation of the differences among various countries. The empirical estimates suggest that

276 Technological Change, Productivity and Economic Integration

the convergence hypothesis applies for industrialised countries. Research on why growth rates differ has a long history that goes well beyond growth accounting exercises. The idea that poorer countries eventually catch up with richer ones was advanced as early as in the nineteenth century to explain continental Europe's convergence with Britain. In the 1960s one of the most basic models was the Marx±Lewis model of abundant labour supplies, which explained the divergent growth experience of the Western European countries. To achieve safe results, it is necessary to conduct a cross-country, multisectoral analysis of how technological activities affect the different sectors. According to our estimates there is a relationship between the level of economic growth and the growth of technological activities. Technological activities (best measured by patents) appear to contribute considerably to economic growth, unless this is a negative demand effect. Speci®cally, our results con®rm that there is a close relationship between the level of economic growth (as measured by per capita GDP) and the level of technological development (as measured by the number of external patents). Our results indicate that both imitation and innovation activities have a signi®cant effect on the growth of GDP and productivity. Countries that are technologically backward might be able to generate more rapid growth than even the advanced countries, if they are given the opportunity to exploit the new technologies employed by the technological leaders. The pace of the catching up depends on the diffusion of knowledge, the rate of structural change, the accumulation of capital and the expansion of demand. Those member states whose growth rates are lagging behind could catch up, if they reduced the technological gap. An important aspect of this is that they should not rely only on technology imports and investment, but should also increase their innovation activities and improve their locally produced technologies (as happened in Korea and Singapore). However, our results con®rm that some of the small and medium-sized EU member states have attained high levels of per capita GDP without a large innovation capacity. To explain the differences in growth between these countries in the post-war period, a much more detailed analysis of economic, social and institutional structures should be conducted. When we compare the technologically advanced and less advanced member states, it is not dif®cult to see that the less advanced countries lacked experience of large-scale production, technical education and resources. The catching-up hypothesis is related to economic and technological relations among countries. There are different opportunities for countries

George M. Korres 277

to pursue a development strategy that depends on resource and scale factors. In summary, we can say that the introduction of new technologies has in¯uenced industrialisation and economic growth. Of course, for countries with poor technological apparatus, the impact of new technologies is much smaller. Finally, it seems that the technological gap between the less and more advanced countries is still widening. Notes 1. Growth accounting begins with the measurement of factor accumulation and then imputes output expansion to the inputs that have been accumulated by assuming that market factor prices re¯ect value marginal products. The purpose of growth accounting is to determine from the empirical data how changes in real output (between two periods, say t = 0 and t = T) can be attributed to changes in the inputs (capital, labour and technology respectively). The part of output growth that cannot be attributed to the accumulation of any input is referred as the `Solow residual' and is attributed to technological progress. See, Denison (1962). 2. See, OECD (1991, 1992); UNESCO (1969). 3. Total factor productivity was introduced by Kendrick (1984) and applied to the output of industries, economic sectors or the economy as a whole. It can be measured by the tangible inputs of labour (measured by man hours worked) and capital inputs (measured by capital stock derived as past investment minus depreciation). The labour and capital indexes are then combined into an index of factor inputs by means of weights that represent distributive shares of industry products in the base period. Productivity growth is the difference between the growth of output and the growth of tangible input, as measured by this index. According to this method, the productivity that arises from improvement in the quality of labour and capital is included in residual growth. Denison (1962) concludes that technological change (or advanced knowledge) contributed only 40 per cent to US productivity growth. 4. For a more detailed analysis, see, Fagerberg (1987, 1988). 5. In the basic model we use the variables GDPPC (GDP per capita), INV (investment) and EXPA/GERD (external patent applications/gross expenditure on R&D). However, these variables have some disadvantages because the patent index clearly overestimates the absolute differences in technological level between countries and GERD data do not exist for several periods. 6. The method we use attempts to overcome the problem of spurious correlation, and to estimate the relationship between the rates of change of the variables rather than between their absolute levels. The result of looking at the rate of change in a variable is typically to remove any trend element. That is, many non-stationary economic time series become stationary when they are ®rst differentiated. Unfortunately, when attention is concentrated on the relationship between rates of change there is a real danger that valuable information on the long-term relationship between the levels of the variables will be lost. Firstdifference is an unsatisfactory method of dealing with a spurious correlation problem. See, Berndt (1991). 7. This causes multicollinearity and interpretation problems; for instance, there is feedback from dependent to independent variables. See, Fagerberg (1988).

278 Technological Change, Productivity and Economic Integration

References Abramovitz, M. (1986) `Catching-up, foreign ahead and falling behind', Journal of Economic History, vol. 46. Baily, M. N. and A. K. Chakrabarti (1988) Innovation and the productivity crisis (Washington, DC: Brookings Institution). Berndt, E. R. (1991) The practice of econometrics: classic and contemporary (Reading, MASS: Addison-Wesley). Berndt, E. R. and J. Triplett (1990) `Productivity and economic growth', in D. Jorgenson (ed.) Fifty years of economic measurement studies in income and wealth (Cambridge, MASS: NBER), pp. 19±119. Bhalla, A. and A. Fluitman (1985) `Science and technology indicators and socioeconomic development', World Development, vol. 13, no. 2. Boskin, M. J. and L. J. Lau (1992) `Capital, technology and Economic growth', in Rosenberg, Landau and Mowery (eds), Technology and the wealth of nations (Stanford, CA: Stanford University Press). CEC (1987) Research and technological development for Europe (Brussels: CEC). CEC (1991) Government ®nancing of R&D 1980±1990 (Brussels: Eurostat). Davidson, R and G. J. Mackinnon (1993) Estimation and inference in Econometrics (Oxford: Oxford University Press). Denison, E. F. (1962) The sources of economic growth in the United States (Washington: Committee for Economic Development). Diewert, W. E. and T. J. Wales (1987) `Flexible functional forms and global curvature conditions', Econometrica, vol. 55, no. 1, (January), pp. 43±68. Fabricant, S. (1954) Economic progress and economic change (Princeton, NJ: Princeton University Press). Fagerberg, J. (1987) `A technology gap approach to why growth rates differ', Research Policy, vol. 16, pp. 87±99. Fagerberg, J. (1988) `Why growth rates differ', in G. Dosi, C. Freeman, R. Nelson and L. Soete (eds), Technical Change and Economic Theory (Cambridge University Press). Fagerberg, J. (1991) `Innovation, catching-up and growth', in Technology and productivity ± the Challenge for Economic Policy (Paris: OECD). Fagerberg, J. (1994) `Technology and international differences in growth rates', Journal of Economic Literature, vol. X X X I I , no. 3. Fagerberg, J. (1995) `User-producer interaction, learning and comparative advantage', Cambridge Journal of Economics, vol. 19, pp. 243±56. Gittleman, M. and E. N. Wolf (1995) `R&D activity and cross-country growth comparisons', Cambridge Journal of Economics, vol. 19, pp. 189±207. Griliches, Zvi (1973) `Research expenditures and growth accounting', in B. R. Williams (ed.), Science and technology in economic growth (London). Griliches, Zvi (1979) `Issues in assessing the contribution of R&D to productivity growth', Bell Journal of Economics, vol. 10. Griliches, Zvi (1980) `R&D and the productivity slow down', American Economic Review, vol. 70, no. 2. Griliches, Zvi (1988) Technology, education and productivity (Oxford: Basil Blackwell). Griliches, Zvi and F. Lichtenberg (1984) `R&D and productivity growth at the industry level: Is there still a relationship?', in Zvi Griliches (ed.), R&D, patents and productivity (Cambridge, MASS: Harvard University Press).

George M. Korres 279 Jorgenson, D. W., F. M. Gollop and B. M. Fraumeni (1987) `Productivity and U.S. economic growth', Growth in Aggregate Output (Cambridge, MASS: Harvard University Press). Keller, A. (1990) `Econometrics of technical change: techniques and problems', in P. Hackl (ed.) Statistical analysis and forecasting of economic structural change (International Institute for Applied Systems Analysis). Kendrick, J. W. (1984) International comparisons of productivity and causes of the slowdown (Cambridge, MASS: Ballinger). Kendrick, J. W. (1991) `Total factor productivity ± what it does not measure: an overview', in OECD, Technology and productivity: the challenge for economic growth (Paris: OECD), pp. 149±57. Naxakis Xarris (1997) `Internationalisation of economy and environmental degrading', Market Without Borders, Review of the Institute of Economic Relationships, vol. 3, no. 1. Landau, R. (1989) `Technology and capital formation', in D. Jorgenson and R. Landau (eds), Technology and capital formation (Cambridge, MASS: MIT Press). OECD (1991) Basic Science and Technology Statistics (Paris: OECD). OECD (1991) Economic-Outlook: Historical Statistics, 1960±1990 (Paris: OECD). OECD (1991) Industrial policy in OECD countries: annual review 1990 (Paris: OECD). OECD (1991) Technology and productivity ± the challenge for economic growth (Paris: OECD). OECD (1992) TEP ± technology in a changing world (Paris: OECD). OECD (1992) TEP ± technology and economy: the key relationships (Paris: OECD). Quinn, A. (1990) Trends in scienti®c R&TD in the EEC (Brussels: CEC). Sato, R. and G. Suzawa (1983) Research and Productivity (Boston, MASS: Auburn House). Schmookler, J. (1966) Invention and economic growth (Cambridge, MASS: Harvard University Press). Schumpeter, J. A. (1934) The theory of economic development (Cambridge, MASS: Harvard Economic Studies). Schumpeter, J. A. (1942) Capitalism, socialism and democracy (New York: Harper). Solow, R. M. (1957) `Technical change and the aggregate production function', Review of Economics and Statistics, vol. 39, pp. 312±20. UNESCO (1969) The measurement of scienti®c and technological activities (Paris: UNESCO). UNESCO (1992) Scienti®c and technology in developing countries: strategies for the 1990s (Paris: UNESCO). Varian, R. Hall, Theodore C. Bergstrom, and R. Hal (1993) Workouts in intermediate microeconomics, 3rd edn. (New York: W. W. Norton). Vernon, R. (1966) `International investment and international trade in the product cycle', Quarterly Journal of Economics, vol. 80, no. 2, pp. 190±207. Zander, U. (1991) Exploiting a technological edge ± voluntary and involuntary dissemination of technology (Stockholm: Institute of International Business, Stockholm School of Economics).

16

Privatisation: Recent Experiences and Prospects in a World Without Borders George C. Bitros

16.1

Introduction

The world-wide slowdown in economic growth, which started in the early 1970s, immediately launched experts into a concerted effort to ®nd explanations and devise appropriate remedial policies. Prominent among the explanations of the slowdown was the unprecedented expansion of state sectors. This eventually, in conjunction with numerous other ominous developments such as the ever-rising budget de®cits, the relentless bureaucratisation of individual initiatives and the pressure to adjust to the increasingly competitive international environment, caused public opinion all over the free world to swing in favour of rolling back the state. Initially the policies in this direction assumed various forms that did not involve the transfer of state-owned assets to the private sector, for example the deregulation of telecommunications and airlines in the United States, the contracting-out by local authorities in many countries of refuse collection and other essential services, and the franchising to private ®rms of productive activities for various common-pool goods.1 Then, in the early 1980s, and particularly after the successful sale of several major state companies by the Thatcher government in the UK, the thrust of the effort to reduce the size of the public sector took the form of privatisation. The fact that privatisation came to be considered in the 1980s as an equally orthodox approach to restructuring economies as the more conventional industrial policies is amply brought out by the political manifestos presented by conservative and socialist parties all over the world. Two rather obvious examples in this respect are those of the Conservative Party in Greece and the Socialist Party in Spain. By most

280

281

accounts, the conservatives in Greece won the 1996 elections not despite but because of the extensive liberalisation and privatisation policies they advocated. Likewise, the socialist government in Spain launched an extensive privatisation programme but this did not cause it to fall from power. So, after several years of world-wide experiments with privatisation, it is time to take stock and attempt an appraisal of the various experiences and the prospects that lie ahead. This chapter ®rst summarises the evidence on the international diffusion of the policy, then highlights the forces that determined its acceptance, and ®nally, in the light of these comments, makes certain educated guesses about likely developments in the foreseeable future.

16.2

The picture at the end of 1992

As already mentioned, the ®rst serious attempt to privatise took place in the UK in the early 1980s. The ®rst signi®cant privatisations were conducted there with the sale of Cable and Wireless in 1981 and the National Freight Company in 1982. Thereafter, privatisation proceeded on a wider, as well as on a more international, scale. A good, albeit rough, idea of what has transpired since then can be gleaned from Table 16.1. Relevant to the diffusion rate and the regional distribution of privatisation efforts are the following points. First, whereas privatisation initially spread at an increasing rate, it later slowed down to a signi®cant extent ± the number of countries that adopted some or other form of privatisation grew from ®ve in 1984 to 14 in 1989,but only 16 in 1992. Second, although privatisation spread at an increasing rate in the 1980s, only a handful of countries implemented privatisation programmes. Without going into the factors that may have hindered the international diffusion of the policy, it is interesting to note that an extension to the list of the less developed countries would include few of the poor countries of Africa, Latin America and Asia.2 Third, few of the countries that did embark on a privatisation programme did so to a signi®cant extent. According to the relative measure shown in the ®nal column of Table 16.1, the proceeds from privatisation as a percentage of average annual GNP amounted to more than 10 per cent in only four countries (Mexico, Argentina, the UK and New Zealand). Finally, the great majority of the countries concerned are located in Western Europe. In particular, of the 14 countries listed in the table, the majority are members of the European Union.

282

Table 16.1 The diffusion of privatisation policy1

Austria Canada France Germany2 Greece3 Italy Japan Netherlands New Zealand Portugal Spain Sweden Turkey UK

Privatisation period

Cumulative proceeds (£billion)

Proceeds as % of average annual GDP

1987±90 1984±90 1983±91 1984±90 1991±92 1983±91 1986±88 1987±91 1987±91 1989±91 1987±90 1987±90 1988±91 1979±91

0.60 1.60 8.24 3.34 0.45 6.25 47.80 1.50 3.00 1.50 1.20 1.30 0.30 44.50

0.9 0.6 1.5 0.5 ± 1.4 3.1 1.0 14.1 4.3 0.5 1.2 1.6 11.9

Notes: 1. The calculations are rough because they do include the value of assets that were transferred to the private sector at practically no cost to the purchaser. 2. Area of Federal Republic of Germany before uni®cation. 3. Author's calculations. Source: Stevens (1992).

283

In view, then, of the limited scope of privatisation in the 1980s, it is natural to infer that states all over the world continue to hold considerable assets in many sectors of their economies. A rough indication of the situation in the OECD countries in 1990/91 is provided by the breakdown of state-owned assets by country and broad sectors of economic activity in Table 16.2. Looking at these data, a number of interesting observations come immediately to mind. First, in the most developed country of the world, the United States, the state owns only one asset, the postal service. Second, in Japan, which is the second most developed country in the world, the government holds a majority interest in the railways and total ownership of the postal service. However, given that the railways are to be totally privatised shortly, state ownership there will also be restricted to the postal service. Third, as we move to the next tier of developed countries (Canada, Germany before uni®cation, Italy and so on) state ownership spreads into more and more sectors of the economy. Fourth, with a number of exceptions in each case, state ownership in the various OECD countries is concentrated in railway transport, gas, post and communications. Finally, the assets that remain under state control are substantial. In the Mediterranean countries in particular, there is a signi®cant state presence in key manufacturing sectors such as shipbuilding, metals, textiles, chemicals, oil re®ning and banking. The conclusion that, in most OECD countries, the state continues to control sizeable productive assets is strongly reinforced, if we look at the less developed countries. A good case in point is the bloc of Central and Eastern European countries, which are in the process of introducing the open market as the predominant form of economic organisation. According to Andreff (1992), in Hungary, Poland and Czechoslovakia, where privatisation is under way, no more than 60 companies have been transferred to the private sector so far. Hence, the overwhelming majority of productive facilities in these countries, and in Russia, continue to be under state control. Moreover, reports from international organisations and individual researchers show that the situation in most underdeveloped areas is little different from that in the former socialist countries. Working with a sample of 37 less developed countries, Shirley (1988) found that, at the time, Bangladesh and Chile had privatised more than 20 enterprises each, whereas privatisation in all other countries was relatively insigni®cant. For a more representative sample, World Bank reports indicate that for 90 less developed countries where privatisation was taking place, the overall number of privatised enterprises did not exceed 530.

Steel Australia Austria Belgium Canada Finland France Germany Greece Italy Japan Netherlands New Zealand Norway Portugal Spain Sweden Switzerland Turkey UK United States

0 100 50 0 90 100 0 n.a. 60 0 14 0 50 100 60 40 0 75 0 0

Ship Motor Building vehicles 0 n.a. 0 0 n.a. n.a. 25 70 n.a. 0 n.a. n.a. n.a. n.a. 90 75 n.a. n.a. 0 0

0 n.a. 0 0 100 40 0 n.a. n.a. 0 70 n.a. n.a. n.a. 10 0 n.a. n.a. 0 0

Oil petroleum

Coal mining

Railways

Airlines

0 85 n.a. n.a. 100 n.a. 0 n.a. n.a. 0 n.a. 0 n.a. 100 50 n.a. n.a. 87 25 0

0 100 n.a. 0 n.a. 100 n.a. 70 n.a. 0 n.a. n.a. n.a. n.a. 50 n.a. n.a. 100 100 0

100 100 100 75 100 100 100 100 100 66 100 100 100 n.a. 100 100 100 100 100 (f)

75 52 52 0 70 100 60 100 75 0 39 0 n.a. 100 100 n.a. 30 85 0 0

Electricity production 100 50 25 100 n.a. 100 n.a. 100 75 0 n.a. 94 n.a. n.a. 30 n.a. n.a. n.a. 40 n.a.

284

Table 16.2 Estimated percentage of state-owned assets in selected sectors for various OECD countries, 1990±91 Gas

Post

Comm.

Banking

0 100 25 0 100 100 0 100 80 0 n.a. 100 100 n.a. 100 0 n.a. 100 0 n.a.

100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100

100 100 100 25 100 100 100 100 100 46 100 0 100 n.a. 100 100 100 100 20 0

(a) 10 (b) n.a. n.a. n.a. n.a. 70 40 0 n.a. n.a. n.a. 75 12 (c) (d) (e) 0 0

Notes: (a) The state has signi®cant holdings in the banking sector; for example, the Commonwealth Bank of Australia, which is one of the four largest banks in the country. (b) Semi-public savings banks and credit institutions. (c) Nordbanken, the second largest of the country's banks, is state-owned. (d) District banks owned by regional authorities. (e) State interests in four major banks. (f) The Amtrak railway is federally owned. Source: Stevens (1992).

285

In summary then, the available evidence suggests that, due to the sluggish adoption rate of privatisation in recent years, at the beginning of the 1990s states all over the world had extensive holdings in key segments of manufacturing, transport (air and rail), public utilities and ®nancial services. This was particularly the case with respect to the former socialist countries of Central and Eastern Europe and all the less developed nations. Hence, the prospects should be limited only by the net balance of forces that drive privatisation. It is to this question that we now turn.

16.3

The main determinants of privatisations3

The transfer of public assets to the private sector on a deliberate policy basis is subject to several con¯icting forces. Some of them hamper the process of privatisation, whereas others work favourably. Consequently, its prospects depend on the view that one takes with regard to their relative course and strength. 16.3.1

Negative in¯uences

Privatisation can be slowed by several impediments. At top of the list are the interests of politicians, the interests of state-related private enterprises and the power of the unions in the public sector. Looking ®rst at politicians and political parties, despite whatever ideological or scienti®c reasons to the contrary they may proclaim, their inherent attitude towards privatisation is negative.4 This inference stems naturally from several profound considerations. First, each privatisation reduces the ability of politicians to offer employment opportunities to their immediate political associates as well as voters, and thus affects their ability to be reelected. Second, each privatisation reduces the power of politicians to issue state contracts to particular industrial or ®nancial groups, and thus eliminates potential sources of support during election periods. Finally, politicians are negatively disposed towards privatisation because of the political capital involved in the job losses that occur whenever an overstaffed public enterprise is transferred to the private sector.5 Turning next to state-related private enterprises, it is no secret that although they are publicly in favour of deregulation and privatisation, in actuality they perceive that their interests lie in the maintenance of the big state. The fact, for example, that business concerns contribute large sums of money to political parties suggests that they ®nd this investment worthwhile, and the fact that businesspeople, in direct and indirect ways, have unloaded their bankrupt enterprises onto the state sector reveals

286

quite clearly that entrepreneurs are in favour of competition, but only for everybody else. Hence, a large segment of private enterprises that are presently tied to the productive activities of the state are slowing down privatisation. Finally, a large amount of opposition to privatisation comes from the powerful unions of civil servants and workers in public enterprises and organisations. The tremendous expansion of the state in the post-war period brought with it a tremendous expansion of unions in the public sector, and now governments have to deal with their power. As a matter of fact, drawing on past experiences in the UK, France, Italy and elsewhere, the opposition of unions is so stiff that it will require strong counterbalancing pressures on governments for privatisation to accelerate. In addition to the above, privatisation is constrained by several purely technical factors; for example, inadequate legal and tax provisions, underdeveloped money and capital markets, and sheer ignorance of the technology of privatisation. Even recession can seriously hamper privatisation in two ways. First, in the face of stringent stabilisation policies, major local and international corporations may face signi®cant liquidity constraints that prevent them from buying state properties. Second, given the instability of public ®nances, the degree of uncertainty about future state actions may be too large to allow the necessary risk taking. Last, but not least, privatisation is hampered by the antimarket mentality cultivated in the post-war period by the politics and economics of the welfare state. As long as the majority of citizens ®nd it unnecessary to demand the redrawing of limits between the state and the market, the coalition of interests that bene®t from big government will oppose this policy and try to vitiate it. However, with the recent failure of the welfare state and the collapse of communism, attitudes are shifting in favour of the market, so the constraints on privatisation from this source are easing. 16.3.2 Positive in¯uences Turning to the forces that favour privatisation, there are at least three important ones: the increasing pressure of budget de®cits, which continuously raise public debt; the need to improve the ef®ciency of the economy; and the continuing trend towards globalisation and the intensi®cation of international competition. Hence, there is a need to gauge their course and relative strength in the coming years. If privatisation for ideological or scienti®c reasons is considered unlikely, then it must be motivated by the pressure to reduce budget de®cits ± either by shedding loss-making state enterprises or by selling

287 Table 16.3 Central government budget surplus/de®cit as percentage of GNP

Australia Austria Canada Belgium Denmark Finland France Greece Ireland Italy Japan Netherlands Norway New Zealand Portugal Spain Sweden Switzerland Turkey United Kingdom United States

1985

1990

1995

±2.9 ±4.0 ±3.8 ±11.2 ±0.6 0.1 ±3.4 ±12.7 ±13.8 ±12.5 ±3.9 ±6.5 12.1 ±4.7 ±9.8 ±7.0 ±3.3 ±0.2 ±7.4 ±2.1 ±4.9

1.5 ±3.8 ±2.3 ±6.9 1.2 1.6 ±1.0 ±18.5 ±2.6 ±10.3 ±1.3 ±4.2 4.3 2.1 ±3.1 ±2.0 4.0 ±0.1 ±5.9 0.4 ±3.0

0.2 ±4.2 ±6.6 ±4.5 2.7 ±2.0 ±1.3 ±5.8 ±1.8 ±8.3 ±1.1 ±4.2 5.1 0.5 ±2.1 ±1.3 ±1.3 ±0.6 ±3.7 ±0.7 ±2.0

Source: DRI/McGraw Hill, World Markets Executive, Summary (second quarter, 1991).

pro®table public assets to repay debts, thus reducing interest payments and bringing public ®nances into better balance with the state of the economy. Therefore, the extent to which state budgets may be expected to in¯uence privatisation depends on the expected evolution of de®cits/ surpluses. Table 16.3 shows the status of 21 central government budgets in 1985 and 1990, and the 1995 projections for the sample of OECD countries. Looking at these ®gures, one can hardly fail to draw the following conclusions. First, for the great majority of countries budget de®cits were expected to continue well into the 1990s. Second, ceteris paribus, had it not been for the proceeds from privatisation, the improvement shown in the case of, say, the UK in 1990 would not have taken place, and in several other countries the situation would have been far worse than it is today. Third, under current policies, de®cits are expected to shrink, and in a few countries even to turn into surpluses. In summary, therefore, it may be inferred that the expected course of budget de®cits will continue to favour privatisation, but not as much as in the past. The second factor that may positively in¯uence the process of

288

privatisation is the need to improve the ef®ciency of the economy. In this regard, research opinion is divided between property right theorists, who argue that private ownership is more ef®cient than public ownership,6 and industrial organisation researchers,7 who maintain that ef®ciency is less related to the type of ownership (public or private) than to the competitive conditions that privatisation may or may not create. The implication of the latter argument is that if public enterprises can be made to operate under tough competitive conditions, changing their ownership may not be necessary to secure increases in ef®ciency. However, even if politicians did not view public enterprises as extensions to their authority, there is an insurmountable problem of asymmetrical information here that cannot be confronted in the best way by administrative means as compared to the best results of the market. My view, then, is that the perceived positive relationship between private ownership and ef®ciency will continue to prevail over alternative hypotheses, and will reinforce the drive for privatisation in direct proportion to the economic growth desired by the country in question. Last, and most important, is globalisation and the intensi®cation of international competition. In the 1980s economic, political and technological interdependence increased as never before. International trade expanded at a much faster rate than output, foreign direct investment surged, with the advent of technological developments in communications and computers the size of markets expanded on a worldwide scale, and competition greatly intensi®ed. Under these circumstances, private companies ceased to look solely inwards and started to design and implement strategies to achieve regional and international competitive advantages. The question that therefore arises is: does ownership matter in the struggle for international markets? Or, to put it more bluntly, do stateowned companies have what it takes to thrive in a ®ercely competitive global environment? If the nature of the policies adopted by the EU with regard to public utilities are anything to go by, the predominant view is that state companies are in many ways inferior when pursuing global strategies than private corporations. To corroborate this view, note the following trends. First, at present the practice of subsidising nationalised industries is quite widespread in the EU. However, with the process of economic and monetary union under way the European Commission is subjecting all direct and indirect assistance to ever increasing scrutiny, and as government subsidies are eliminated the pressure will mount to return state-owned enterprises to the private sector.

289

Second, the doctrine of free market access will be invoked during international trade negotiations much more forcefully than in previous rounds, and adherence to it by nation states will be monitored much more closely than in the past. As a result, the insulation of state-owned companies from international competition will reduce, and as this happens governments will be forced to consider privatising these companies. Third, aside from pro®ts and growth, state-owned companies are obliged by governments to pursue social objectives such as supplying services to outlying villages, offering employment to certain categories of people for humanitarian reasons and so on. But with the pressure to internationalise, state companies will be less and less able to ful®l these mostly non-economic objectives. By implication, their usefulness to politicians for disbursing political favours will diminish and this will ease their path to privatisation. Finally, technological advances and new insights in economic theory continue to undermine the foundations upon which state companies are based. For instance, in communications it is now considered that the only part of the industry that should remain a natural monopoly is telephone lines. In light, therefore, of expanding technological interdependence, the demand for state enterprise products will continue to decline, and in turn this will put pressure on governments to proceed with privatisation. Overall then, the trend towards globalisation and the intensi®cation of competition should act as a prime factor in privatisation in the coming years.

16.4

Future prospects

The existing social, political and intellectual inertia continues to favour the maintenance of sizeable public sectors. However, increasing numbers of people are advocating structural policies that will revitalise the international economy and put it on a sustainable growth path. The experience with privatisation in the 1980s has shown that it is a promising structural policy that is likely to continue. The question is, where it will take place and at what pace? It certainly will not take place in countries that have very little to privatise, such as the United States and Japan, or in countries that experienced a large wave of privatisations in the 1980s, such as the UK and France. Nor would it be reasonable to expect fast and massive privatisation in, say, the countries of Central and Eastern Europe8 or the poor countries of Africa. In such environments the overwhelming balance of the forces

290

mentioned above is against privatisation and it will take many years for the ethics and attitudes of market economics to take root, if ever. Instead, the next signi®cant wave of privatisations can be expected to occur in countries that: (1) are eager to integrate into the more competitive international environment; (2) have oversized and inef®cient public sectors; and (3) continue to experience large budget de®cits. This pro®le primarily ®ts Turkey and the countries of Southern Europe. As Table 16.2 shows, government ownership in Portugal, Spain, Italy, Greece and Turkey extends deep into industry, transport, communications, energy, other utilities and banking. So, privatisation in these sectors is likely to be substantial. However, the timing of privatisation is hard to gauge because there are various technical and political dif®culties. An obvious case in point is Greece. The Greek stock market is underdeveloped, hence sales must be directed primarily at foreign buyers, with all the dif®culties that entails. Yet, given the pressure of budget de®cits and the need to integrate further into the European Union, privatisation will be hard to resist. And the same is more or less true of all the other countries of Southern Europe. The above in no way implies that privatisation will cease in the core countries of Europe. France, Belgium, the Netherlands and Germany will proceed further, particularly with regard to transport, communications and utilities. But the assets of these industries are dif®cult to privatise, so the process is bound to be slow. Among the other countries of continental Europe the one that holds particular promise is Austria, which has substantial state holdings and has been under pressure from budget de®cits. As the holding company OIAG breaks up and segments are transferred to the private sector, the process should accelerate. A similar inference would seem to be warranted with regard to Sweden and Finland, as both countries have privatisation plans. However, in light of broad public support for the continued presence of the state in key sectors of the economy, it is doubtful whether the activity will pick up speed. With respect to the rest of the world, the countries that may show increased privatisation activity are Argentina, Chile, Mexico and Egypt.

16.5

Summary and conclusions

At the beginning of the 1990s, states all over the world continued to control large chunks of national assets. These assets were spread over several key sectors but were mostly concentrated in transport (rail and air), communications, energy and utilities. While there are substantial grounds for the acceleration of privatisation, the balance of negative

291

and positive forces that determine the direction and rate of privatisation is not everywhere favourable. Hence, privatisation is expected to accelerate in a select group of countries ± located mainly in the Mediterranean region, beset by large budget de®cits and eager to integrate into the international economy ± whereas it may slow down elsewhere. At the global level therefore, the expectation is that the transfer of statecontrolled assets to the private sector is likely to decelerate. Notes 1. An excellent taxonomy of the various systems for the production and delivery of goods and services is provided by Savas (1985). 2. Surveys by the Adam Smith Institute (1986) and more recent publications by the Agency for International Development show an acceleration of privatisation efforts in these countries, but with limited results so far. 3. For a more detailed list of determinants of the decision to privatise, see, Cameron (1992). 4. Looking back, probably the only privatisation programme that was undertaken not because of budgetary pressures but for ideological and political reasons was that of the UK under Margaret Thatcher. 5. For an insightful analysis of the fundamental reasons why politicians should be inherently against privatisation, see, Buchanan and Wagner (1977). 6. See, for example, Alchian (1965), Baumol (1982) and Demsetz (1983). 7. See, for example, Hartley et al (1991). 8. An excellent evaluation of the dif®culties that liberalisation and privatisation face in these countries, and how slow and protracted the process, are presented in Hines (1991).

References Adam Smith Institute (1986) Privatisation World-wide (London: Adam Smith Institute). Alchian, A. A. (1965) `Some economics of property rights', Il Politico, vol. 30, pp. 316±29. Andreff, V. (1992) `French privatisation techniques and experience: a model for Central-Eastern Europe?', in F. Targetti (ed.) Privatisation in Europe: West and East Experience (Vermont: Darmouth). Baumol, W. J. (1982) `Contestable markets: an uprising in the theory of industry structure', American Economic Review, vol. 52, pp. 1052±69. Buchanan, J. M. and R. E. Wagner (1977) Democracy in De®cit (New York: Academic Press). Cameron, C. W. (1992) `A comparison of privatisation in capitalist and socialist countries', International Journal of Social Economics, vol. 19, pp. 55±60. Demsetz, H. (1983) `The structure of ownership and the theory of the ®rm', Journal of Law and Economics, vol. 26, pp. 375±90. Hartley, K., D. Parker and S. Martin (1991) `Organisational Status, Ownership and Productivity', Fiscal Studies, vol. 12. Hines, M. (1991) `The Transition in Central and Eastern Europe', Discussion Paper (Washington, DC: World Bank).

292 Savas, E. S. (1985) `Privatisation: A Powerful New Tool for Government', The Privatisation Review, vol. I, pp. 4~6. Shirley, M. (1988) `L' expeÂrience de la privatisation', Finance et DeÂveloppement, vol. 25. Stevens, B. (1992) `Prospects for privatisation in OECD countries', Quarterly Review, August, (London National Westminster Bank), pp. 2±22.

17

The Balkans and the European Union Nicholas V. Gianaris

17.1

Introduction

Throughout history the Balkan countries ± Albania, Bulgaria, Greece, Romania, Turkey and former Yugoslavia ± have been in¯uenced by ethnic, political and economic elements of larger nations, mainly because of their location at the cross-roads of three continents (Europe, Asia and Africa). After the Second World War the collectivisation of farms, the nationalisation of industries and central planning were imposed in the northern Balkan countries, primarily Albania, Bulgaria, Romania and the former Soviet Union. With the collapse of the communist system in the late 1980s, gradual steps were taken towards a free market system, similar to that of the European Union (EU). The economies of the former socialist nations in the Balkans face the problems of low production incentives and inef®cient resource allocation, mainly because of the remaining state controls. Production incentives are greater under private ownership than under common or state ownership ± as Aristotle said some 25 centuries ago, common ownership means common neglect.1 From that standpoint, more political and economic reforms will be needed and the denationalisation of more state enterprises will be required in these countries if they are to join the EU. The Balkan nations, which for more than four centuries were under Turkish occupation and subject to internal con¯icts, introduced free market and democratic reforms similar to those of the other East European nations. Former Yugoslavia ± which for four decades practised a selfmanagement system whereby the workers appointed the managers ± split into independent republics after severe ethnic and religious con¯icts, and is in the process of economic and political adjustment. There is a current

293

294 The Balkans and the EU

trend towards greater cooperation and integration between the Balkan nations, which may eventually lead to support for the `convergence theory'. In order to revitalise their economies, the Balkan countries need Western investment and new technology. By helping them, the West would eventually be helping itself. All these countries offer attractive terms to EU ®rms, other foreign investors and multinational corporations in return for new technology and know-how. Despite criticism that such corporate `beasts' interfere with the economic and political life of the host countries, they are considered to be necessary agents of managerial expertise, capital investment and employment. From the standpoint of income distribution, per capita income conceals greater inequalities in Greece and Turkey than in the other Balkan countries. However, there is a trend towards wage and price determination (income policies) to bring about better income distribution. Moreover, the gradual acquisition of corporate shares and the spread of insurance, retirement and similar mutual funds to large segments of the population help alleviate the problem of inequality and tend to support the spread of what may be called democratic capitalism, which can be seen as another path towards convergence with the EU economies.2

17.2

A comparative review

For a number of decades, almost all Balkan countries have enjoyed high rates of investment, measured as a percentage of GDP. On average, former Yugoslavia and Greece invested more capital per unit of output than the other Balkan countries. The ratio of investment to additional output was close to 5.0 for former Yugoslavia, 4.0 for Greece, 3.7 for Romania and around 3.0 for Albania and Bulgaria. All the Balkan countries provide free education at all levels, with minor exceptions in Greece and Turkey. However, Greece has the lowest annual educational expenditure as a percentage of GDP and the highest number of students (about 40 000) relative to its population. Nevertheless, because of the large budget de®cits compared with the EU limits (no more than 3 per cent of GDP), there is a trend towards private, non-pro®t universities and other schools where students pay for their tuition, as practised in, for example, the UK, France and the United States.3 Table 17.1 shows the main indicators for the Balkan countries. Greece has the highest per capita GNP, followed by Slovenia, Bulgaria, Romania and Albania (the poorest nation in Europe). The agricultural sector makes the highest contribution to GDP in Albania, followed by Croatia and

Table 17.1 Economic indicators for the Balkan countries, 1995 Population (millions) Albania Bulgaria Bosnia-Herzegovina Croatia FYROM Greece Romania Slovenia Turkey Yugoslavia (new)

3.3 8.4 4.1 4.8 2.1 10.5 22.7 2.0 61.1 10.9

Per-capita GNP (US $) 670 1330 4639 3250 860 8210 1480 8200 2780 3060

Agriculture as a share of GDP 56 13 n.a. 12 n.a. 21 21 5 16 12

Current account balance ($ million) ±12 334 n.a. ±1712 n.a. ±2864 ±1342 ±37 ±2339 n.a.

Note: The population of Yugoslavia (new), which now consists of Serbia, Montenegro and the two provinces, Kossovo and Vojvodina, was reduced when some republics broke away in 1991 (Bosnia-Herzegovina, Croatia, Former Yugoslav Republic of Macedonia, or FYROM, and Slovenia). Sources: World Bank, World Development Report, 1997 (Washington, DC: Oxford University Press for the World Bank); Nicholas V. Gianaris, Geopolitical and Economic Changes in the Balkan countries (Westport, CT: Praeger, 1996), ch. 5.

295

296 The Balkans and the EU

Bulgaria. Greece, Turkey, Croatia and Romania have a high de®cits in their current account balances, whereas Albania and Slovenia have small de®cits and Bulgaria has a surplus. Comparatively speaking, the EU countries had a high per capita GNP in 1995, varying from $8210 in Greece and $9700 in Portugal to $27 510 in Germany and $29 890 for Denmark, as shown in Table 17.2. In the same year the United States had $26 980, Japan $39 640 and Switzerland as much as $40 630. The share of agriculture in GDP for most of the EU countries was very small, ranging from 2±6 per cent. Germany and Austria had high de®cits in their current account balances, whereas Italy, France and the Netherlands had high surpluses in 1995, compared with a high de®cit ($148.2 billion) for the United States and a high surplus ($111.2 billion) for Japan. Regarding participation in the common EU currency (the euro), under the terms of the Maastricht Treaty the member states, including Greece, must reduce their de®cits to less than 3 per cent and their public sector debt to less than 60 per cent of GDP. The other Balkan countries should adjust their policies accordingly for eventual EU membership. The Balkan countries are relatively small and their economies are weak. This suggests that they should join the large group of EU countries for their own protection and rapid development. As Thucydides said some 25 centuries ago, `In the judgement of human beings, justice counts when there is equal power for implementation, but if it does not occur, the powerful do whatever their power allows and the weak retreat and accept' (translated from the Hellenic). Closer relations with, and eventual membership of, the EU would stabilise the economies of the Balkan nations and increase trade and investment in the area. This, in turn, could help eliminate ethnic, religious and territorial con¯icts, particularly in the republics of former Yugoslavia.

17.3

Privatisation and investment

Since the drastic changes in Eastern Europe and the former Soviet Union in 1989, the winds of freedom and democracy have blown through the Balkan region. Despite the reappearance of ethnic rivalries, which were suppressed under communism, gradual and painful efforts are being made towards a market economy through the privatisation of public enterprises. Over the decades, Greece has made many attempts to bring the Balkan countries into closer cooperation, and more so recently. Now, as a full member of the EU, Greece can play a decisive role in economic and geopolitical developments in the region.

Table 17.2 Economic indicators for the EU countries, 1995

Austria Belgium/Luxembourg Britain Denmark Finland France Germany Greece Ireland Italy Netherlands Portugal Spain Sweden

Population (millions)

Per capita GNP (US $)

8.1 10.1 58.5 5.2 5.1 58.1 81.9 10.5 3.6 57.2 15.5 9.9 39.2 8.8

18 720 24 710 18 700 29 890 20 580 24 990 27 510 8 210 14 710 19 020 24 000 9 740 13 580 23 750

Current account balance ($ millions) ±5 113 14 960 ±46 32 1 413 5 642 16 443 ±20 976 ±2 864 1 379 25 706 16 191 ±229 1 280 4 633

Source: World Bank, World Development Report 1997 (Washington, DC: Oxford University Press).

297

298 The Balkans and the EU

There are many opportunities for trade and investment in the area and scores of Greek, other EU and US companies are moving into the pro®table Balkan markets, primarily in joint ventures with local enterprises. The growth of large corporate enterprises requires expansion into many countries, and the Balkans cannot be excluded from consideration. The Greek government provides tax and other incentives for Greek enterprises to invest in the northern Balkan countries, which strengthens the economies of the countries involved and creates economic bridges between the EU and the Balkan nations. A number of Greek banks, as well as the European Investment Bank, are participating in this investment expansion by providing much needed loans and venture capital. In addition, the Cross-Balkan and Euxene Centre in Salonika provides information and other services to companies that wish to invest in the Balkan and Euxene countries. The Union of Euxene countries (a similar formation was called `The Common Market of Euxene' by Aristotle in the ®fth century BC) comprises neighbouring Balkan Sea countries ± Armenia, Bulgaria, Georgia, Greece, Romania and Turkey.4 Albania, a tiny and poor country that mainly produces agricultural goods and mineral products such as chrome (Albania is the world's third largest producer) and oil (about 20 million barrels a year), has started to privatise its state enterprises and is forming joint ventures with companies from the EU and other countries. Such joint ventures include that between the Albkrom state enterprise (producing chrome) and six British companies (including GEC-Alsthom and Compair Holman) and the American Mecalloy Corporation. In addition, Chevron, the large US petroleum company, has reached an agreement with DPNG (an Albanian petroleum enterprise) to extract and produce coal and petroleum products in an area close to Tirana, the capital of Albania, and Coca-Cola has opened a $10 million bottling plant. Recently, the former Yugoslav Republic of Macedonia introduced legislation for the privatisation of state enterprises. Firms with fewer than 50 employees are to be sold to the highest bidder, with preference given to the buyer who can obtain 29 per cent of the capital value, provided that this ®rst buyer eventually obtains at least 51 per cent of the total. The Bulgarian Organisation of Privatisation plans to privatise more than 300 state enterprises, in addition to the more than 100 that have already been sold. However, Bulgaria, where the bulk of the economy is still in the hands of inef®cient state ®rms, is proving slow to privatise. Among the state ®rms that have been sold, or are undergoing privatisation, are Elcabel (wire and similar products), Eltos (machines and tools), Panoyot Volvo (plastics) and Plastchin (plastic products). The Euro-

Nicholas V. Gianaris 299

merchant Balkan Funds has minority stakes in some Bulgarian and other Balkan companies, mainly focusing on food processing and retailing. Also, Global Finance of Greece is engaged in investment ventures in Bulgaria, with some capital contributed by the International Finance Corporation and the European Bank for Reconstruction and Development. Moreover, Bulgaria and Greece have signed an agreement for Russian oil to be carried by tanker from the Black Sea to the Bulgarian port of Burgas, and from there by pipeline to the Greek port of Alexandroupolis. Through the EU Phare and TACIS programmes, EU ®rms (mainly from Greece and Germany) have increased their investment activities in Bulgaria. Although the pace of privatisation in Romania is slow, more than half a million small state enterprises have been sold. Foreign investment is not encouraged, but a number of foreign ®rms have entered Romania and formed joint ventures. The previously unproductive centrally planned economy of Romania, which guaranteed permanent employment but had a low standard of living, is being exposed to ef®cient but often brutal capitalism. As the joke about the many surplus workers in state ®rms used to go, `They pretend to pay us, and we pretend to work'. Foreign investors in Romania include the Greek ®rms Triton and Ellconsult. Investments include the expansion of Black Sea ports; the French bank BNP, in alliance with the Dresdner Bank of Germany, which has opened a subsidiary in Bucharest; and Peugeot Citroen of France, has formed a joint venture with Automobile Dacia (a state-owned carmaker) to assemble small family cars. Moreover, New Holland, a subsidiary of Fiat of Italy, has acquired 60 per cent of Semanatoarea, a Romanian stateowned group, to make tractors for the Eastern European market. The new Yugoslavia, which comprises Serbia, Montenegro and the provinces of Kosovo and Vojvodina, has kept more or less to the former system of self-management, where workers and employees manage the ®rms in which they work. Mainly because of the persistent con¯icts in the region, no signi®cant changes have been introduced. The relatively rich republic of Slovenia, which produced more than 30 per cent of former Yugoslavia's exports, reformed its economy with great speed and was among the ®rst, along with Poland and the Czech Republic, to apply to join the EU. Westernised Croatia, which accounted for 90 per cent of former Yugoslavia's tourist income, declared its independence from Yugoslavia in June 1991, as did Slovenia, but because of its con¯ict with Serbia and its troubled federation with Bosnia-Herzegovina, no signi®cant economic changes have occurred in this republic lately.

300 The Balkans and the EU

Bosnia-Herzegovina, with a mixed population of Muslims (39 per cent), Serbs (32 per cent) and Croats (18 per cent), is still suffering from the aftermath of the ethnic and religious con¯ict and, as a result of the Dayton Agreement, NATO troops remain in the region to try to keep peace as the `Balkanisation' of the area continues.

17.4

Closer cooperation

From an economic point of view, a Balkan common market and the infant Union of Euxene countries cannot be expected to produce impressive results, principally because these countries mainly produce goods that are subject to strong competition. Nevertheless, from a socio-political standpoint, such unions may prove bene®cial as pioneering movements in an eventual pan-European union, under the auspices of the EU. Greece is already a full member and Turkey an associate member of the EU, and other neighbouring nations are expected to become associate or full members in the near future. Again, closer cooperation with the EU may be bene®cial, although painful, for a transitional period of economic and political adjustment. During meetings in Athens and other venues, the Balkan and Euxene countries have pledged further cooperation among themselves and with the EU. EU enlargement will not necessarily lead to a weak and chaotic Europe but to a prosperous Europe and a more stable world. As former Chancellor Helmut Kohl of Germany once said, `Poles and Hungarians, Czechs and Slovaks, and many other people and nations in Central, Eastern and South-eastern Europe place their hopes in the Community. We cannot disappoint them.'5 In order to join the EU, all the former socialist Balkan countries, as well as Turkey, will have to speed up their economic and political reforms. In particular, Turkey will have to respect the human rights of its more than 12 million Kurds, stop its expansionary policy in Cyprus and the Aegean Sea and pay greater attention to improving the economic circumstances of its people. All the Balkan countries should sell inef®cient state ®rms to the private sector, with some discount to their employees, thereby increasing productivity. Privatisation is a universal phenomenon and the Balkans cannot be excluded. Even communist China is selling off the majority of its 13 000 large and medium-sized state enterprises, as well as most of its smaller companies.6 At the same time the EU leaders should rigorously promote the entry of the Balkan countries into the EU, not least because

Nicholas V. Gianaris 301

the costs involved would be less than its ongoing military expenditure through the United Nations and NATO.7 Notes 1. 2. 3. 4. 5. 6. 7.

Aristotle, Politics, 2.5, 1263a 38; Barker (1959). Gianaris (1996, 1993a). Gianaris (1997a, 1997b). Gianaris, (1993b). Kohl (1993). Faison (1997); Walker (1997). For the answers to a related letter by the present writer from the British prime minister and the president of France, see, Gianaris, (1996, 1998).

References Aristotle, Politics, 2.5, 1263a 38. Barker, Ernest (1959) The Political Thought of Plato and Aristotle (New York: Dover). Faison, Seth (1997) `China's Leader Announces Sell-Off of State Enterprises', New York Times, 13 September. Gianaris, Nicholas V. (1996) Modern Capitalism: Privatisation, Employee Ownership, and Industrial Democracy (Westport, CT: Praeger). Gianaris, Nicholas V. (1988) The North American Free Trade Agreement and the European Union (Westport, CT: Praeger), ch. 10. Gianaris, Nicholas V. (1993a) Contemporary Economic Systems (Westport, CT: Praeger), ch. 4. Gianaris, Nicholas V. (1993b) The European Community, Eastern Europe, and Russia (Westport, CT: Praeger), ch. 8. Gianaris, Nicholas V. (1996) Geopolitical and Economic Changes in the Balkan Countries (Westport, CT: Praeger), p. 181. Gianaris, Nicholas V. (1997a) Ellada and Evropaiki Enosi (Hellas and the European Union) (Athens: Antonios Sakkoulas), ch. 4. Gianaris, Nicholas V. (1997b) Paidia kai Economia (Education and the Economy) (Athens: Antonios Sakkoulas), chs 6±8. Kohl, Helmut (1993) `European Integration', Presidents and Prime Ministers, vol. 2, no. 1, p. 14. Walker, Tony (1997) `Chinese President Embraces Market in Spirit of Deng', Financial Times, 13 September.