Investment, Growth and Employment: Perspectives and Policy

Investment, Growth and Employment

Investment—in both facilities and know-how—is essential for growth. Economists try to understand the forces that determine investment, but its behaviour is unruly; often the term ‘animal spirits’ is used to explain the resulting volatility. This volume presents a new set of studies to explain international investment behaviour and assess its impact on growth and jobs. The authors also examine policy measures aimed at reversing the climate of low investment that has characterised recent decades. The contributors examine how well standard models of investment work, the role of finance constraints, the effect of risk and uncertainty, the impact of alternative forms of corporate governance, the forces shaping the adoption of new technology, the impact of foreign direct investment, the effect of investment on the NAIRU, and the causal structure of investment and growth. Editors’ introductions to the different sections of the book provide comprehensive overviews of the main theories of investment, the impact of investment on growth and employment; they also examine the main questions raised for policy makers. Investment, Growth and Employment brings together in a single volume the main strands of work on investment and surveys the existing frontiers of knowledge. It will be of value to all students of, and researchers in, political economy, macroeconomics, the UK and European economies, and business economics. Finally, its argument that governments need actively to promote a climate of growth in the new Europe will be of interest to policy makers. Ciaran Driver is a Reader in Economics at Imperial College Management School, University of London and is the author, with D.Moreton, of Investment, Expectations and Uncertainty (1992). Paul Temple is a Lecturer at the Department of Economics, University of Surrey and is, with T.Buxton and P.Chapman, an editor of, and contributor to, Britain’s Economic Performance (2nd edn, 1998).

Investment, Growth and Employment Perspectives for policy

Edited by Ciaran Driver and Paul Temple

London and New York

First published 1999 by Routledge 11 New Fetter Lane, London EC4P 4EE Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 Routledge is an imprint of the Taylor & Francis Group This edition published in the Taylor & Francis e-Library, 2005. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” © 1999 Selection and editorial matter Ciaran Driver and Paul Temple; individual chapters © their authors All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Investment, growth and employment: perspectives for policy/ edited by Ciaran Driver and Paul Temple. p. cm. 1. Investments, Foreign. 2. Economic development. 3. Job creation. I. Driver, Ciaran. II. Temple, Paul, 1952–. HG4538. I65457 1999 332.67'3–dc21 99–19536 CIP ISBN 0-203-98420-X Master e-book ISBN

ISBN 0-415-19779-1 (hbk) ISBN 0-415-19780-5 (pbk)

Contents

PART I

List of figures

vi

List of tables

vii

List of contributors

ix

Preface

xi

Acknowledgements

xii

The determinants of investment

1

1

Overview: a survey of recent issues in investment theory CIARAN DRIVER AND PAUL TEMPLE

2

Finance, profitability and investment in manufacturing BRIAN HENRY, ANDREW SENTANCE AND GIOVANNI URGA

17

3

Credit rationing versus consolidation of financial structure JEAN-BERNARD CHATELAIN

30

4

Investment, uncertainty and industry structure RINA BHATTACHARYA AND PAUL HOPE

46

5

Corporate governance, investment and economic performance SIMON PECK AND PAUL TEMPLE

64

6

Uncertainty, macroeconomic volatility and investment in new technology OTTO TOIVANEN, PAUL STONEMAN AND PAUL DIEDEREN

81

PART II

The consequences of investment

2

95

7

Overview: investment, feedback and spillover CIARAN DRIVER AND PAUL TEMPLE

8

Components of investment and growth JERRY COAKLEY AND ANDREW WOOD

106

9

Foreign direct investment, innovation and economic growth within Europe RAY BARRELL AND NIGEL PAIN

117

Investment, growth and unemployment: modelling the supply side of the UK economy JAMES NIXON AND GIOVANNI URGA

131

10

PART III The policy lessons

96

148

11

Overview: a survey of key policy issues CIARAN DRIVER AND PAUL TEMPLE

149

12

Supply constraints and inflation CIARAN DRIVER AND DAVID SHEPHERD

157

13

Long-run effects of investment incentives MICHAEL SUMNER

174

v

14

Investment policy and the employers’ perspective KATE BARKER

180

15

The capacity to tackle unemployment JONATHAN MICHIE

185

16

The UK’s investment problem IAN BRINKLEY ANDSOTERIOS SOTERI

189

17

Investment and capital productivity in Europe and the US JAEWOO LEE

197

Index

207

Figures

1.1 1.2 1.3 2.1 2.2 4.1 5.1 7.1 7.2 7.3 7.4 8.1 8.2 10.1 10.2 10.3 10.4 10.5 10.6 11.1 16.1 16.2 16.3 16.4 17.1 17.2 17.3 17.4 17.5

Investment articles counted Rates of return and the cost of capital in the UK The marginal product of capital in a stochastic model Total UK fixed investment (volume index: trough of recession=100) Manufacturing investment recoveries (trough of recession 1992Q1=100) The marginal revenue product of capital and uncertainty Return on capital employed Gross domestic fixed capital formation as a percentage of GDP Machinery and equipment as a proportion of gross capital stock (all industries) Machinery and equipment as a proportion of gross capital stock (manufacturing) The impact of capital deepening on employment UK GDP and investment in buildings UK GDP and investment in equipment UK unemployment rate UK gross domestic product and trend: constant prices (log scale) The determination of equilibrium unemployment Factor volumes: percentage deviation from base Employment: percentage deviation from base after a 10 per cent increase in the money stock Unemployment rate: percentage deviation from base after a 1 per cent increase in the real cost of capital Industrial support in the UK, 1946/7–1990/1 Private sector and business investment (percentage of GDP using 1990 constant prices) Gross capital stock (£bn at 1990 prices) Public investment as share of national income (net public sector investment as percentage of GDP) Investment over two recoveries, 1981–1992=100 (1990 prices) Gross fixed income Investment in machinery and equipment Non-equipment investment Share of investment: goods versus services Capital productivity

2 4 10 17 18 48 72 97 99 99 103 108 108 132 132 136 142 143 143 151 190 191 191 193 198 199 199 201 203

Tables

1.1 2.1 2.2 2.3 2.4 3.1 3.2 3.3 4.1 4.2 4.3 4.4 4.5 4.6 4.7 5.1 5.2 5.3 5.4 6.1 6.2 6.3 6.4 6.A.1 6.A.2 6.A.3 7.1 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 8.10 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 10.1a 10.1b 10.1c

Alternative investment regimes Orders of integration Cointegration tests on alternative models Hybrid model: likelihood ratio tests REH models for investment Different regimes for debt level Sample sizes Estimation results Basic equation Impact of uncertainty Impact of uncertainty Impact of uncertainty Decomposition of percentage change in the investment— output ratio, 1984–1992 average Distribution of concentration ratios, 1980–1992 average Industries in sample A taxonomy of governance structures Stock market capitalisations, 1997 Ownership of common stock in 1990 (percentage of outstanding shares owned) Empirical studies of short-termism Country-wise descriptive statistics Variable definitions Three sets of results Estimated derivatives and elasticities of ln (St/St–1) with respect to explanatory variables Correlation matrix of some explanatory variables Wald restriction tests Results from ARCH estimations Reasons for foreign investment Data coverage Pairwise regressions: trace and eigenvalue test statistics Pairwise regressions: tests for significance of loading parameters (t-ratios) Trace and maximum eigenvalue test statistics α and β matrices from cointegrating VARs Restrictions on the α matrix (x2(r)) Trace and maximum eigenvalue test statistics α and β matrices from cointegrating VARs Tests for restrictions on the β matrix Tests for restrictions on the α matrix Global foreign direct investment stocks FDI flows in selected OECD economies ($ billion, period totals) Cross-border mergers and acquisitions, 1991–1995 The importance of foreign-owned firms, manufacturing sectors Accounting for inward-investment growth in Europe Selected data for US non-bank foreign affiliates in Europe 3SLS estimates of a West German production function with endogenous technical progress Within-industry spillovers from foreign firms in manufacturing Cointegration tests on unrestricted levels system Cointegration tests when linear homogeneity with respect to output is imposed Cointegration tests when homotheicity is imposed

2 24 25 26 27 35 41 42 52 53 54 54 55 56 58 66 67 67 75 85 86 87 88 91 91 92 101 109 110 110 111 111 112 112 113 113 114 117 118 119 120 122 123 126 128 138 138 138

viii

10.1d 10.2 10.3a 10.3b 10.3c 10.3d 10.4 10.5 12.1 12.2 12.3

Cointegration tests when Harrod neutral technical progress is imposed Levels and dynamic estimates of main coefficients Long-run Allen elasticities of substitution Short-run Allen elasticities of substitution Long-run price elasticities Short-run price elasticities Cointegrating regression for wages with productivity Dynamic wage equations based on labour productivity Stationarity tests: sample 1976Q4–1997Q1 Granger causality tests on ΔP and ΔC Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1, IV(2SLS) estimation 12.4 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1, IV(2SLS) estimation 12.5 Significance of the utilisation and constraint variables 12.6 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1980Q1–1992Q4 12.A.1 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1 12.A.2 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1 13.1 Measures of fiscal policy 13.2 Order of integration 13.3 Cointegration tests 13.4 Error correction models 16.1 Investment shares compared, 1960–1997 16.2 Productivity and real wage growth (annual percentage change) in the 1990s 16.3 Wages, investment and profits, 1990–1997 17.1 Average growth rates in gross fixed investment 17.2 Sectoral investment shares

138 138 139 139 139 139 140 140 160 161 162 164 165 166 169 170 175 175 176 177 190 192 192 197 201

Contributors

Ciaran Driver is a Reader in economics at Imperial College College Management School, University of London. He holds degrees in electrical engineering and systems, with a Ph.D. in economics. He has worked in the UK government economic service and as consultant to public and private bodies. He has held visiting positions in the US and Australia and has published widely in leading international journals and has authored and edited several books. Paul Temple is a Lecturer in economics at the University of Surrey which he joined in 1997 after working as Research Fellow in the Centre for Business Strategy, London Business School. He has worked as an economic policy adviser and has contributed to a wide variety of issues relating to the competitiveness of the UK economy. He is an editor of Britain’s Economic Performance (Routledge). More details of publications etc. can be obtained from website: http:// www.econ.surrey.ac.uk Kate Barker is Chief Economic Adviser at the CBI. She joined the CBI in 1994 from Ford of Europe where she was Chief European Economist. She is a former member of the Treasury Panel of Independent Forecasters. Ray Barrell is a Senior Research Fellow at the National Institute, and he has led the World economy team there for a decade. Prior to that he was an economic adviser at HMT for three years and an academic for ten. His research interests include the determinants of growth and location, European integration and the analysis of policy. He has published widely in these areas in books and in journals such as European Economic Review, Economic Journal, the Journal of Policy Modelling, and the Review of Economics and Statistics. Rina Bhattacharya is currently working for the International Monetary Fund in Washington. She holds a Ph.D. in Economics from Yale University in New Haven, Connecticut, USA. Her main areas of interest are open-economy macroeconomics and international finance. She has worked as a lecturer at the University of Sussex, and also at the UK Treasury and the Bank of England prior to joining the International Monetary Fund. Ian Brinkley is Senior Policy Officer in the TUC’s Economic and Social Affairs Department, with particular responsibility for developing and presenting the TUC’s economic policies and developing the TUC’s work programme on the labour market. He previously worked as a researcher on labour market policy at the University of Kent at Canterbury and at the Centre for Environmental studies. Jean-Bernard Chatelain (Ph.D.) is a researcher at the Centre de Recherche, Banque de France. Jerry Coakley is Reader in Financial Economics at London Guildhall University. He publishes and does research on the Feldstein—Horioka puzzle and capital mobility, tests for PPP, and estimating and testing nonlinear (threshold autoregressive) models of financial and other markets. Paul Diederen studied econometrics at the University of Amsterdam. At MERIT in Maastricht he wrote his Ph.D. thesis on ‘Technological progress in enterprises and diffusion of innovations’. As a post-doc. researcher at Warwick University he worked on investment and diffusion theory and on technology policy. He is currently employed at the Agricultural Economics Research Institute in The Hague. Brian Henry is Director of Research at the Centre for Economic Forecasting, London Business School. Previously he has been senior economic adviser at the Bank of England and an economic adviser at the IMF, as well as holding teaching posts at the London School of Economics, University College London and Cambridge. His research interests are in macro and monetary economics and in econometric modelling, areas in which he has published widely. Paul Hope is currently a Senior Economist at the Office of Water Services in Birmingham, UK. He previously worked at the UK Treasury and at the Bank of England. He holds degrees in economics from the Universities of Leicester and Manchester, and his interests include industrial economics and European Monetary Union. Jaewoo Lee currently works at the International Monetary Fund, and was previously Professor at the University of California at Irvine. Jonathan Michie is Professor of Management at Birkbeck College, University of London. Recent books include Firms, Organizations and Contracts: A Reader in Industrial Organization (edited with Peter Buckley, Oxford University Press,

x

1996) and Contracts, Co-operation, and Competition: Studies in Economics, Management, and Law (edited with Simon Deakin, Oxford University Press, 1997). James Nixon is a Research Fellow at the Centre for Economic Forecasting, London Business School. His research interests include macroeconometric modelling and the analysis of the economic policy. Previously James has been Chief UK Economist at the LBS, producing regular forecasts and commentary of the UK economy. Before joining the LBS he was an Economic Adviser at HM Treasury, working on the Budget Forecast team. Nigel Pain is a Senior Research Fellow at the National Institute of Economic and Social Research (NIESR). Formerly at HM Treasury, he joined NIESR in 1988. He has published widely on the determinants of international trade and investment, and on the impact of multinational corporations on the structure and performance of national economies. Further details are available on the NIESR internet site at http://www.niesr.ac.uk Simon Peck is currently a Research Fellow at City University Business School. His research embraces various aspects of the links between corporate governance and economic performance. Andrew Sentance is Chief Economist, British Airways plc. He joined BA in 1998 from London Business School where he was Director of the Centre for Economic Forecasting. He is a former Director of Economic Affairs at the Confederation of British Industry and a former member of the Treasury Panel of Independent Forecasters (‘the wise men’). He is also a Visiting Professor at Royal Holloway, University of London, and Chairman of the Society of Business Economists. David Shepherd is Senior Lecturer in macroeconomics in Imperial College College Management School, University of London. He has published in a number of leading economics journals, and is co-author of a key work on British manufacturing investment overseas. Soterios Soteri is a Policy Officer at the Trades Union Congress where he works on macroeconomic policy related issues. He has previously worked as an economist at the National Institute of Economics and Social Research and as a research officer at the Civil and Public Services Association. Paul Stoneman (BA (Warwick), M.Sc. (Econ.), (London), Ph.D. (Cambridge)) is Research Professor in Warwick Business School with research interests centring on the Economics of Technological Change and Technology Policy. He has published many articles in these fields and a number of books including the Handbook of the Economics of Technological Change (Basil Blackwell, 1995). He has also been involved on a practical level with policy making in the UK and has undertaken various studies for national and international government bodies as well as other private sector organisations. Michael Sumner has been Professor of Economics at Sussex since 1983. His main research interests are in macroeconomics, public finance, and firm behaviour. Otto Toivanen researches at the Department of Economics, Helsinki School of Economics, Finland. He has published in leading economics journals. Giovanni Urga is a Research Fellow in the Centre for Business Strategy of the London Business School. He is also Visiting Professor in Econometrics at the Economics Department of Bergamo University (Italy) since 1992 and at New Economic School (Moscow) since 1996. He was formerly a Lecturer at Queen Mary and Westfield College and Research Officer at the Institute of Economics and Statistics (Oxford). His research centres on econometric modelling, the economics of investment decisions, and the econometrics of financial markets and panel data analysis. Andrew Wood is Senior Lecturer in Economics at London Guildhall University. He has previously worked for the National Economic Development Office, the Treasury and Civil Service Committee, and was research fellow at Birkbeck College and South Bank University.

Preface

This book has its origins in a realisation that interest in capital investment, particularly in Europe and the UK, is growing rapidly. Moreover, it is an area where there is still a wide gulf between academic debate and the concerns of economic policy. Nowhere is this more readily apparent than in the contentious issue of whether, in the current context, greater investment would contribute to employment growth. To stimulate a debate aimed at bridging this gap, the editors called a conference in May 1997 at Imperial College London organised in cooperation with the London Business School. Key contributers to the investment literature, together with more policy oriented economists were invited to participate. Some of the contributions that day, updated and revised appear here. Although the book grew out of the conference it is, we hope, more than just a conference volume. Indeed we are pleased to include several pieces not presented at the conference and we have also added (by way of a guide to the literature) three overviews on the themes of investment theory; investment consequences; and on the implications of both for economic policy. The book is international in scope and contains several chapters with an international dimension. There is also however, a major focus on UK investment behaviour and the policy chapters reflect this. Indeed, many of the lessons in the UK may be becoming increasingly relevant for Europe more generally, where a major slowdown in rates of capital formation has attracted widespread attention. We certainly hope that the book will be of interest to more than the narrow spectrum of academic economists.

Acknowledgements

The authors express their gratitude to participants at a conference at Imperial College, May 1997, and all those especially at Imperial, Surrey, and London Business School, whose discussions have made this work possible. We would particularly like to thank the editorial staff at Routledge, whose encouragement made this book possible. We would like to remember our families, whose forbearance was an essential ingredient.

Part I The determinants of investment

Figure 1.1 Investment articles counted. Source: Social Science Citation Index.

1 Overview A survey of recent issues in investment theory Ciaran Driver and Paul Temple

Introduction Interest in capital investment has risen sharply in the 1990s after stagnating at a low level during the 1980s as may be seen from the international count of relevant journal articles in Figure 1.1.1 The preceding low level of interest may have something to do with the difficulty of breaking out of the narrowly defined ‘modern’ approach to investment, originating in Abel (1980). By contrast in the 1990s we have seen a divergence in approach, which has widened the scope of enquiry not only in respect of the causes of capital investment but also its consequences. In this first overview we deal with the determinants of investment. The overview for Part II addresses the consequences, while that for Part III considers some key issues for economic policy. Understanding investment Despite the rise in research activity shown in Figure 1.1, there has been no breakthrough on the empirical front; the ability to forecast investment expenditures seems as elusive as ever. The average root mean squared errors one year ahead for seven forecasting models is four times larger for non-residential real fixed investment as it is for GDP growth (Granger 1994). In some ways this should not surprise given the volatility of investment expenditure which entails a stock adjustment. Nevertheless, the forecasting performance of consumer durables, which has an even higher volatility than investment is rather better. This discrepancy may perhaps be due to the long-lived nature of capital equipment and structures or to the complex interactions between firms which characterise the investment decision. These complexities may be illustrated by listing some of the information requirements of an investment appraisal. These include: 1 Construction of forecasts and sensitivities in respect of macroeconomic variables, industry demand and market share, including conjectures of rival responses, public policy, effect of investment on existing activity and availability of complementary inputs. 2 Estimation of option values associated with acting (or not), including the technological and market positioning of the firm.

RECENT ISSUES IN INVESTMENT THEORY

3

3 Construction of a financial view including forecast cost of capital and long-run hurdle rates; conceptualisation of a market-based or managerialist philosophy on risk and liquidity constraints. In the absence of a social context there would almost certainly be huge variation in the way in which firms assess the future. In practice, although there is variation, it tends to be tightly contained. One explanation of this is the way in which political and social influences generate a robust consensus on the external environment facing firms. These long-run expectations characterise the regime that firms face in carrying out investment decisions. When these expectations change there is unusually high uncertainty due to the possibility of a regime change. We may illustrate this idea by adapting the theory of Marglin and Bhaduri (1989) to construct the matrix of investment regimes shown in Table 1.1. Table 1.1 Alternative investment regimes Public policy focus Investment

Demand

Profit

Less sensitive to current profit than to demand More sensitive to current profit than to demand

Co-operative 1950s Conflictual 1960s/early 1970s

Potentially co-operative? Post- 1980s Conflictual Late 1970s/1980s

The table illustrates a near-orthodox model of investment where the main determinants are demand and profitability with the role of demand crucial in translating current profit margins into expected profitability. Cooperative outcomes between labour and capital are possible where the likely influence of demand growth is to increase confidence in sustained profitability. In Kaleckian fashion, cooperative outcomes are unstable since they lead to a profit squeeze. The restoration of profitability tends to occur through a focus on profit-enhancing policies and accompanying destruction of demand. The resultant rise in profitability does not automatically entail cooperative growth as confidence may only be restored by a timely switch to a demand-led policy focus and a reduction in uncertainty. However, many economic variables—capital intensity, wage adjustment and expectations of inflation—adjust very slowly (Bean 1989; Wardlow 1994; Blanchard 1997). Public policy is also slow to adjust because credibility is related to policy consistency. These considerations limit the power of equilibrium analysis. If such forces do in fact underlie the longer-term cycles in advanced economies it can readily be understood why the formal models of investment, despite their sophistication, tend to predict poorly in the medium term.2 Most theories of investment are, at bottom, simply dynamic versions of an equilibration between the expected marginal revenue product of capital and its expected marginal cost. But neither of these schedules are likely to evolve in easily predictable ways. In forming expectations about the marginal revenue product, the corporate players have to weigh up, as in the Marglin and Bhaduri model, the relative influence of current profitability and demand conditions—no easy task. Certainly, the ex post realisations do not always bear out the simple story of equilibration between return and cost even in the medium run. In the UK, for example, as Figure 1.2 shows, the series for average profitability and average cost of investment have been diverging for over twenty years, suggesting the need for a disequilibrium analysis and the presence of constraints (Schultze 1987). At face value it might appear that the rising gap between the marginal revenue product of capital and the cost of funds should have produced an investment boom. The fact that this has most definitely not occurred needs to be explained. Possible candidates are measurement errors, the existence of constraints on investment or the occurrence of a regime shift. One important idea is that as capital flows have been liberalised, the opportunity cost of funds needs to be calculated as the rate of return on foreign assets (Young 1994; Cowling and Sugden 1996).3 If domestic investment has to meet a hurdle rate equal to this return, it could explain the coincidence of rising profitability and stagnant investment. Other explanations such as financial constraints or uncertainty are discussed in some of the chapters below. In this overview we review some of the main aspects of modern investment theory before commenting on the individual chapters in the book. Among the most influential (and relatively recent) developments in orthodox theory have been: 1 the development of dynamics and the testing of q-theories; 2 the incorporation of constraints; 3 models of investment under uncertainty. In reviewing these topics we will use their simplest representative form and offer a flavour of intuition along with the formal structure of a model.

4

CIARAN DRIVER AND PAUL TEMPLE

Figure 1.2 Rates of return and the cost of capital in the UK. Source: Bank of England.

Dynamics Much recent empirical work on investment follows the ‘adjustment cost’ approach which sees the investment decision as the optimal forward-looking tracking of a stochastic target. Earlier work—the flexible accelerator— stressed that there was a tradeoff between immediate adjustment to a target and the cost of rapid adjustment. The more modern version originating with Abel (1980) is in the same mould and can be simply described by a brace of equations obtained by differentiating the value of the firm (Vt) conditional on the previous period capital stock (Kt–1). The value of the firm may be written in dynamic programming form as the current value of profits and the continuation value representing the expected value from time t+1. To save on notation we ignore depreciation and discounting and consider only the fixed factor K. Thus: (1) given the assumption of zero depreciation. Differentiating Vt with respect to the ‘state variable’ Kt–1 gives the Euler equation: (2) The term on the left is the shadow value of one unit of capital. Repeated substitution gives the familiar condition that the shadow value of capital is the sum of the future marginal values of that capital. Differenting Vt with respect to the ‘control variable’ It gives: (3) Combining (2) and (3) gives (4) This states that the shadow value of capital is the marginal effect of investment (rate of change of capital) on current profits; this is the cost of investment. In a static framework that cost of investment is just the purchase price. In a dynamic framework, the right-hand side is the sum of the purchase price and an adjustment cost that is some convex function of investment, C(I). Equations (1) to (4) offer a variety of approaches to estimating investment equations. Three basic models have been used (a) First, equation (4) has been estimated directly by using stock market data to represent the left-hand term which is the shadow value of capital. The average valuation of the capital stock as a ratio of its replacement value (Tobin's average q) has been used to proxy the marginal value of q. If adjustment costs C(I) are assumed quadratic in the investment rate (I/K) so that marginal adjustment costs are linear, it is easy to obtain an equation from (4) which relates (I/K) log-linearly to q and to the real cost of capital.4

RECENT ISSUES IN INVESTMENT THEORY

5

(b) An alternative method is to dispense with observed stock market data and to attempt to proxy marginal q by an auxiliary equation which sums future marginal values of capital. This can be obtained by forecasting future marginal revenue products using an assumed representation for the production function and an estimated vector autoregression which links successive time periods (Abel and Blanchard 1986). (c) A third method is to eliminate the shadow value of capital from the investment equation by using a differenced version of (4). Using (2), the left side can be written as δπ/δKt. Given a representation for marginal revenue product and the adjustment cost function, and using rational expectations to eliminate the expectation, this equation can be estimated without recourse to data on marginal q. However, it may pertinently be asked how account is taken of the under-utilisation of capital in this framework. With constant returns, perfect competition and stable prices for capital goods, the equation reduces to a condition relating adjustment cost now to that in the next period, i.e. investment is simply driven by the expected path of adjustment costs. There has been some disagreement in the literature over the most promising approach. It seems generally agreed that the qtheory approach has disappointed empirically, either because of a failure of stocks to mirror fundamentals or for broader reasons of investment strategy (Blanchard et al. 1993; Chirinko 1993).5 As for the other approaches, Blundell et al. (1992) favour the Euler equation approach partly because it avoids the auxiliary assumptions that have been used to measure the shadow value of capital. It appears, however, that most estimated Euler equations do not satisfy their own theoretical restrictions and may be ‘mongrel’ relationships (Nickell and Nicolitsas 1996; see also Chirinko 1993, note 40). A serious problem with much of the above theory is the reliance on restrictive functional forms, in particular the assumed convex quadratic form for the adjustment cost. While this makes the solution of the differential equation system tractable, deviations from it can render the solution unstable. Furthermore the a priori basis for it is thin (Rothschild 1971; Maccini 1987). Convex costs may be an element of firm behaviour, e.g. in the sense that rapid expansion induces managerial overload (Penrose 1959) or as in R&D projects where parallel rather than serial expenditures might reduce learning effects (Scherer 1986). More generally the model is explained in terms of disruption to production flow. But if these aspects of cost were binding constraints on investment at firm or business level we might expect there to be an extensive managerial discourse on the problem. This appears not to be the case. Recent empirical work has confirmed the lack of evidence for convexity (Hamermesh and Pfann 1996). It has also been known for some time that using quadratic adjustment costs in macro equations implies an implausibly large adjustment cost (Bosworth 1981). It seems therefore that convex costs of adjustment can be rationalised only to the extent that they mimic aggregate investment response.6 An alternative approach to explaining lagged adjustment is external adjustment costs, representing, for example, a rising supply price for capital goods in a monopsonistic setting or where competitive firms face similar demand variation. Chirinko (1994) shows that in this case the q-theory formulation is modified to include future investment expenditure rates and supply elasticity of capital goods. However, for this theory to be a serious contender in explaining general investment lags, there would have to be marked differences between the investment lag structures of small open capital importing economies and those relatively self-sufficient in capital. This possibility does not seem to be observed, though neither has it been extensively studied. Yet another approach explains lagged adjustment by time to build (Taylor 1982; Pindyck 1993). However, while this may be important, it cannot be the full story; delivery lags seem similar for different industries but the lagged adjustment shows considerable cross-industry variation (Abel and Blanchard 1983). In any event, the delivery lag model cannot be used to explain the lag structure of investment authorisations. While the lags here are shorter than for expenditures, there is still a lag structure to be explained. Some of the more recent work on investment stresses the role of nonconvexities such as irreversibility and also the role of lumpiness of capital expenditure.7 Research at individual plant level has confirmed that most firms are characterised by ‘zeros and ones’, i.e that investment tends to be a discrete activity, at any rate for large projects (Nilsen and Schiantarelli 1996). The probability of an investment burst for a plant increases with the time since the previous burst. Firms also invest disproportionately heavily when capital shortage is high than when it is low or negative. The aggregation of such lumpy investments across heterogeneous firms can be shown to create a smoothed adjustment that appears to explain aggregate dynamics better than existing theories. Simulations also suggests that the longer-term effect of tax policy may be much stronger than conventionally measured because of the problems in aggregating dissimilar responses (Caballero et al. 1995).

6

CIARAN DRIVER AND PAUL TEMPLE

Investment under constraints Micro-constraints The starting point for many theories of investment is some version of a neoclassical model. Confronted with the data these models perform poorly. Data for the main OECD countries suggest that the three main variables, capital stock, output and relative factor price, are not cointegrated.8 However, there is evidence that an increment of output affects the growth rate of the capital stock permanently (Ford and Poret 1991). Other research has found conflicting evidence as to whether gross investment (regarded perhaps as a proxy for replacement investment in long-run equilibrium) is cointegrated with output and the cost of capital. (Lomax 1990; Henry et al. in this book, Chapter 2; and Sumner, Chapter 13). In response to these findings, theorists have argued that the adjustment process is in some way supply-constrained, since the adjustment of capital is so slow. This has naturally led to the incorporation of constraints in the investment specification. Both the management and method of financing investment projects offer possible lines of development on the basis that such factors cannot as easily be ‘bought in’ as other factors such as physical plant and equipment.9 Both managerial and financial constraints are of interest because they may be both highly non-linear and firm specific in nature. While therefore of great potential importance in resolving some of the paradoxes of investment behaviour, this has led to considerable problems when it comes to empirical testing. In basic models of finance, the provision of finance plays a role orthogonal to the investment decision. According to the Modigliani and Miller theorem, the firm’s capital structure is irrelevant for its market value. This striking result can be thought of in terms of an arbitrage condition in a situation of zero transactions costs: if firms are maximising profits subject to a production function, then any change in value resulting from a different financial structure would give rise to arbitrage opportunities. While the list of violations to the assumptions of the Modigliani-Miller result is a substantial one, the literature has focused on the idea that capital structures reflect information asymmetries between management and owners and/or lenders. Typically, management may be expected to possess more information about the value of the firm’s investments (and of their own efforts) than either debt or equity holders. Schianterelli (1996) has pointed to two major results from this literature. The first is that uncollateralised loans will be more costly than internal finance because lenders will demand a ‘lemons premium’. The second is that the size of this premium will be inversely related to the borrower’s net worth—defined to include cash, liquid assets, and the portion of illiquid assets (such as the future profits of the firm) deemed suitable as collateral. The latter aspect will tend to amplify any financial shock to production and investment, since these impact on net worth as well as through the cost of capital via the conventional market rate of interest. This is the so-called ‘financial accelerator’ effect. Given the fact that at any point in time individual firms may or may not be suffering from a financing constraint poses empirical estimation problems in the testing of either of the two hypotheses above. In practice much of the methodology has stemmed from the contribution of Fazzari et al. (1988) who used an a priori initial classification of firms into constrained and unconstrained regimes on the basis of dividend pay-out ratios. Arguably, high dividend firms are likely to fall into the latter regime. Other means of proxying for the probability of a particular firm falling into one or other regimes include firm size and age (Devereux and Schianterelli 1990), bond rating (Whited 1992), and affiliation to a Japanese industrial group (Hoshi et al. 1990). Many of the studies assume that any given firm is permanently allocated to one regime or another over the sample period; this runs counter to the idea of important financial accelerator effects. An exception here is Hu and Schianterelli (1994) who employ a ‘switching’ regression which is dependent upon both firm characteristics and macroeconomic conditions. In general, empirical implementation has followed the approach outlined above using either Q or an Euler equation. Typically, convex adjustment costs are assumed. A fundamental problem with the Q formulation is in devising a suitable control for investment opportunities, especially if it is given that stock market valuations may not reflect all that is known about future profitability. In that event, measures such as cash flow may be proxying for this missing information. Gilchrist and Himmelberg (1994) have developed the Abel and Blanchard (1986) approach described above (page 7) in an attempt to control more tightly for investment opportunities. The other alternative is of course the Euler equation itself. For firms which are financially constrained, the standard Euler equation (2) is mis-specified; in theory, as Hubbard (1998) points out, an augmented Euler equation framework allows for the impact of changes in cash flow or net worth to vary systematically. For example, in Whited’s 1992 model, the shadow cost of external financing varies depends upon the firm’s coverage ratio. This is broadly the approach developed by Chatelain in Chapter 3 below. The research paradigm on financing constraints described above is now a considerable one, as the surveys of Schianterelli (1996) and Hubbard (1998) make clear. Results are probably robust enough for the impact of information problems in capital markets to be taken seriously, although a dissenting note was recently expressed by Kaplan and Zingales (1998). However, thus far implementation has been restricted to the basic neo-classical model described above and have not begun to explore the implications of irreversibility and options based investment models outlined below.

RECENT ISSUES IN INVESTMENT THEORY

7

Managerial constraints of the sort introduced by Penrose (1959) are discussed in Peck and Temple below (Chapter 5) but have not featured strongly in the empirical literature, although they have informed a debate centred on the motivation of modern management which remains important. Indeed the positive relationship between cash-flow found in the asymmetric information approach may also be a consequence of discretionary managerial behaviour favouring sales or growth (which are linked to salaries and promotion opportunities). As Schianterelli (1996) observes, costly actions taken to control management may be difficult to distinguish from the costs attached to adverse selection. Saving as a macro-economic constraint Financial constraints clearly operate at the micro-level. That individual firms may find themselves financially constrained does not imply that profitable investment opportunities run to waste, since other firms may take them. On the other hand, such an optimistic scenario cannot be assumed since such opportunities may be highly firm specific. Whatever, financial constraints need to be clearly distinguished from a constraint imposed by saving—the resources created for investment— because the incomes created by current sales are not immediately consumed. In contrast to the idea of a financial constraint on individual firms, the idea of a savings constraint is primarily a macro-phenomenon, impacting on firms via the cost of capital. However, the level of aggregation at which such a constraint may operate—be it nationally or globally—is a matter of considerable interest. In the standard economic model of a closed economy with full employment, subsidies to investment or a reduction in taxes on saving lead to similar results. This conclusion needs to be severely modified, of course, under Keynesian conditions of less than full employment where attempts to stimulate savings are likely to prove self-defeating because the deflationary impact of the rise in saving more than offsets any tendency for rates of interest to fall. Policies directed at investment directly contribute to effective demand and employment, and must under these conditions be favoured. As noted graphically by William Vickery (1993), savings are not like a sack of potatoes—if they are not used they disappear in reduced income. This point of departure is of course well known; it implies that interest rates are not solely determined by the interplay of demand and supply for loans and that an independent investment equation is needed to obtain model closure.10 The orthodox view, by contrast, argues that savings have an important effect on investment at the global level and that the reduction in savings rates, especially by loose fiscal policy since the late 1960s, has contributed to lower investment in OECD economies (Jenkinson 1996). This orthodoxy might not apply to a small open economy in the presence of a global capital market. In such a case, attempts to subsidise domestic saving in order to raise investment merely result in a substitution of domestic saving for capital imports; domestic investment would again be independent of levels of domestic savings. However, capital markets may not yet be fully global— international portfolio investment is not greatly diversified internationally partly because of exchange rate risk (Feldstein 1994).11 It is difficult to distinguish empirically between these views because they each comprise a complex of hypotheses. However, the failure of a standard model of investment to perform adequately suggests that there is a need for an independent investment equation. Furthermore, the weak elasticities of investment with respect to interest rates also undermines the orthodox view. Finally work on public dis-saving in the US suggests that fiscal deficits have actually contributed to gross private domestic investment and to national saving (Eisner 1994). Uncertainty The work on dynamics has to some extent been superseded by work on irreversibility and uncertainty. Some two decades after Nickell first argued that irreversibility combined with risk could offer an alternative explanation for adjustment lags, the idea has become relatively commonplace and forms the basis for some recent models (Dixit and Pindyck 1994; Abel et al. 1996). The traditional literature on investment under uncertainty consisted of a myriad of models in which there could be a positive or negative bias or no bias at all to investment depending on the assumptions chosen (Aiginger 1987; Driver and Moreton 1992). To simplify, but not unduly, the three main ways in which demand uncertainty matters are: first, when there is risk aversion, second, where marginal profit with respect to capital depends non-linearly on demand and, third, when the capacity utilisation rather than price is the equilibrating variable in the event of a shock. In the first case, demand uncertainty usually results in a lower optimal capital stock. Survey-based evidence also suggests that this may be the primary method by which uncertainty has an effect on investment, at any rate for large projects (Aiginger 1987). In the second case, a mean-preserving spread will cause a bias, but the direction and magnitude will depend on technology, demand and the form of the stochastic error; some examples of bias are given below.12 The most quoted effect is the result that uncertainty raises investment for price-taking firms, but more generally the effect is ambiguous. To see this, consider first the case where, under risk neutrality uncertainty will not cause the optimal decision to differ from that under certainty. In the case of uncertainty over the output price (p), with capacity or output given by q we may write expected profit as:

8

CIARAN DRIVER AND PAUL TEMPLE

where the integral is over the full range of p. Differentiating this w.r.t. q Equating to zero gives E[p]=c′(q). This output rule is the same as under certainty if E(p) is the same as the certainty price. However, changing the context slightly, e.g. introducing a tax rebate which makes the downside different from the upside, would introduce a bias. A bias would also be introduced if we relax the assumption that p is exogenous: If we make q=upε, with mean of u=1: The derivative of this with respect to q will contain a term in u, not present in the certainty case, where u=1. For these kinds of models, standard results with q the decision variable and p the stochastic variable may be derived (Rothschild and Stiglitz 1971; Aiginger 1987)

The intuition behind this can be seen by assuming that the marginal impact on profit of an extra unit of q is convex in the price. This means that the expected value of the marginal profit with respect to q is greater than the marginal profit under certainty. Under certainty the point of zero marginal profit will be met sooner, i.e. with lower q and so, q(opt) > q(cert). For example, consider a constant elasticity demand curve shifted by a random term u. The marginal revenue function is: This is concave in u for elastic demand and so the optimal q is smaller than under certainty. Extension to a two-factor setting with perfect competition and price uncertainty is sometimes known as the Hartman-Abel model (Hartman 1972; Abel 1983). Then optimal level of K will be determined by: where K is capital fixed in advance of the price but where L/K could be varied ex post. The function in the square brackets is linear in p if L does not vary with p, e.g. as in a fixed coefficients model. If L rises in response to higher P—as would be indicated by the usual marginal productivity conditions if there is flexibility—FK will rise too and the function in square brackets will be convex in p, imparting an upward bias to capital input.13 The intuitive reason for the result is that as the price rises and labour input has to be increased to suboptimal levels, the value of a unit of capital is increasing non-linearly in p.14 The third and perhaps most interesting case is the Newsboy inventory model applied to capital input. The previous models have assumed that there is no rationing. Firms can always meet demand and price adjusts upward so that demand is met. This would not appear to be always sensible at least at the level of the individual firm, where a forecasting error could cause a firm to run out of capacity. Because of this we must consider a new set of models where sales are distinct from production. This turns out to have the radical implication that we cannot use the usual convexity/concavity formulation of the RothschildStiglitz condition because we now have three distinct variables: the decision variable, e.g. output; the stochastic variable, e.g. price; and a new variable distinct from output-demand. This context is sometimes known as a stochastic rationing context and the usual condition imposed is that Sales=min[Production, Demand] Outlined below is the simple Newsboy model, adapted for the case of capital input, where the following notation applies: Output price (p); capacity (y); capacity cost per unit of production (c); demand under certainty (D0(p)); stochastic shift parameter (a). Setting marginal cost with respect to y equal to expected marginal revenue, with an absolute capacity constraint at y=aD0,

where Z—the ratio of capacity to mean demand—is an inverse indicator of expected utilisation. Note that Z>1 implies that the firm plans to hold excess capacity. For a symmetrical distribution of a, F(1)=1/2. It then follows that:

RECENT ISSUES IN INVESTMENT THEORY

9

Not surprisingly, a distribution with larger tails will for any given c/p ratio have an optimal Z that is further away from unity. For example, if p>2c, there will be planned excess capacity.15 With fixed price margins, the effect of uncertainty is to encourage investment only if the unit profit at full capacity is above a threshold; otherwise the bias is negative. Real options In the Newsboy model above, irreversibility of capital is assumed. But the model is limited to two stages: investment followed by a realisation of demand. In recent years a new literature based on real options has added a new element in the context of a model with at least three stages: investment possibility one; investment possibility two; realisation of demand. (Dixit and Pindyck 1994). The basic idea of real options is that firms possess valuable opportunities which can be exploited with some discretion as to time but which once exercised are at least partially irreversible. The exercise of the option (investment) thus implies a cost which should be added to the usual user cost of capital when deciding on the timing of investment. There is a value to waiting where the uncertainty is partly resolved by waiting. Waiting makes costly (irreversible) mistakes less likely and this explains lagged adjustment.16 The basic Dixit and Pindyck model is set out in Appendix 4.2 to the chapter by Bhattacharya and Hope, Chapter 4 in this volume. The derivation of an investment equation from this perspective is not straightforward. Risk will induce caution by raising the threshold or investment trigger. But increased risk will also raise the variance of demand (and value of the firm) thus increasing the chance that the threshold will be exceeded in any period. Typically investment is delayed by risk (Pindyck 1991). However, the firm also has to assess the cost of waiting in terms of the erosion of market opportunities whether by time or by competitors’ actions. This measures the cost of not exercising the option to invest which is difficult to measure. One strategy is to relate the threat of new entry to the realisation of a favourable profit outcome.17 The effect of risk on the size of incremental capacity choice is also negative in imperfect competition (Pindyck (1988, 1991, 1993), though the effect on the capital output ratio is less clear cut. The Hartman-Abel convexity effect which tends to increase investment under risk dominates only under perfect competition and constant returns. In that case the marginal revenue product of capital is independent of the capital stock and thus irreversibility can never be problematic for the firm. More recently, real option theory has been subjected to an important extension in that the early version of it assumed no constraint on expansion. A more general approach can take into account the possibility of both irreversibility and constraints on expansion (Abel et al. 1996). The firm then has to bear in mind, when making its investment decision, not only irreversibility but the prospect of being short of capital if the realisation of demand turns out to be high. Abel et al. argue that this requires a consideration of an option to expand. To put it somewhat differently, earlier discussions such as Pindyck (1991) focused mainly on the decision of when to exercise existing options.18 The advance on this is to suggest that the decision to obtain an option is itself part of each investment process. Abel et al. explain that in the context of a two-period investment model, the ex ante investment may be no longer appropriate in the light of the realisation of the stochastic variable e. In the second period, one might prefer to sell part of the capital invested or exercise a right to buy more at a prearranged price. This complication results in the following expression (Abel et al., expression 17) in place of the Jorgenson user cost term (c): where b, bL, bH are the first period purchase price of a unit of capital and the corresponding selling and buying prices respectively; F(e) is the distribution function of the underlying stochastic variable; and RK is the marginal return on capital installed which may have to be evaluated at a nonoptimal level of the capital stock. The terms eL and eH are the critical values of the stochastic variable at which the original capital is no longer optimal ex post: either because the return is no longer greater than the return from selling the capital or because the return has risen to the rental on new capital purchased at the option price. Capital should then be bought or sold until the marginal return equals the lower or upper rentals bL or bH. This is represented by Abel et al. in their figure 1 reproduced below as Figure 1.3. There are some hints here as to how risk might affect investment. The firm has to bear in mind the cost of exercising the option and the cost of failing to obtain options.19 Abel et al. (1996) remark that ‘although the NPV rule is theoretically correct it is very difficult to implement in practice’ (p. 761). The reason is that the option values are difficult to compute. The theory here reflects, in a dynamic context, the same dilemma illustrated in the Newsboy model: how to balance the likelihood and net gains of excess capacity on the one hand and insufficient capacity on the other. So far we have seen little empirical work that takes these points fully into account.20 And the theory so far developed, despite its increased realism in dealing with options, is limited to market clearing behaviour. The models of uncertainty reviewed above are essentially models of risk in that the probability distribution or stochastic process is well determined. This ignores the problem of fundamental uncertainty stressed in post-Keynesian writing, where it is argued that the properties of the stochastic system are not time invariant. Some empirical evidence for both positions is available in that business decision-makers offered a choice between a vision of calculable probabilities and fundamental

10

CIARAN DRIVER AND PAUL TEMPLE

Figure 1.3 The marginal product of capital in a stochastic model (Abel et al. 1996).

uncertainty seem to split roughly two-thirds in favour of the former (Aiginger 1987). One implication of fundamental uncertainty is that investment decisions may need to be modelled in terms of bounded rationality or in terms of conventional behaviour (Heiner 1983; Driver 1992; Conlisk 1996). A useful discussion of these issues may be found in the debate between Crotty (1996) and Fazzari and Variato (1996). Authors’ contributions In this part of the book we take forward some of the themes discussed above, in particular those of constraints and uncertainty. We add a further emerging theme—that of investment as an aspect of technology adoption. The chapter by Brian Henry, Andrew Sentance and Giovanni Urga (Chapter 2) is a splendid example of the opportunities and challenges in estimating stable investment relationships using the standard cost-of-adjustment approach. Their models offer considerable flexibility, allowing for imperfect competition, financial constraints both internal and external, represented by terms in undistributed profits, a liquidity ratio and a gearing ratio. They are catholic in their use of proxies for or alternatives to the Tobin-q variable using, for example, profitability. Finally they consider whether direct survey-based expectations offer additional explanatory power and whether firms make investment decisions on the basis of rational expectations of the explanatory variables. The results of this study may be divided into the tests on long-run cointegrating relationships and the findings from the dynamic estimation. The authors seek a cointegrating relationship between gross UK manufacturing investment (which is treated as an I(1) variable) and various real and financial indicators. One result is that the neo-classical model and the averageq model (with gearing) do not cointegrate. The best-performing long-run models are those involving output and profits and (with an incorrectly signed relative price term) output and liquidity.21 What are we to make of the finding that profitability and output are possibly the main long-run determinants of investment? On the one hand these variables hardly surprise in empirical investment functions—they have been offered as competing or complementary determinants at least since the 1950s when serious econometric work on investment began. But on the other hand it is instructive that these variables are only loosely related to the economic theory with which the chapter began—the standard cost-of-adjustment approach. In the chapter the authors introduce profitability as an empirical counterpart to Tobin’s q. This seems defensible enough in a disequilibrium context where profitability is diverging from the cost of capital. The theoretical relationship between profitability and firm investment behaviour is, however, not well understood. There may well be a variety of responses depending on whether a firm is undertaking defensive investment in response to low profitability or reacting with expansionary investment to higher profitability. It seems likely also that some firms will operate in submissive mode where investment is curtailed in response to rising profitability elsewhere. The balance of these effects is picked up in this chapter as a positive relation between investment and profitability; it would be of interest to investigate this in a panel context. As for the real output variable, the authors give four separate reasons for its inclusion. On the one hand there is the standard argument that firms may be output constrained in which case the desired capital stock is directly determined by (exogenous) output. Second, under imperfect competition, output affects the marginal revenue product of capital. Third, a similar effect follows in a union-based model. Finally, anticipated output constraints may alter the optimal path of adjustment. It would be good to be able to discriminate between these explanations of the oldest and most robust empirical fact in investment studies; in the current chapter it seems that only the union-based interpretation is rejected.

RECENT ISSUES IN INVESTMENT THEORY

11

In some ways we may interpret this chapter as a first-stage clearing of the ground in an attempt to find a set of variables which are data admissible. The reader must judge to what extent the results are compatible with standard theory. Certainly the difficulty in finding robust cointegrating relationships—and the presence of a time trend in those identified—suggests that it is not only the dynamics of investment behaviour that remain poorly understood. The dynamic results that the author present are important and interesting. They find evidence that forward-looking behaviour matters, but that the forward-planning horizon is limited to three quarters.22It would be interesting to see whether the VAR relationship for the determinants of investment imply forecastability beyond a three quarter horizon. If this were the case it is not clear that the authors’ findings can be said to confirm rational expectations behaviour. In any event the use of a three quarter ahead horizon poses a difficulty for the cost-of-adjustment model which generally assumes an infinite horizon. If investment is based on short-term forecasts only it is easy to see why the omission of terms in animal spirits and long-run uncertainty could render the investment equation unstable. Such omissions may account for the model’s failure to explain adequately the behaviour of investment in the 1992–1996 period.23 Jean-Bernard Chatelain (Chapter 3) also addresses the issue of financial constraints. He distinguishes between a rationing ‘debt ceiling regime’ (where the debt-equity ratio is endogenised as a function of expected profits acting as collateral) and a ‘consolidation regime’ where there is some incentive for firms to repay debt rather than invest because their average cost of capital is thereby reduced. With firms in the consolidation regime, tax policy will not be very effective, as the provision of funds will simply allow firms to repay debt faster. This contrasts with the case of finance rationing where tax policy should have a strong impact. Certainly firms in France and elsewhere have been retiring debt and decreasing the debt-equity ratio at the same time as showing considerable restraint in regard to investment. The question is to what extent this can be rationalised as a response to the differential movement in interest rates and profit rates such that debt is retired as the former rises relative to the latter. For empirical implementation Chatelain considers a sample split based on the extent of debt repayment over the latter years of the sample; high debt repayment suggests consolidation. Another split is based on whether the interest rate and profit rates are such that the consolidation would be indicated. From a panel of 292 French firms, Chatelain seeks evidence of an endogenous debt ceiling for the non-consolidation regime. The results are clearly preliminary, but at this stage the significance of the important ζ parameters are not as clear-cut as might have been hoped. This may reflect the assumptions made in respect of how collateral values are taken into account. Chatelain’s chapter is largely concerned with the possibility of debt repayment as an alternative to investment in a scenario of high interest rates. It is, however, of interest to note that firms have increasingly taken to buying back their own equity as well. While the latter phenomenon can clearly not be explained in the Chatelain model, it perhaps indicates that there are nonfinancial factors driving the slowdown in investment which in turn get reflected in the retirement of both debt and equity. Information related problems are of course at the root of the short-termism debate which surfaced in the UK and the US during the 1980s, and this is also discussed by Simon Peck and Paul Temple in Chapter 5. These economic systems differ substantially from those found elsewhere in Europe or Japan in that ownership is not so concentrated and hence incentives to monitor management tend to be weaker. That these type of systems may accentuate information related problems is at the heart of the short-termism debate. The result is a tendency for financial markets to systematically under value more distant profit flows. The role of uncertainty in investment is addressed in Chapter 4 by Rina Bhattacharya and Paul Hope. They use panel data on 103 manufacturing industries in the UK, using the survey data from the main UK Business database, that of the Confederation of British Industries (CBI). Their theoretical section brings out clearly the many ways in which uncertainty can influence investment and the ambiguity of the direction of its influence. Although they do not consider the disequilibrium or rationing framework (see above) they discuss how risk aversion, non-linearities, and option theory can influence investment. The uncertainty variable used is a direct survey question in the CBI database. Although there is some debate as to whether this variable represents uncertainty or lack of confidence, the authors condition the results on a real variable (capacity utilisation) to take account of a possible variety of responses. Liquidity constraints and cost of capital effects are also entered with significant coefficients and expected signs. An interesting aspect of this chapter concerns testing the sensitivity of investment to uncertainty according to the degree of imperfect competition. Caballero (1991) has argued that a negative irreversibility effect can only be obtained for the case of imperfect competition and the strength of this should depend on the degree of monopoly power proxied by the elasticity of demand. Ghosal and Loungani (1996) and Guiso and Parigi (1996) have investigated this with contradictory results.24 In this chapter the finding is that the influence of uncertainty on investment is negative for less concentrated industries. The authors note that this is the reverse of that predicted by Caballero and suggest that the discrepancy may be due to increasing returns to scale in the highly concentrated industries.25 Another possibility is that demand shocks are accommodated by quantity adjustments, i.e. variation in capacity utilisation rather than price. In that case, firms with market power will be more likely to hold more reserve capacity, as in the Newsboy model, than firms with lower profit margins.

12

CIARAN DRIVER AND PAUL TEMPLE

The relationship between technology and investment has often been mooted in the literature but rarely developed. Paul Stoneman has long argued for an approach that sees the duality of gross investment and technological diffusion. Indeed it can be argued that diffusion theory anticipated some recent concerns in investment theory in that the Mansfield or David models emphasise the roles of risk and firm-specific factors in explaining adoption (Stoneman, ed. 1995). In Chapter 6, Otto Tovanen, Paul Stoneman and Paul Diederen conclude Part I by investigating the effect of uncertainty and volatility on a more narrowly defined investment category—that of industrial robots. These authors use unbalanced panel data on 16 countries’ stock of robots over the period 1981 to 1993 to estimate an epidemic diffusion model in which volatility variables affect the adjustment speed of adoption to close the gap between the actual and target stock. Conditional volatility variables, i.e. estimates of the forecast variance (ARCH) are constructed, but at this aggregate level most of the forecast variances turned out to be time invariant and accordingly unconditional variance was employed instead. The results suggest that taking account of volatility doubles the mean adjustment lag from about 5 years to 10. This result is driven by volatility in the price of robots which also includes exchange rate volatility. Other indicators of volatility such as GDP are not significant although negatively signed. It is possible that this reflects the correlation between volatility and structural change with new technology investment such as robots being favoured by a loosening of the industrial structure (Sentance and Urga 1997). Apart from volatility, the authors allowed the rate of inflation to affect the adjustment speed to test the prevailing orthodoxy of the virtuous effect of low inflation. The result here indicated a perverse effect with high inflation encouraging faster adoption of robots. This is perhaps not all that surprising. It has been noted elsewhere that high inflation might encourage quality enhancing investment in a context where inflation might facilitate a corresponding widening of the profit margin (Wardlow 1994). Overall these contributions take the debate on investment forward in a number of respects and certainly provide a snapshot of available knowledge. More detailed empirical work will be needed at all levels of aggregation, but particularly at the micro level, to fully resolve the issues raised in this introduction and in the chapters which follow. There is space for both formal econometric work, simulations, decision theory, and case studies to help resolve the many puzzles that remain. We hope that the gathering interest in the area revealed in Figure 1.1 will prove sustained. Notes 1 Figure 1.1 shows the number of articles identified in the Social Science Citation Index using the keywords ‘capital investment, fixed investment, capital expenditure’. Although this fails to capture some important contributions, it excludes articles purely concerned with financial investment. 2 See Chick (1998) for a discussion of formalism and ‘mock’ precision: Chick, V. (1998) ‘On knowing one’s place: the role of formalism in economics’. 3 The overseas operations of UK companies have tended to generate profits not only larger than the domestic profit rate but significantly higher than other countries overseas return on foreign assets, especially Europe and Japan. It is quite possible that this has led to a submissive response by UK based business in the manner described for US industries in Scherer (1991). 4 Chirinko (1987) derives the relationship between marginal and average q under non-constant returns and imperfect competition. 5 Investment and q may also be jointly determined as in real option models (see below) or where merger activity is affecting both. 6 Quite apart from the convexity assumption, it is unclear why cost of adjustment involved in building or modernising a plant should always depend on the firm’s existing capital stock. Nor is it clear that a rational manager would always decide to overbuild in order to reduce future adjustment costs, though that is implicit in the models. 7 With non-convexities alone, it is possible to reformulate the q-model to incorporate irreversibility (Abel and Eberly 1994; Eberly 1997). 8 The cost-minimising alternative specification where output is determined exogenously by demand suggests that capital, output and the ratio of the price of capital to output price should be cointegrated; this also appears not to be the case. 9 Another possible constraint on adjustment is an overhang of excess capital which might constrain new capital formation (Whittaker 1998). It is not clear even in theory how excess capital will be eliminated (Ghemawat and Nalebuff 1985; Grant 1990; Pindyck 1991). It seems unlikely, however, that this situation characterises more than a few industries in the UK given the depth of recent recessions, but it may apply to some extent in Europe. 10 More complex models have wealth effects which depend on savings parameters (Skott 1988). 11 There is a large literature on the so-called Feldstein-Horioka paradox—a close and apparently robust statistical correlation between rates of domestic saving and rates of domestic investment (Feldstein and Horioka 1980; for a review see Devereux 1996). One approach to the puzzle is simply to see the relationship as the converse of a tendency for the balance of payments on current account (i.e. the amount that needs to be borrowed on capital account) to be stable in the medium to long term. One reason for this might be an exchange rate risk that tends to increase as capital imports (or exports) mount; an alternative might be endogenous government policy responses. In any event, this approach suggests that the relationship is symptomatic of a binding balance of payments constraint, driven by Harrod’s foreign trade multiplier (for a discussion see Fagerberg 1996). An implication of this is that new investment if it is to be successful in stimulating growth should ultimately relieve this balance of payments constraint, though that does not imply that investment must be directly concentrated in the tradable sector.

RECENT ISSUES IN INVESTMENT THEORY

13

12 The bias arises because of Jensen’s inequality. 13 Where there are sharply decreasing economies of scale it is possible that the rise in FK could be reversed: the exact condition is that σ>1/(1–η) where σ is the elasticity of substitution and η is the economies of scale parameter. However this is considered unlikely. 14 Note, however, this model does not consider the case of capital being unused, i.e. it makes no distinction between the marginal return on capital in use and the marginal return return on capital purchased. (The K in the equation for πK should relate to capital purchased, whereas the K in the equation for πL relates to capital in use.) 15 Where variable cost w is included replace p by p-w. For a flexible-price version of the Newsboy model see Driver et al. (1996). 16 In some models, information can only be obtained by commitment. This may be the case with R&D investments. In such cases the use of option values suggests that investments should be judged at a lower hurdle rate than the user cost because an option is being acquired. A similar reversal can take place when there are long delivery lags. Here higher uncertainty may tend to speed up uncertainty (Bar-Ilan and Strange 1996). 17 Some industries may have coordination mechanisms to facilitate high capacity utilisation (Gilbert and Lieberman 1987). In this case the timing of expansion depends on a conjecture in respect of new entry. 18 The acquisition of options is treated more briefly. ‘How do firms obtain investment opportunities?…they arise from a firm’s managerial resources, technological knowledge, reputation, market position, and possible scale, all of which may have been built up over time’ (Pindyck 1991, p. 1111). Some have argued that the dependence of investment on these previously acquired options is a main issue in investment theory (Gordon 1994). 19 This clarification explains why there has traditionally been a split between the finance function and the strategy function in the planning of capital investment— traditionally financial analysis (NPV) did not include option values (Kaplan 1986). 20 Guiso and Parigi (1996) look at the effect of irreversibility using Italian firm-level data on investment intentions, output expectations and measures of subjective uncertainty as well as three separately defined measures of irreversibility including direct perception, cyclicality of the industry and extent of second-hand sales. In panel data estimation they show that the more irreversible the investment context, the greater the negative effect of uncertainty. See also Driver (1999). 21 The failure of the relative price terms in the equations is not unusual. It is often argued that the cost of capital is an endogenous cyclical variable which makes it difficult to observe causal effects. In a study of Irish investment, where the cost of capital varies considerably due to variation in capital grants and where the interest rate may be less correlated with the domestic cycle, a strong cost of capital effect is observed in a similar cointegrating relationship (Driver and NESC Secretariat 1998). 22 The finding that forward looking behaviour matters is also supported by the finding that direct survey-based expectations add to the explanatory power of the equations. This finding mirrors a similar result for Australian manufacturing investment (Driver and Dowrick 1997). 23 Whitaker (1998) also considers other possible reasons for the slow recovery of investment in the post-1992 period, including measurement error in real investment, the role of capacity utilisation and financial variables. 24 Guiso and Parigi find evidence that the negative sensitivity of investment to uncertainty increases with market power. However, they proxy demand elasticity by the price margin which may be correlated with capital intensity. 25 It may also be noted that the data employed is industry data. This raises complications because as Pindyck (1993) has noted in response to Caballero’s argument, industry-level uncertainty can have a negative effect on investment even in perfect competition. This arises because industry price uncertainty is accompanied by uncertainty over entry which biases down the upside.

References A.B.Abel (1980) ‘Empirical investment equations: an integrative framework’, 12th Carnegie Rochester Conference on Public Policy, Vol. 12, pp. 39–91. A.B.Abel (1983) ‘Optimal investment under uncertainty’, American Economic Review, March, 73, 1, 228–33. A.B.Abel and O.J.Blanchard (1983) ‘Investment and sales’, working paper No. 2050, National Bureau of Economic Research. A.B.Abel and O.J.Blanchard (1986) ‘The present value of profits and cyclical movements in investment’, Econometrica, 54, 2, 249–73. A.B.Abel and J.Eberly (1994) ‘A unified model of investment under uncertainty’, American Economic Review, 84, 5, 1369–84. A.B.Abel, A.K.Dixit, J.C.Eberly and R.S.Pindyck (1996) ‘Options, the value of capital and investment’, Quarterly Journal of Economics, 111, 3, 753–77. K.Aiginger (1987) Production and Decision Theory under Uncertainty, Oxford: Basil Blackwell. P.Arestis and M.Sawyer (1998) ‘The macroeconomics of New Labour’, mimeo, University of East London. A.Bar-Ilan and W.C.Strange (1996) ‘Investment lags’, American Economic Review, 86, 3, 610–22. C.R.Bean (1989) ‘Capital shortage and permanent unemployment’, Economic Policy, April, 11–54. C.R.Bean (1991) ‘An econometric model of manufacturing investment in the UK’, Economic Journal, March, 91–121. O.Blanchard (1997) ‘The medium run’, Brookings Papers on Economic Activity, 2, 89–158. O.Blanchard, C.Rhee and L.Summers (1993) ‘The stockmarket, profits and investment’, Quarterly Journal of Economics, 108, 1, 115–36. M.Blomstrom, R.E.Lipsey and M.Zejan (1996) ‘Is fixed investment the key to economic growth?’, Quarterly Journal of Economics, 111, 3, 269–76. R.Blundell, S.Bond and C.Meghir (1992) ‘Econometric models of company investment’, in L.Mayyas and P.Severstre (eds), The Econometrics of Panel Data: handbook of theory and applications, Dordrecht: Kluwer Academic Publishers.

14

CIARAN DRIVER AND PAUL TEMPLE

G.Bombach (1985) ‘Post-war economic growth revisited’, Professor Dr.de Vries Lectures in Economics; theory; institutions; policy, Vol. 6, Amsterdam: North Holland. S.Bond and T.Jenkinson, (1996) ‘The assessment: investment performance and policy’, Oxford Review of Economic Policy, 12, 2, 1–29. B.Bosworth (1981) ‘Comment’ on L.H.Summers, ‘Taxation and Corporate Investment: a q-theory approach’, Brookings Papers on Economic Activity, 61, 128–32. R.J.Caballero (1991) ‘On the sign of the investment-uncertainty relationship’, American Economic Review, March, 81, 279–88. R.J.Caballero, E.M.Engel and J.C.Haltiwanger (1995) ‘Plant-level adjustment and aggregate investment dynamics’, Brookings Papers on Economic Activity, 2, 1–54. M.Catinat, R.Cawley, F.Ilzkovitz, A.Italianer and M.Mons (1987) ‘The determinants of investment’, European Economy, 31. V.Chick (1998) ‘On knowing one’s place: the role of formalism in economics’, Economic Joumal, 108, 451, 1859–69. R.S.Chirinko (1987) ‘Intertemporal constraints, shadow prices and financial asset values’, NBER working paper No. 2247, May. R.S.Chirinko (1993) ‘Business fixed investment spending modeling strategies, empirical results, and policy implications’, Journal of Economic Literature, 31, 1875–1911. R.S.Chirinko (1994) ‘On the Keynesian investment function and the investment function(s) of Keynes’, in P.Davidson (ed.), Can the Free Market Pick Winners?: what determines investment, Armonk, NY: M.E.Sharpe. P.Clark (1979) ‘Investment in the 1970s: theory, performance and prediction’, Brookings Papers on Economic Activity, 1, 73–113. P.Clark (1993) ‘Tax incentives and equipment investment’, Brookings Papers on Economic Activity, 1, 317–47. J.Conlisk (1996) ‘Why bounded rationality?’, Journal of Economic Literature, 34, 2, 669–700. K.Cowling and R.Sugden (1996) ‘Capacity, transnationals and industrial strategy’, in J.Michie and J.Grieve Smith (eds), Creating Industrial Capacity, Oxford: Oxford University Press. J.Crotty (1996) ‘Is new Keynesian investment theory really “Keynesian”? Reflections on Fazzari and Variato, Journal of Post Keynesian Economics, 18, 3, 333–57. J.B.DeLong and L.H.Summers (1992) ‘Equipment investment and economic growth: how strong is the nexus? ’, Brookings Papers on Economic Activity, 2, 157–99. M.P.Devereux (1996) ‘Investment, saving, and taxation in an open economy’, Oxford Review of Economic Policy, 12, 2, 90–108. M.P.Devereux, and S.Schianterelli (1990) ‘Investment, financial factors and cashflow: UK panel data’, in G.Hubbard (ed.), Asymmetric Information, Corporate Finance and Investment, Chicago: University of Chicago Press. A.Dixit and R.Pindyck (1994) Investment Under Uncertainty, Princeton: Princeton University Press. S.Dowrick (1994) ‘Fiscal policy and investment: the new supply side economics’, mimeo, Australian National University, Canberra. C.Driver (1992) ‘The origin of predictable behaviour: comment’, American Economic Review, 82, 1, 368. C.Driver (1999) ‘Contrasting effects of risk on investment in two sectors: evidence from Ireland on real options’, Economic Letters, forthcoming. C.Driver and D.Moreton (1991) ‘The influence of uncertainty on UK manufacturing investment’, The Economic Journal, 101, 409, 1452–59. C.Driver and D.Moreton (1992) Investment, Expectations and Uncertainty, Oxford: Basil Blackwell. C.Driver and S.Dowrick (1997) ‘Investment intentions and investment realisations in Australian manufacturing’, Australian Economic Papers, 36, 68, 90–105. C.Driver and NESC Secretariat (1998) ‘Econometric analysis of investment and risk’, Chapter 4 of NESC Private Sector Investment in Ireland, pp. 49–56, Dublin: National Economicand Social Council. C.Driver, S.Abubacker and G.Argiris (1996) ‘Capacity choice under monopoly, flexible price and demand uncertainty’, Southern Economic Journal, 63, 2, 526–32. J.Eberly, (1997) ‘International evidence on investment and fundamentals’, European Economic Review, 41, 6, 1055–78. R.Eisner (1994) ‘National savings and budget deficits’, Review of Economics and Statistics, February, 181–5. J.Fagerberg (1996) ‘Technology and competitiveness’, Oxford Review of Economic Policy, 12, 39–51. S.M.Fazzari and A.M.Variato (1994) ‘Asymmetric information and Keynesian theories of investment’, Journal of Post Keynesian Economics, 16, 3, 351–70. S.M.Fazzari and A.M.Variato (1996) ‘Varieties of Keynesian investment theories: further reflections’, Journal of Post Keynesian Economics, 18, 3, 359–68. S.M.Fazzari, R.G.Hubbard and B.C.Petersen (1988) ‘Financing constraints and corporate investment’, Brookings Papers on Economic Activity, 1, 141–95. M.Feldstein (1994) ‘Tax policy and international capital flows’, NBER working paper No. 4851. M.Feldstein and C.Horioka (1980) ‘Domestic saving and international capital flows’, Economic Journal, 90, 314–29. R.Ford and P.Poret (1991) ‘Business investment: recent performance and some implications for policy’, OECD Economic Studies, 16, Spring, 79–131. P.A.Geroski and S.Machin (1994) ‘Innovation, profitability, and growth over the business cycle’, in K.Aiginger and J.Finsinger (eds), Applied Industrial Organisation, Dordrecht: Kluwer Academic Publishers. P.Ghemawat and B.Nalebuff (1985) ‘The devolution of declining industries’, The Quarterly Journal of Economics, 105, 1, 167–86. V.Ghosal and P.Loungani (1996) ‘Product market competition and the impact of price uncertainty on investment: some evidence from US manufacturing industries’, Journal of Industrial Economics, 44, 217–28. R.J.Gilbert and M.Lieberman (1987) ‘Investment and coordination in oligopolistic industries’, R Journal of Economics, 18, 17–33.

RECENT ISSUES IN INVESTMENT THEORY

15

S.Gilchrist, and C.P.Himmelberg (1994) ‘Evidence on the role of cash flow for Investment’, mimeo, Board of Governors, Federal Reserve System. M.J.Gordon (1994) ‘The neoclassical and a post-Keynesian theory of investment’, in P.Davidson (ed.), Can the Free Market Pick Winners?: what determines investment, Armonk, NY: ME Sharpe. C.W.J.Granger (1994) ‘Forecasting in economics’, in A.S.Weigend and N.A. Gershenfeld (eds), Time Series Prediction Wokingham: Addison-Wesley. R.M.Grant (1990) ‘Exit and rationalisation in the British cutlery industry 1974–1984’, in C.Baden-Fuller (ed.), Managing Excess Capacity, Oxford: Basil Blackwell. L.Guiso and G.Parigi (1996) ‘Investment and demand uncertainty’, discussion paper No. 1497, CEPR London. D.S.Hamermesh and G.A.Pfann (1996) ‘Adjustment costs in factor demand’, Journal of Economic Literature, 34, September, 1264–92. R.Hartman (1972) ‘The effects of price and cost uncertainty on investment’, Journal of Economic Theory, October, 5, 258–66. R.Heiner (1983) ‘The origin of predictable behavior’, American Economic Review, September, 560–95. T.Hoshi, A.K.Kashyap, and D.Scharfstein (1990) ‘Corporate structure and investment: evidence from Japanese panel data’, Quarterly Joumal of Economics, 106, 1, 33–60. X.Hu, and F.Schianterelli (1994) ‘Investment and financing constraints: a switching regression using US firm panel data’, Boston College working paper. R.G.Hubbard (1998) ‘Capital market imperfections and investment’, Journal of Economic Literature, 36, March, 193–225. N.Jenkinson (1996) ‘Savings, investment and real interest rates’, Bank of England Quarterly Bulletin, February, 51–60. S.N.Junankar (1988) ‘The CBI Industrial Trends Survey: 30 years of interpretation and analysis’, CBI Economic Situation Report. S.N.Junankar (1990) ‘How do companies respond to the Industrial Trends Survey?’, CBI Economic Situation Report. R.S.Kaplan (1986) ‘Must CIM be justified by faith alone?’, Harvard Business Review, March-April 87–95. S.N.Kaplan, and L.Zingales, (1998) ‘Do investment-cash flow sensitivities provide useful measures of financing constraints’, Quarterly Journal of Economics, 113, 1, 169–215. M.Kitson and J. Michie (1996) ‘Britain's industrial performance since 1960: under-investment and relative decline, The Economic Journal, 106, 434, 196–212. M.Lieberman (1985) ‘Capacity utilisation in the chemical processing industries: theoretical models and empirical tests’, Research Paper No. 817, Stanford University Graduate School of Business. J.W.Lomax, (1990) ‘A model of ICC's dividend payments’, Bank of England Discussion Paper No. 52, London UK. L.J.Maccini (1987) ‘Adjustment costs’, in J.Eatwell, M.Millgate and P.Newman, The New Palgrave: a dictionary of economics, Vol. 1, 23–5. S.A.Marglin and A.Bhaduri (1989) ‘Profit squeeze and Keynesian theory’, in S.A.Marglin and J.Schor (eds), The Golden Age of Capital Accumulation, Oxford, Clarendon. D.Mayes and G.Young (1993) ‘Industrial investment and economic policy’, discussion paper No. 56, National Institute of Economicand Social Research, London. S.J.Nickell (1978) The Investment Decisions of Firms, Cambridge: Cambridge University Press. S.J.Nickell and D.Nicolitsas (1996) ‘Does innovation encourage investment in fixed capital?’, Institute of Economics and Statistics, Oxford. O.A.Nilsen and F.Schiantarelli (1996) ‘Zeros and lumps in investment’, mimeo, Boston College and World Bank. E.Penrose (1959) The Theory of the Growth of the Firm, Oxford: Basil Blackwell. R.S.Pindyck (1988) ‘Irreversibility, investment and the value of the firm’, American Economic Review, December, 969–85. R.S.Pindyck (1991) ‘Irreversibility, uncertainty and investment’, Journal of Economic Literature, 29, 1110–48. R.S.Pindyck (1993) ‘A note on competitive investment under uncertainty’, American Economic Review, March, 273–7. R.S.Pindyck and A.Solimano (1993) ‘Economic instability and aggregate investment’, in O.J.Blanchard and S.Fischer (eds), NBER Macroeconomics Annual 1993, Cambridge, MA: MIT Press. S.Price (1995) ‘Aggregate uncertainty, capacity utilisation and UK manufacturing investment’, Applied Economics, 27, 147–54. P.Robinson (1996) ‘Skill shortages and full employment’ in J.Michie and J.Grieve Smith, Creating Industrial Capacity, Oxford: Oxford University Press. D.Rodrik (1994) ‘King Kong meets Godzilla: the World Bank and the East Asian miracle’, CEPR discussion paper No. 944, London, April. S.Roper et al. (1996) Product Innovation and Development in UK, German and Irish Manufacturing, Northern Ireland Economic Research Centre, University of Strathclyde, and IFO Institute, Munich. M.Rothschild (1971) ‘On the cost of adjustment’, Quarterly Journal of Economics, 85, 4, 605–22. M.Rothschild and J.Stiglitz (1971) ‘Increasing risk: its economic consequences’, Journal of Economic Theory, 3, 1, 66–82. R.E.Rowthorn (1995) ‘Capital formation and unemployment’, Oxford Review of Economic Policy, 1,1, 26–39. R.E.Rowthorn (1996) ‘Unemployment, wage bargaining and capital-labour substitution’, mimeo, University of Cambridge. F.M.Scherer (1986) ‘Time-cost trade off in uncertain empirical research projects’, Chapter 4 of Innovation and Growth, Cambridge, MA: MIT Press. F.M.Scherer (1991) ‘International R&D races: theory and evidence’, in L.G. Matteson and B.Stymne (eds), Corporate Industry Strategies for Europe, Barking: Elsevier Science Publishers. F.Schianterelli (1996) ‘Financial constraints and investment: methodological issues and international evidence’, Oxford Review of Economic Policy, 12, 2, 70-87.

16

CIARAN DRIVER AND PAUL TEMPLE

C.L.Schultze (1987) ‘Saving, investment and profitability in Europe’, in R.Lawrence and C.Schultze (eds), Barriers to European Growth: a transatlantic view, Washington: Brookings, 509–39. A.Sentance and G.Urga (1997) ‘Profitability, structural change and business investment in the UK’, London Business School, October. D.J.Sharp (1991) ‘Uncovering the hidden value in high-risk investments’, Sloan Management Review, Summer, 69–74. P.Skott (1988) ‘Finance, saving and accumulation’, Cambridge Journal of Economics, 12, 339–54. P.Stoneman, ed. (1995) Handbook of the Economics of Innovation and Technological Change, Oxford: Basil Blackwell. J.B.Taylor (1982) ‘The Swedish investment funds system as a stabilisation policy rule’, Brookings Papers on Economic Activity, 1, 57–105. W.Vickery (1993) ‘Today’s tasks for economists’, American Economic Review, 83, 1, 1–11. A.Wardlow (1994) ‘Investment appraisal criteria and the impact of low inflation’, Bank of England Quarterly Bulletin, August, 250–4. T.M.Whited (1992) ‘Debt liquidity constraints and corporate investment: evidence from panel data’, Journal of Finance, 47, 1425–6. S.Whittaker (1998) ‘Investment in this recovery: an assessment’, Bank of England Quarterly Bulletin, February, 38–47. G.Young (1994) ‘The influence of foreign factor prices and industrial taxation on fixed investment in the UK’, National Institute Discussion Paper, 66.

2 Finance, profitability and investment in manufacturing Brian Henry, Andrew Sentance and Gioυanni Urga1

Introduction The sluggishness of capital investment is one of the more puzzling features of the UK economic recovery from the 1990s recession. Investment is normally much more volatile than the economy as a whole, increasing strongly in recoveries and falling sharply in recessions. Yet between 1992 and 1996, total fixed capital formation increased by just 4.4 per cent—an average rate of increase of 1.1 per cent. In the same period, GDP growth averaged over 2.5 per cent. Investment actually fell in 1995 and rose by just over 1 per cent in 1996 according to the current official estimates. This situation is very different from the early years of the 1980s recovery. In the first three years of that recovery alone— between the first half of 1981 and the first half of 1984—fixed investment rose by 20 per cent, accounting for over a third of the 8.5 per cent rise in GDP seen over that period. The 1980s appear unusual; as Figure 2.1 shows, the current investment recovery is more in line with the experience of the 1970s. Then, however, investment was held back by weak profitability and industrial relations problems in British industry. The industrial relations scene has since been transformed. But the weakness of investment does not appear to be due to any shortage of funds for business to invest, as company profits have risen by over 50 per cent. In manufacturing, a similar story can be told except that this recovery now appears to be the outlier. Figure 2.2 shows that since 1992 investment first grew at roughly the same rate as during the 1980s recovery but, from mid-1995, fell away. So far, the 1990s recovery in manufacturing investment has not matched that of the two previous episodes. Possible explanations for this unusual behaviour are investigated in this chapter. To evaluate possible explanations of sluggish investment, we concentrate on the case of manufacturing. The approach is to estimate models of manufacturing investment, which take into account demand effects and financial constraints in different ways. In short, we seek to establish an econometric specification which appears to fit the data well, is consistent with conventional economic theories of investment behaviour, and then test its implications for investment during the recovery.

Figure 2.1 Total UK fixed investment (volume index: trough of recession=100).

The plan of the chapter is to review the relevant theory next. We go on to describe the estimation techniques used. We then present data analysis and econometric results before offering conclusions.

18

HENRY, SENTANCE AND URGA

Figure 2.2 Manufacturing investment recoveries (trough of recession 1992Q1=100).

Theory The background Both the influence of financial factors and the effects of expectations of future profitability and demand upon investment have been continuing concerns in the empirical literature on investment determination. Concerning the success of this work, the view expressed by two of the more distinguished economists in the field is salutary. The explanation of aggregate and sectoral investment spending has been one of the less successful endeavours in empirical economies. The problem is not just that these models have been unable to explain or predict more than a small proportion of the movements of investment. In addition, constructed quantities that in theory should have strong explanatory power—such as Tobin’s q, or various measures of the cost of capital—in practice do not. (Dixit and Pindyck 1994, p. 419) As the quote from Dixit and Pindyck makes clear, the present state of research on investment, and on the empirical role of financial effects on investment in particular, is highly unsatisfactory. There is little evidence which supports neo-classical or ‘q’ variants of cost of capital and/or aggregate valuation effects on investment. (For similar conclusions see also Chirinko 1993.) Hence, one of the emphases which this chapter has is in how financial effects may work, including the nature of financial constraints (meaning broadly where firms may not be able to borrow at prevailing lending rates). It needs to be made clear at the outset that there is no attempt to rigorously derive an encompassing framework within which the alternative hypothesis— of financial effects or otherwise—can be tested against each other. The models are too diverse, and underlying theoretical assumptions vary too much for that. Among the most obvious of these differences are assumptions about imperfections in product and financial markets, the detail with which the tax structure is modelled, the alternative hypothesis about expectations formation, and finally different assumptions about technology. There are also differences which arise in how firms’ behaviour is analysed. In most models of investment, the underlying model is that of optimal factor demands, but in others a bargaining framework is used, determining both investment and wages (e.g. Denny and Nickell 1992). Indeed, given this diversity, some of these issues are not even reviewed in this chapter. We do not for example investigate the effect of different assumptions about technology, or the effects of joint modelling of factor demands. Nor, in spite of quoting from Dixit and Pindyck does uncertainty figure in what we do from here on. A recent paper providing empirical results for capital accumulation in a joint factor demand framework based on a generalised cost function is provided in Nixon and Urga (1997). For an approach to the issue of technology, see Young (1996) who discusses shortcomings in vintage production function models and contrasts these with more orthodox technologies. Econometric studies with some support for effects from uncertainty are reviewed in Driver et al. (1996). Although we do not intend to provide a review of the many theories, and different empirical results in the investment literature, none the less it is our intention to provide an assessment of the empirical merits of a selection of models which are presently in existence. For this we need to describe the alternatives we aim to cover in a unified framework. We proceed to this now, starting with a basic model, and amending this to incorporate what, in our judgement at least, are the most important additions to this basic model.

FINANCE, PROFITABILITY AND INVESTMENT

19

The basic model We start with a few basic assumptions about the behaviour of the firm. In the models discussed here and later, the assumption underlying the analysis is that of a representative firm, which is intertemporally optimising subject to the standard assumptions of quadratic adjustment cost. We do this because it is then convenient to marshall the different models of investment which follow based on differing assumptions about product market competition, financial effects, and bargaining behaviour, in terms of their general implications for the future expected rental stream accruing to a unit increase in capital (net of adjustment costs). As will be seen, different models can be described in terms of what they say about this. So to start with a basic competitive model (the extensions produced by assumptions of imperfect competition, financing constraints, and expectations formation are described below), the firm's expected cashflow in period t(CF)t can be expressed. (1) where the output price is treated as numeraire, E refers to expectations formed in period t, F is the production function dependent on labour and capital (L and K respectively), and G is an adjustment cost function, arising because capital is quasi-fixed, hence we assume G(.) depends on K and investment (I). Lastly, W and PI are the real prices of labour and investment respectively (we make the assumption of perfect factor markets for the present). Constrained maximising proceeds along familiar lines. That is, the firm maximises (2) where δ is the exogenous rate of decay in the capital stock, and (1+r) is the discount factor. To obtain an equation which may be estimated, the optimal solution for (2) needs to incorporate a specific functional form for G(.), the adjustment cost function. Most often this assumes costs are quadratic in gross investment, linear homogenous in I and K, and affected by a technology shock (τ) (see Summers 1981). Hence, (3) where α is the adjustment cost parameter (the larger is α, the more slowly does investment respond), and time subscripts are suppressed for convenience. Assuming goods, factor and financial markets are perfect, substituting (3) into (1), optimising (2) w.r.t. K and solving subject to a transversality condition gives the investment equation (4) where ^ is defined as . In turn, λ is the shadow value of a change in the return to capital—the marginal revenue product of capital net of adjustment costs—and ρ equals (1−δ)/(1+r). Hence (4) shows that investment depends on the discounted sum of ‘spot’ marginal revenue products, and investment (relative to the capital stock) increases when the present value of benefits exceeds the present costs of investment, at a rate dictated by the adjustment cost parameter (α). This basic relationship is fundamental, and in the models which follow, alternative assumptions regarding product market competition, and financial constraints are summarised in terms of their implications for this net rental flow. Before proceeding to consider alternatives to (4) note that it can be cast as a q model, which typically sets the investment ratio as a function of the ratio of the financial value of the firm to the replacement value of its capital stock. According to this interpretation, Equation (4) is (5) where qA is average q, equal in turn to the ratio of the financial value of the firm to capital stock at replacement cost ((V/P1K) where V is the financial value of the firm). Although there are many problems in (5), one of its attractions is that it gives a straightforward way of relating the unobservable forward looking variable E(^) in (4) to observables, which in principle incorporate expectations about the future. None the less, this is not the only way to deal with expectations, and below other ways to include expectations in these equations, including direct measures of expectations, are illustrated. Before that, we describe how this simplified model can be extended, and it will be these extensions which will form the basis of the empirical models tested later. These extensions will deal with the introduction of imperfectly competitive product markets, the role of internal financing, alternative ways to include financial constraints, effects on investment arising from the presence of unions with bargaining power, and lastly the possible effects of expectations. Extensions to the basic model To confine the alternatives considered here to manageable proportions, we will make a few assumptions which will be common to each. Apart from the assumptions of quadratic adjustment costs, the models estimated later will be based on the

20

HENRY, SENTANCE AND URGA

assumption of monopolistic competition. This assumption is one of the possible rationales for incorporating output in the investment equation, something for which there is empirical support. Thus we suppose that technology is given by the linear homogeneous production function, depending upon capital (K) and labour (L), (6) and we further suppose that there are internal adjustment costs G (I, K) which are assumed to be quadratic. We assume the representative firm also faces a demand function (7) where is the general price level, η the elasticity of demand and a a shift parameter. Hence, the implications of imperfect competition means that the static marginal conditions for capital are amended from, e.g. pFi=pi, where p is the product price, and pi the factor price of the ith input, to be p(1−η−1)Fi=pi, where η is the elasticity of product demand. In this model, it is via this last term that output can be introduced. Below we comment on a further possible justification, arising from an expected output constraint (see below) Allowing for retentions and dividends Building further on the baseline model, to allow for retentions and dividends, as well as providing a rudimentary account of the company tax system, suppose the firm chooses K, L, Vn and I to maximise wealth of its shareholders, i.e. (8) Vn

where D are dividends, the value of new share issues, and Bj the discount factor. The γ term measures the relative tax advantage of dividends versus retained earnings (see Schiantarelli and Georgoutsos 1990). The first order condition for maximising Vt w.r.t. I, subject to the usual non negativity constraints on dividends, shares, and the capital accumulation equation Kt=(1−γ) Kt−1+It, where γ is the rate of depreciation, is that the cost of a marginal unit of capital must equal its shadow value λk (the non negative multiplier on the capital accumulation constraint). This, however, now has different implications as compared with the baseline model, (4). Now, the marginal shadow value of capital λk becomes (suppressing future time subscripts for simplicity) (9) where β is the discount factor. Other terms in equation (9) are as follows. • A is the discounted present value (DPV) of tax savings due to investment allowances. • H is the DPV of the cash flows which are due on the firm's debt, including all interest payments. • C is the DPV of the decrease in revenues produced by an additional unit of capital, since—given (7)—product price is lowered to sell additional output. The C terms capture the capitalised value of this effect. This way of introducing output into the investment decision rule may not be particularly appealing from an a priori viewpoint. Assuming firms are output constrained on the other hand, implies a cost minimising model, where output and relative factor prices determine investment. A more promising approach is described in Precious (1987). Here the forward looking firm may experience regimes where output (sales) is constrained. Even if the firm is currently unconstrained, it may none the less expect at a future date to encounter such a constraint. Expected output will therefore enter its investment decision; but the investment equation is not just the familiar cost minimising equation dependent on output and relative factor prices only. • γ is a parameter which determines whether the firm finances itself through new issues or retentions at the margin. • By simplifying somewhat the equations (8) and (9) above, the investment equation in this case depends upon average q (qA), output (or more precisely, expected output) and a set of other variables. (10) Schiantarelli and Georgoutsos (1990). Xt includes any variables which drive a wedge between investment, output and qA. In the present example this is current cash flow relative to debt. But there are other contenders for inclusion in X, and in what follows, we group these into financial and real variables. Financial effects on investment Equation (10) includes a further financial variable in cash flow. But other variables have been proposed to proxy financial imperfections. The approach we take to this is a pragmatic one, starting from the proposition that with capital market imperfections, internal and external sources of finance are not perfect substitutes. This is not the place to review the relevant arguments in detail. Suffice to note that general arguments suggest that it may be an optimal strategy for banks to set interest

FINANCE, PROFITABILITY AND INVESTMENT

21

rates where an excess demand for credit exists due to adverse selection since banks cannot fully monitor the net worth of firms (Stiglitz and Weiss 1992). Firms may also face a rising supply price of debt depending upon, e.g. firm size and risk category. Also there may be agency costs in issuing and maintaining debt. Finally, on the debt side, there may be moral hazard involved, where with a high debt/ income ratio, the incentives of the firm (managers acting on behalf of shareholders) and creditors diverge, so managers have an incentive to undertake higher risk and return projects. In similar vein there are related imperfections in raising funds via equity; adverse selection and transactions cost (especially for smaller firms) being problems. Financial effects which enter Xt in (10) are: • Cash flow. This assumes that firms may either use retained earnings or new issues (but in which model the decision is exogenous), so the investment equation includes a cash flow term. where B=the market value of debt, ζ the corporate tax rate, and i the interest rate on debentures. • Capital gearing. The investment model also can be extended to include debt. In this case, agency costs are assumed as well as the usual interest costs on debt. These lead to an additional term in capital gearing, defined as the stock of debt as a proportion of the capital base (see Cuthbertson and Gasparro 1995). • Liquidity. A further argument is that firms may face a rising supply price of capital when borrowing and the effects of this may be proxied by liquidity (defined as liquid assets to total liabilities) (Kelly and Owen 1985). Real factors in investment Here, the main additional elements in X are real wages and measures of union strength. This results in a ‘hold-up’ model, where part of the rental stream accruing to the new investment is captured by unions in the form of higher real wages. Rather than take the firm’s factor demand decision, this alternative is a two-stage solution, where the firm is party to a conventional Nash wage bargain, the solutions to which determines wages, based on which, the firm then sets prices, investment and employment to maximise profit (see Denny and Nickell 1992). In summary, the solution to the wage bargain takes the form where is the outsider wage. We suppose (∂ ln w)/(∂ ln k)=γ(UN) , where UN is an index of union power, (γ1>0), that employment is set to maximise profit, and that prices can then be inferred from (7). The expected profit stream is then (11) where ε=1−1/η, and where the expected value in (11) depends on the solution (W, P, L, Y) from the wage bargain, and profit maximising above. The marginal benefit of extra investment is then the DPV of ∂П/∂k over the planning horizon of firms, which is equated to cost (p∂c/∂I). The model then yields an investment equation conditioned upon aggregate demand (output), and the real wage as well as indices of union power (proxied by UN here), i.e. (12) In this model, and I3>0 are expected signs. Explicit expectations In the above models expected values of the determinants of investment enter the dynamic investment equation. Here, we note how this can be developed into a dynamic equation for estimation, using equation (10) as an example. As quadratic adjustment costs are assumed, write the Euler equation above as, in unrestricted form (13)

22

HENRY, SENTANCE AND URGA

Upon factorising in conventional fashion (13) may be written (14) , Xt and (Y/K)t and λ is the stable root of equation where A is the vector of variables determining I/K. In the present case (13). Equation (14) now has an expected forward convolution of the variables determining the investment capital ratio. As is well known, the expectational terms in (14) can be introduced in several ways. In the empirical section we will use two: model based which we refer to as rational expectations hypothesis (REH) expectations, and, where future output is concerned, survey based (CBI) measures. Later, we describe how both of these may be estimated. Summary of alternative specifications To sum up, in what follows we consider dynamic models of investment behaviour based on the standard quadratic adjustment cost framework, where firms have informed, forward looking expectations. The variables included in the investment models purposely reflect a wide range of alternatives which have figured in the literature. On financial influences, we consider • Cost of capital. • qA variables, including proxy variable such as profitability (Sentance and Urga 1995). • Variables proxying financial constraints. These include undistributed profits, and liquidity (Woods 1995), and capital gearing (Cuthbertson and Gasparro 1995). • Additional variables are suggested by enlarging the model to include bargaining behaviour (Denny and Nickell 1992). This extension introduces indices of union power, the real wage and real price of investment goods as possible additional determinants of investment. Finally, as expectations play a central part in the empirical work, two broad approaches are employed: rational or informed expectations mechanisms, and direct survey based measures. Estimation methods Before moving to describe the estimation results, we briefly note some of the technical matters which crop up in estimating models where future expectations of demand, factor prices, etc. occur. Where we use direct measures of output expectations based on CBI surveys of business expectations the main problem which arises is the conversion of the survey responses to a quantified series. This is an extensive topic, and we do not intend even to try to outline it here (Pesaran 1984 provides an introduction to this subject). Of the two most favoured methods—the regression and the probability distribution based methods, we have opted for the regression based approach. Pesaran (1984) discusses the rival merits of these methods, and recommends the regression method as preferable. The rest of this section then discusses some of the basic estimation problems which arise when applying a rational expectations assumption, and how these may be addressed. We also show how cointegration results can be captured with a forward looking model of this sort, using a transformation of the model first described in Callen, Hall and Henry (1990). Estimating the forward looking model under the REH The optimal dynamic equation for net investment (14) can be written (15) (ignoring expectational and other errors) where A is a linear function of the long run determinants of investment (i.e. the cost of capital, output, etc.), and R is the discount factor. As background to what is used in the next section, recall that two approaches to estimating rational expectations models have proved popular in the applied literature; the errors in variables model (EVM), and the extrapolative model. In the former, where the model in (15) is simplified to an unlagged, two period model for expositional purposes only, we would have (16) where is a weakly exogeneous expectations variable, and μt is white noise. Then to apply the EVM method, we note that by the REH, (17) i.e. the actual future variable is equal to its rationally expected value plus a random error (the ‘expectations’ error). Substituting (17) into (16) we get

FINANCE, PROFITABILITY AND INVESTMENT

23

(18) where εt=μt−δ1wt+1–δ2wt+2. Clearly, at+j (j=1,2) and εt are correlated. Furthermore, the composite error term is serially correlated, having moving average error. Methods normally applied to deal with these problems involve the use of instrumental variables (IV), giving consistent parameter estimates, together with a correction to the variance—covariance matrix to ensure valid inference (see Cuthbertson, Hall and Taylor 1992, for a short review). The alternative assumes that there is a model determining the expected variables, which may be estimated jointly with the structural model (15). Thus suppose this model for can At+j be represented by the VAR (19) 1 where δ (L) is a polynomial in the lag operator. Then by substituting (19) into (15) we may write the investment equation as (20) (21) A variety of routes are open when applying this, including using predictions from (19) as instruments in estimating (15). (So introducing similar estimation problems to those noted above), or in joint full information maximum likelihood (FIML) estimation of (15) and (19). Estimating (15) with extrapolated predictions from (19), will further need to apply the restrictions embodied in (21) for rationality to apply. In this way it can be seen that an unrestricted version of (20) would give a lagged model for optimal investment, but applying the restrictions in (21) gives the REH version. In principle, conditional tests of the application of the REH can be based on this. In the applications below, the techniques of using an IV version of the extrapolative model is employed, and we also propose tests of the role of forward looking expectations formation following on from our earlier remarks. Before moving on to that, we indicate how cointegration methods can be brought to bear on the estimation of dynamic models with forward looking terms. Cointegration and forward looking behaviour For this, we need to extend (15) to allow it to have longer lags. This is easily justified, and can arise from non-quadratic adjustment costs, or a respecification of the firms’ objective function to penalise deviations of growth in investment away from its planned path. Specifically if It follows an AR(2) process, then by supposing that the long run equilibrium value for investment is I*, the investment equation can be reparameterised into an Error Correction Model (ECM) form. (22) where λ2 is the stable root, and where the di are dependent upon λl, λ2 and R. I* now refers to the long run (cointegrating) value of investment, which depends on the long run cost of capital, long run profitability, etc. In other words, it corresponds to the long run version of the linear function A in equation (15). The ECM (17) incorporates the feedforward–feedback restrictions of the REH into this reparameterised model (see Callen, Hall and Henry 1990). Indeed, as in the case of extrapolated predictions, reviewed earlier, this enables tests of forward looking versus backward looking versions of the model to be performed, a point we take up below. Finally, as is evident from (22), the dynamic equation can be estimated by a two step procedure, even where there are rational expectations. The first step involves estimating an equation for the long run determinants of investment (I*). We use cointegration techniques for this. The second step introduces these estimates into (22). This enables us to concentrate in the first step upon evidence for cointegration in the alternative models reviewed here. Hence, in the next section, to estimate the model we use a Johansen procedure, for identifying long run, cointegrating relations for investment, before moving to dynamic equations, where these may be backward or forward looking. Data analysis and estimation results The plan of this section on empirical results is as follows. It aims at an evaluation of the investment model using data on the manufacturing sector. Although increasingly less important in determining aggregate investment behaviour, manufacturing is none the less important both in its own right, and because the availability of sector specific variables (such as the CBI data we

24

HENRY, SENTANCE AND URGA

use later) means a rich set of tests of expectations formation and financial effects can be made. First, we describe the time series properties of the data, before moving to cointegration tests. Time series and other preliminary analysis of the data Time series analysis of the data Univariate tests for the orders of integration for the variables used in this chapter are given in Table 2.1. Among these tests some give ambiguous, and in some cases, counter-intuitive results. The first four, the cost of capital (CC), the Tobin’s ‘q’ variable (Qa), output in manufacturing (YM) and expected output ( ) reject stationarity in levels. Of the next five, only capital gearing (CG) is unambiguously non-stationary. Investment in manufacturing (MI), has a suggestion of stationarity in the ADF(4). In what follows we treat this as I(1), given the ambiguity of these tests. Relative factor prices in manufacturing (FPM), and liquidity (L) are I(0) according to both DF and ADP tests. In the cointegration tests below, we have experimented with treating them as I(0) or I(1) in the cointegration vector: with little variation in the results. Finally union density (UN) is evidently not I(0), and although there is some evidence that it could be I(2), in what follows we treat it as I(1). The remaining three variables—the real wage (RW), capacity utilisation (CU) and profitability—are taken to be I(1). Capacity utilisation might be expected to be I(0), but for this sample appears not to be. Cointegration tests for manufacturing investment We provide first a set of results for range of specifications for the investment equation, using output as the scaling variable, with alternative cost variables and variables proxying financial constraints. The purpose of this part of the exercise is twofold. First, to assess the empirical grounds for Table 2.1 Orders of integration Levels DF

ADF (4)

Differences DF

ADF (4)

CC Qa YM

−2.58 −2.82 −10.2 −6.2 −1.3 −1.5 −11.58 −4.9 −1.38 −1.81 −10.0 −4.3 −1.58 −1.74 −7.2 −4.09 MI −2.20 −3.68 −12.0 −4.6 CG −1.09 −1.26 −10.4 −5.3 FPM −2.99 −3.56 −9.4 −5.87 L −3.01 −3.81 −9.7 −4.1 UN 2.6 0.17 −3.88 −2.66 RW −1.9 −1.4 −11.9 −3.9 CU −2.3 −2.9 −10.5 −4.16 Π −2.17 −1.9 −14.0 −5.1 Notes: Variables, except the qA, liquidity and cost of capital variables, are in logs. All tests are run with a deterministic time trend. The relevant critical values are −2.88 (DF) and −3.44 (ADF (4)).

financial influences on investment, as well as familiar cost of capital and q variables. Second, to derive a well specified levels equation which will be used in the next section to test between forward and backward formed expectations models. The strategy used here is to set up several alternative models, and use the cointegration results as a test of their basic features. We consider a ‘union’ model familiar from the hold-up literature, where we concentrate on the evidence for real wage and strength of unions effects. Next, we formulate a basic ‘neo-classical’ model based on the cost of capital. We also experiment with a cost minimising model together with direct liquidity effects. Next, we investigate the role of profitability in an investment model which some people have suggested as an empirical counterpart to a ‘q’ model. Finally, we consider a ‘q’ model, incorporating gearing in this as an additional explanatory variable. After the cointegration results we move to evaluating a hybrid model based on what appear to be the most promising of the results, and this version of the model is the one we use to explore how the recent behaviour of investment may be accounted for. The results of the cointegration tests on the alternative models are shown in Table 2.2.

FINANCE, PROFITABILITY AND INVESTMENT

25

Model 1 is the best result we were able to obtain for the union model. Output is restricted to unity in this model, although it actually fails this restriction (x2(1)=7.07). So the implied parameter restrictions accord with theory when, and only when, this restriction is applied, and we conclude that overall this model is not generally supported by the data. Model 2 is a pragmatic ‘neo-classical’ model, and again does not fare well. The cost of capital term is incorrectly signed. Moreover, the model does not cointegrate at the 5 per cent level. Table 2.2 Cointegration tests on alternative models (1) (2) (3) (4)

MI MI MI MI

YM

RW

UN

1.0 1.2 1.0 1.0

0.39

−0.01

CC

CU

0.05

−0.008

FPM

L

Π

LR

LT

20.1* 27.4* 37.3** 29.5**

40.9** 44.4 63.9** 45.7**

0.01 0.73 – – 0.42 a Q CG (5) MI 2.1 0.56 −0.46 18.1 37.2 Notes: Accepts hypothesis of cointegration at the 10 per cent (*) or the 5 per cent (**) level. All models were estimated using a VAR of length 2. Equations (3)–(5) included restricted deterministic time trends, as these gave the best version of the model. The remaining equations used restricted intercepts only. Equation (1) was estimated to the end of 1992 due to limited availability of data on unionisation. All other equations used data ending in 1996Q4. LR is the eigenvalue test, LT the trace test for the presence of cointegration.

Model 3 is based on the cost minimising model reported by Woods (ibid.). It has an important effect from liquidity (L) but the relative price term (FPM) is incorrectly signed. Further variants of this are discussed below. Model 4 is a model of the form which Sentance and Urga used in the business sector case (Sentance and Urga (1998)). The profitability parameter (Π) is correctly signed and the model cointegrates. The imposed unit coefficient on output is satisfactory (The Wald test gave x2(1)=2.6). Model 5 is the qA model coupled with gearing (G) model described in Cuthbertson and Gasparro (1995). Again this is an example of a model which does not appear to cointegrate, but the parameter values accord with prior theory. In the sample period used originally by the authors (1968Q1–1988Q4), the model does cointegrate. But overall, the conclusion from our limited exercise is that it appears that the model is not robust when the original data set is extended. A hybrid model Although this exercise has been quite informal, it is instructive, in that some models appear to be ruled out. Models (1)–(2), and model (5) appear not to have cointegration and/or broad confirmation of their a priori assumptions. So next, more formal testing is conducted between models (3) and (4). These both introduce additional financial influences on investment: liquidity and profitability respectively. Can we choose between these two variables? For this we use the hybrid model (23) To test between these alternatives we use overidentification tests on the cointegration vector (which is just identified), restricting the hybrid model (23) above to conform to one or other of the models on which it is based. The tests we consider are first that a1=1, a4=0. This would reject a separate influence from relative factor prices. Next, we consider evidence that a2=0. If upheld, this implies profitability and relative factor prices can be excluded. Finally, we consider the other alternative; that it is liquidity and relative factor prices which can be excluded Table 2.3 shows the results of likelihood ratio tests conducted on a hybrid model which uses all the variables from (3) and (4) in the model. It also shows that the first model with the relative factor price excluded and a unit co-efficient on output is accepted. Dropping profitability (Π) is strongly rejected. The alternative, keeping Π in the model, but dropping liquidity (L) is also rejected, but less strongly than the previous case. So the tests are not conclusive as between the choice between liquidity and profitability. They clearly reject the use of relative factor prices. So where does that leave us? We conclude that a parsimonious model for the level of investment, based on output, profitability and liquidity is the preferred version according to our results. We will use this long run result in the dynamic model which follows. Finally, in this section, a dynamic investment equation is estimated. Given the results from Table 3 the first equation is used, which is a hybrid model, incorporating financial variables both as suggested by Woods (ibid.) and Sentance and Urga (ibid.). A dynamic ECM, estimated for the period 1968Q3–1992Q1, with the first equation from Table 2.3 used as the ECM term is shown next

26

HENRY, SENTANCE AND URGA

(24)

This equation passes the usual misspecification tests: there is no evidence of significant serial correlation inappropriate functional form non-normal errors or heteroscedastic errors Interestingly, it suggests that there are short run negative effects from the financial variables, although these effects are positive in the long run. Table 2.3 Hybrid model: likelihood ratio tests a1

a2

α3

a4

1.0 1.0

0 .53 —

1.12 1.14

— —

Restrictions

a1=1, a4=0 a1=1, a4=0 a2=0 1.0 0 .6 — — a1=1, a4=0 a3=0 Note: Equations included restricted time trends, which are not shown for convenience.

LR X2(2)=1.8 X2 (3)=20.4 X2 (3)=10.3

Having obtained what appears to be the most successful of the models (without explicitly incorporating expectations), the final exercise is to consider how well it copes with the expansionary phase post-1992. Standard predictive failure tests can enlighten us on this score. For this we use a further misspecification test on evidence for parameter instability in the model. For the period 1992Q2– (18) of 31.88, which does not reject 1996Q3, the predictive failure test for the model is marginally upheld (i.e. it gives a the null of no change at the 5 per cent level). The role of expectations in manufacturing This final section considers the potential role for expectations. As anticipated earlier, both direct, survey based, measures of expectations, and full REH expectations of the determinants of investment are used. We review the results in turn. CBI output expectations Initially, the dynamic error correction model just discussed is re-estimated using an expected output series generated from the CBI output expectations survey ( ). This uses the expected quarterly change given by quantifying the survey, together with the actual level of manufacturing output to obtain an expected level of output for the (end of) the next quarter. Thus The next step replaces the actual output level in the first equation of Table 2.3 with expected output. This led to a cointegrating vector for investment (LR=45.1*, LT=74.4*). The vector with a unit elasticity for output imposed then gave A result very similar to that obtained when using actual output. Proceeding to the dynamic model also includes an output ‘surprise’ term ( ) as an additional dynamic (I(0)) term. According to this, if output is above expected, than this will encourage investment, although we might expect this to be a slow process given the typical lags in adjusting investment. The empirical result below in equation (25) shows there is support for including such a term, and this version of the model also has a somewhat better predictive performance compared with its predecessor, i.e. it now does not reject the null of no change at 5 per cent level. Although this is not powerful evidence, it is suggestive that indicating introducing explicit expectations improves certain aspects of the model.

(25)

FINANCE, PROFITABILITY AND INVESTMENT

27

Explicit forward looking (REH) models The previous results were suggestive that output expectations might be useful in helping to account for recent investment behaviour. The next set of results takes this further, and tests whether explicit rational forward looking behaviour improves its behaviour even more. To do this, the reparameterised version of the forward looking model described earlier is used, and equation (22) which is the form in which the model is estimated, is repeated here for convenience. (26) There are considerable advantages in using this procedure of reparameterising the dynamic REH model. It enables the earlier cointegration results to be carried over into an REH model for one thing. Also, by comparing the results from the restricted (REH) model with an unrestricted version, a test of the REH restrictions can be made. According to the latter, if these restrictions are upheld, then, for this model., there is evidence that forward looking behaviour matters. Taking the results from Table 2.3 again, with the cointegration equation for investment dependent upon output, profitability and liquidity, a forward looking version of this model is that it is expected values of these variables which determine investment. As (22) above shows, these forward expectations hold over a finite future period. In what follows, up to six quarters is used. The REH model takes forward expected values for investment determinants, but as actual future values are used in estimation, IV methods are employed for reasons described earlier. The REH restrictions are those imposing symmetry between the backward looking and forward looking parts of (15) (the restrictions are between λl, λ2 and the dj in the equation). Hence, to impose the restrictions non linear estimation is also necessary, yielding estimation for the key parameters λ1 and λ2. Results are shown in Table 2.4 below. (Using non-linear least squares). Of these, the second equation, which restricts the forward planning horizon to three quarters, appears preferred. Its serial correlation properties are much improved compared with (1). The models in Table 2.4 are estimated subject to REH restrictions between the lagged part and the future, Table 2.4 REH models for investment λ1 (1)

0.97 (6.9) (2) 1.599 (4.9) Instruments lagged: L, YM

λ2

SE

DW

Horizon (quarters)

−0.35 (3.9) −0.89 (3.87)

0.113

0.8

6

0.14

2.3

3

led, part of the equation. Tests of the validity of these restrictions can be made to judge the applicability of the restrictions, and hence the plausibility of informed forward looking expectations. These tests use quasi-likelihood ratio (QLR) tests, which are more appropriate to non-linear IV equations such as those used here. For model 2 in the table, the QLR gave X2(4)=3.0, which does not reject the REH restrictions. To sum up this section, we have shown that some broadly defined financial variables appear to help in the explanation of manufacturing investment. But, even when these are included, the best version of our lagged model does not appear to account for the slow recovery in investment during the most recent recovery. Including survey based output expectations improves the predictive power of the model though, but not hugely, it needs to be said. Finally we test whether the restrictions necessary to support rational expectations formation are upheld. For the most plausible version of the model we use, the answer appears to be in the affirmative. Conclusions Our conclusions can be divided into two parts: first, we have investigated a number of different models of investment behaviour. We have focused on two issues which have been at the centre of empirical research in this area: the role of financial effects and of expectations upon investment. On the former, our work has tended to confirm that the familiar cost of capital and Tobin’s q variables do not have much explanatory power. However in tests between other financial variables which have figured in recent discussions the results are more encouraging. In particular, by discriminating between a range of variables using cointegration techniques, models using profitability and liquidity appear to be capable of accounting for much of investment behaviour. Building on these results, tests of the probable importance of expectation effects suggest that versions of the model using expected future values of these variables are preferred and we believe this is a fruitful area for further research—particularly using direct measures of expectations from business surveys. This modelling work does shed some light on the second issue, which is the apparently puzzling behaviour of investment over the recovery. Our first conclusion in this area is that there may be much less of a puzzle than is apparent at first sight. We

28

HENRY, SENTANCE AND URGA

have found it possible to establish satisfactory equations for manufacturing investment based on a long-run relationship between investment, output and profitability (or the rate of return). We also find a significant role for liquidity constraints, and in all equations there is support for a unit elasticity for output in the long-run, so we would expect investment to rise broadly in line with GDP, in the absence of major shifts in profitability. Admittedly recent investment performance in the manufacturing sector is more of a problem. Here, the puzzling phenomenon is the sharp drop in manufacturing investment in the first half of 1996. The model including expected output and other variables are only suggestive at this stage. They hint that expected demand and other factors did not improve sufficiently quickly in the recovery for investment to increase. However mismeasurement may be another explanation. The CBI Industrial Trends Survey has continued to show strong investment intentions while the data has been showing that capital spending has been collapsing. It is too early then to conclude from this study that expectations hold the key. Further work is needed to clarify this point. Data appendix

CC

Cost of capital in manufacturing

δ – depreciation rate r – (nominal) interest rate Pk – price of capital goods YM

GDP manufacturing

MI

Fixed capital formation—manufacturing

MK

Gross capital stock at 1990 replacement cost—total manufacturing

RW

Real labour cost=Average earnings manufacturing/GDP deflator

FPM

Relative factor prices (Manufacturing)=(CC * GDP deflator market prices)/average earnings manufacturing

CG

Gearing=ICC’s net indebtedness/(net capital stock at replacement cost+value of stocks)

L

ICS’s liquidity ratio: liquid assets as a proportion of liabilities

UN

Union density

qA

Average Tobins q: market value of equity and debt as a ratio of the replacement value of assets Expected output: quantified series obtained from the CBI output expectations survey

CU

Capacity utilisation (measured by responses to CBI question 4)

Π

Profitability: non oil profits as a percent of capital at replacement cost

Notes 1 Grateful thanks for help in collecting data used in this paper are due to James Clarke, David Upton, and Paul Robson for assistance in applying CBI survey based measures of expectations. Thanks are also due to Rod Whittkaer and Robert Woods HMT, and Nigel Jenkinson and Simon Whittaker, Bank of England for providing data. Michael Summer provided useful comments for which we would like to express our thanks. None of the above share in the responsibility for contents of the paper which rests entirely with the authors. Financial support from ESRC grant number L116251013, Macroeconomic Modelling and Policy Analysis in a Changing World, is gratefully acknowledged by Andrew Sentance.

FINANCE, PROFITABILITY AND INVESTMENT

29

References Callen, T., Hall, S. and Henry, S., (1990), ‘Manufacturing Shocks; Expectations, Risk and Cointegration’, Economic Journal, 100, 756–772. Chirinko, R.S., (1993), ‘Business Fixed Investment Spending: Modelling Strategies, Empirical Results and Policy Implications’, Journal of Economic Literature, 31, 1875– 1991. Cuthbertson, K. and Gasparro, D., (1995), ‘Fixed Investment Decisions in UK Manufacturing: The Importance of Tobin’s Q Output and Debt’, European Economic Review, 39, 919–941. Cuthbertson, K., Hall, S. and Taylor, M., (1992), ‘Applied Econometric Techniques’, London: Harvester Wheatsheaf. Denny, K. and Nickell, S., (1992), ‘Unions and Investment in Britain Industry’, Economic Journal, 192, 874–887. Dixit, A. and Pindyck, R., (1994), ‘Investment under Uncertainty’, Princeton University Press. Driver, C., Abubaker, S., and Argiris, G., (1996), ‘Capacity Choice under Monopoly, Flexible Prices and Demand Uncertainty’, Southern Economic Journal, 63, 2, 526–532. Hall, S.G. and Henry, S.G.B., (1988), ‘Macro Economic Modelling’, Contributions to Economic Analysis, North-Holland. Kelly, C. and Owen, D., (1985), ‘Factor Prices in the Treasury Model’, GES working paper, 83. Nixon, J. and Urga, G., (1997), ‘Unemployment and the Capital Stock: Modelling the Supply Side of the UK Economy’, DP 18–97, Centre for Economic Forecasting, London Business School. Pesaran, M.H., (1984), ‘Expectations Formation and Macro-economic Modelling’, in P.Malgrange, and P.A.Muet (eds.), Contemporary Macroeconomic Modelling, Basil Blackwell. Precious, M., (1987), Rational Expectations, Non Market Clearing and Investment Theory, Clarendon Press. Schiantarelli, F., Georgoutsos, D., (1990), ‘Imperfect Competition Tobin’s Q and Investment’, European Economic Review, 34, 1061–1078. Sentance, A. and Urga, G., (1995), ‘Boom Time for British Manufacturing Investment’, Economic Outlook, 19, 3, London Business School. Stiglitz, J. and Weiss, (1992), ‘Credit Rationing in Markets with Imperfect Information’, American Economic Review, 71, 393–410. Summers, L., (1981), ‘Taxation and Corporate Investment: A q-Theory Approach’, Brookings Papers, 1, 67–127. Urga, G., (1995), Firms in Investment Decisions: Cash Flow, Unions and Ownership Structure. Evidence from Italian Firm Level Data, Centre for Economic Forecasting, London Business School. Woods, R., (1995), ‘Econometric Models of Business Investment: The Role of Factor Prices, and Financial Constraints’, HM Treasury. GES working papers 127, Treasury working papers 69. Young, G., (1996), ‘A New System of Factor Demand Equations for the NIESR Domestic Model’, paper presented to the Macroeconomic Modelling Bureau Seminar, University of Warwick, July.

3 Credit rationing versus consolidation of financial structure Jean-Bernard Chatelain1

Introduction A high sensitivity of investment with respect to cash-flows is quite often related to financial constraints affecting firms. But recently, Kaplan et Zingales [1997] found that low dividend companies in the widely quoted Fazzari Hubbard and Petersen [1988] study can be split in several groups when taking into account additional and mostly qualitative information on the extent to which these firms were effectively facing a financial constraint. It turned out that their ‘financially constrained companies according to their new information’ were exhibiting a very low investment—cash-flow sensitivity with respect to the other low-dividends companies. Recent tests of financial constraints used the Euler equation (the marginal condition for capital) of the neoclassical model of investment. Bond and Meghir [1994] state that it is negative in the unconstrained model, and rising, eventually up to positive values, for financially constrained models. This point of view is similar to Hubbard, Kashyap and Whited [1995] when they make the value of relaxing the debt to capital ratio ceiling a decreasing function of cash-flows. The Lagrange multiplier is also parameterized as an increasing function of the ‘coverage ratio’, which itself is a decreasing function of cashflows, as in Whited [1992]. Finally, when this Lagrange multiplier is substituted in the Euler equation, as done by Jaramillo, Schiantarelli and Weiss [1996], both kinds of financial constraints (debt ceiling and an increasing cost of leverage) have no effect on the investment—cash-flow sensitivity, because the Lagrange multiplier related to a debt ceiling is a common multiplicative factor of both current investment and current cash-flows in the Euler equation. These contradictory pictures covering a central feature of investment behaviour are based on similar neoclassical models. The insights provided by the theoretical model have not been totally investigated, as acknowledged by Chirinko ([1993] pp. 1902–1904) in his research agenda described in the end of his Journal of Economic Literature survey for the economics of investment: ‘further work relating investment and financing decisions to explicitly specified capital market Frictions is clearly needed’. More precisely, Fazzari, Hubbard and Petersen ([1996], p. 26) put forward the argument that Kaplan and Zingales’ firms— tracked as low dividends, low investment—cash-flow sensitivities—needed mostly to use their cashflows to repay their debt instead of financing investment (they labelled this behaviour as relevant for ‘financially distressed’ firms). But in the regime with ‘only an increasing cost of leverage’, the entrepreneur faces a marginal return on investment which is lower than the marginal cost of its current debt (with this one being higher than the opportunity cost of retained earnings). Then, he has the incentive to use his cash-flows to repay part of his debt, so that the lower marginal cost of his new level of debt just matches the marginal productivity of the new level of capital (see Hubbard [1998]). This implies a negative relationship between debt and cash-flows and that investment should be more or less sensitive to cash flows, according to the slopes of investment demand and credit supply. By contrast, in an (endogenous) debt ceiling regime, investment is very sensitive to cash-flows (and the debt ceiling increases with profitability), as all of them should be used to fill the gap between the notional capital stock and effective credit rationed investment as soon as possible. The implications for fiscal policy as well as for monetary policy of taking into account these two different regimes are of major importance. For example, subsidies to investment or decrease on corporate taxation may have a much lower impact on investment, as these funds will be diverted to decrease of the amount of debt of firms facing regime 1, whereas these tax changes will have a strong impact on credit rationed firms. Moreover, a rise in real interest rate could, for example, lead firms to shift from regime 2 to regime 1, and thereby affecting aggregate investment through this composition channel. The ‘consolidation’ behaviour described above is not only the one of ‘financially distressed’ firms as described by Fazzari et al. [1996]. It can be adopted by financially very ‘healthy’ firms (e.g. with low leverage with respect to the sectoral average), when they have considerable cash-flow with respect to their investment demand. These firms may still have an incentive to provide low or zero dividends: the quicker they decrease their debt ratio, the quicker the average cost of capital decreases. In this respect, they are still ‘financially constrained’. Examples are provided by a lot of French firms following the rise of real interest rate in the early 1980s and presumably fostered by the gradual increase of the share of profit in value added and the development of the equity market: they used retained earnings to repay their debt instead of investing. It is

CREDIT RATIONING VERSUS CONSOLIDATION

31

reflected in both national accounts and individual accounting data. At the aggregate level, the debt-equity ratio fell from 140 to 40 in six years (from 1983 to 1989) followed by relative stability. For most years of the 1990s, the aggregate level of retained earnings was more than 100 per cent of aggregate investment. On individual data, the fall is observed on a longer period, from 1983 to 1994. Most of the literature is based on an exogenous debt ceiling (Gertler, Hubbard and Kashyap [1990], Hubbard and Kashyap [1992], Whited [1992] and Hubbard, Kashyap and Whited [1995], henceforth GHKW). But if one assumes that the debt—capital ceiling depends on expected profits (i.e. future cash-flow net of bankruptcy costs are considered as collateral), the Lagrange multiplier affects future investment and cash-flows differently, even though this multiplier does not depend on cash-flows. Kaplan and Zingales variations of investment-cash-flow sensitivities for low dividend firms can be accounted for by the Euler equation, as a difference between firms facing credit rationing and an increasing cost of leverage versus firms facing only an increasing cost of leverage. Although the effects of the endogenous increase of the debt ceiling may be marginal, assuming an endogenous debt ceiling extends the range of structural parameters to be estimated, such as expected bankruptcy costs. This chapter first of all builds a theoretical model which relates investment and financing decisions to capital market frictions as specified by Kiyotaki and Moore [1997], and defined by an endogenous upper limit to the debt to capital ratio. This novelty is introduced into a standard model of investment behaviour for an infinitely long lived firm (Barran and Peeters [1998]). The empirical work which follows in this chapter develops the approach opened by Jaramillo, Schiantarelli and Weiss [1996] by testing the specificity of credit rationing with respect to an increasing cost of borrowing. A more structural test is proposed which simultaneously takes into account the debt and investment behaviour. The Lagrange multiplier on the debt ceiling constraint on dividends is usually parameterised in an ad hoc manner by empirical researchers (GHKW). But this Lagrange multiplier is related to the gap between the standard neoclassical investment behaviour and the (explicit) credit constrained investment. I derive its explicit expression which suggests specifications closer to the structural one for Euler equation tests of financial constraints. A second step towards a more structural modelling is to test simultaneously the Euler equation for capital and the flow of funds equation which determines both debt and investment behaviour for zero dividend firms. The remaining part of the chapter is as follows. The next section presents the intertemporal model of the firm facing a debt ceiling and an increasing cost of borrowing, as the degree of leverage increases. The following section describes the investment regimes. We then parametrize agency costs and adjustment costs for testing. This is followed by an empirical test and the conclusion. Intertemporal behaviour of the firm The model The model allows for an increasing cost of borrowing, as the degree of leverage increases, and for a ceiling on the latter as additional constraints with respect to the neo-classical model of investment. It is identical to the models of Jaramillo, Schiantarelli and Weiss [1996] or Barran and Peeters [1998] except on the point that the leverage ceiling is endogenous. Analysing investment begins with an expression for the value of the firm, which in turn stems from the arbitrage condition governing the valuation of shares. The after-tax return to the owners of the firm at time t reflects capital appreciation and current dividends. In equilibrium, if the owners are to be content holding their shares, this return must equal the after tax nominal return on riskless (government) bonds between period t and period t+1 ( represents the nominal return before income tax and mt is the personal income tax on dividends and interest income in period t).2 (1) where Vit is the value of the firm i at time t, Sit denotes the value of new shares issued at time t+1, ct is the accrual-equivalent capital gains tax rate, θt is the divided received by the shareholder when the firm distributes one unit of post-corporate tax earnings.3 Therefore, the tax rate on dividends is (1−mt)θt. Et is the expectation operator conditional on information known at time t. The after-tax capital gain of the current shareholders thus consists of the change in the market value of the firm less the component of this change due to new share issues. The dividends of the firm at time t+1 are di,t+1. In the absence of bubbles, solving the capital market arbitrage condition yields the following expression for the firm’s market value at time zero: (2) where the firm’s one period nominal discount factor is:

32

JEAN-BERNARD CHATELAIN

(3) We define γt as the tax discrimination parameter that determines the relative tax advantage of dividend income against capital gains: (4) This is the first source of discrimination between retained earnings and new share issues, which implies that in most tax systems, retained earnings are cheaper than new share issues. The entrepreneur in firm i chooses dividends, investment, labour and price of output in maximizing the present value of dividends di,t on date t in an infinite horizon, with the discounted rate of the opportunity cost of internal funds, subject to several constraints. Initial conditions for internal equity Ai, 0 and for debt Bi,0 are given (Ki,0=Ai,0+Bi,0). The first constraint is the capital stock accounting identity, which defines the law of accumulation of a first state variable capital Ki,t. (5) where Iit is its investment at time t, and δ is the constant rate of economic depreciation. The second ‘flow of funds’ constraint defines firm dividends. Cash inflows include sales, new share issues, and net borrowing, while cash outflows consists of dividends, factor and interest payments, and investment expenditures: (6) Where: Nit=a vector of variable factors of production for firm i at time t, F(Ki,t−1,Nit)=the firm’s revenue function (FK>0, FKK<0),4 Ψ(Ki,t−1,Iit)=the cost of adjusting the capital stock (ΨI >0, ΨII >0, ΨK<0, ΨIK<0),5 wt=a vector of nominal factor prices at time t, iit=the nominal interest rate on debt at time t, Bit=the value of net debt outstanding for firm i at time t, =the expected inflation rate at time t, pit=the price of final goods at time t, =the price of capital goods at time t (incorporating tax considerations), τt=the corporate income tax rate, against which interest payments are assumed to be deductible, fit=a transaction charge that has to be paid per unit of new share issues (as in Bond and Meghir [1994].6 This is the second source of discrimination between retained earnings and new share issues. Increasing new share issues by 1/ (1−fit) units allows one additional unit of dividends to be paid to shareholders, which is valued at γt. These transaction costs can eliminate the case for issuing new shares in order to pay dividends. Therefore, it is no longer necessary to include the constraint that dividend payment cannot exceed the excess of revenues over labour cost and interest payments (constraint (4a) in Schiantarelli and Georgoutsos [1990, p. 1064]). We assume the transaction charge to be high enough so that retained earnings are the preferred source of investment finance with respect to new share issues: (7) The third constraint is that the firm faces a downward sloping demand function for its product. Output demanded is assumed to depend upon the price charged by the firm relative to the average price of competitors, pit/Pt and upon a shift parameter Δt that summarizes all the factors determining the position of the demand curve. The average price level is taken as given by each individual firm. Demand is equal to output net of adjustment costs. The inverse of the demand function of the firm can, therefore, be written as: (8) The following positivity constraints describe capital market imperfections. The fourth constraint restricts dividends to be non-negative: (9)

CREDIT RATIONING VERSUS CONSOLIDATION

33

A signalling constraint on dividends floor can also be imposed so that 0
First order conditions The Lagrangean is (with debt ceiling):

the Lagrange multiplier associated with each positivity constraint z,

related to the endogenous

34

JEAN-BERNARD CHATELAIN

(17)

Dividends are substituted directly in the Lagrangean, using the flow of funds constraint as well as investment by the capital accumulation equation. I denote the inverse of the markup where ε represents the elasticity of demand. The first order condition for labour Nit, for capital Kit (Euler equation) for debt Bit and for new share issues Sit are respectively: (18)

(19)

(20)

(21) The changes with respect to Barran and Peeters [1998] marginal conditions are that the credit limit is endogenous, so that new derivatives multiply the Lagrange multiplier of the debt ceiling constraint λx in the marginal conditions for capital and debt. The marginal condition for capital (Euler equation) states that the marginal productivity of capital, corrected by the marginal costs of capital adjustment and investment adjustment is ‘higher’ than the marginal cost of debt, as the Lagrange multiplier of the debt ceiling is strictly positive . Credit rationing implies that the productivity of capital is higher than the marginal cost of debt. The complementary slackness conditions are: (22) (23) (24) Transversality conditions and Euler equations leads to sufficient conditions for the optimal plan of the firm (Stokey and Lucas [1989]): (25) (26)

CREDIT RATIONING VERSUS CONSOLIDATION

35

Investment regimes In order to solve the model theoretially, it is necessary to study all the possible regimes. These regimes are then assessed with respect to their empirical relevance, and we point out features able to provide empirical tests used in the following sections. Unconstrained regimes Regimes are defined with respect to the strict positivity or not of the Lagrange multipliers. We consider the case where a firm . We assume that the remains in the same regime on date t and t+1. First of all, a firm issues new shares when transaction cost on new share issues (ft) is sufficiently high with respect to the marginal cost of retained earnings increased by the shadow price of negative dividends:10 (27) Therefore, we no longer consider the new share issues regimes in what follows. Two of the eight remaining regimes determined according to the strict positivity or not of Lagrange multipliers are impossible (debt cannot be simultaneously at its maximal level and at its minimal level, for dividends equal to zero or not). It is useful to list the different regimes for the debt level in a table (the subscript t,t+1 indicates in Table 3.1 that the constraint is binding on the two dates, a+indicates that the variable is strictly positive). We label regimes 1, 2 and 3 as unconstrained ones, and regimes 4, 5 and 6 as financially constrained ones (as dividends are constrained at their lower bound). In our applied work, we eliminate the empirically irrelevant regimes where firms do not hold debt (regime 1 and regime 4, as well as regime 2.11 Therefore, we retain the unconstrained regime 3, the credit rationing regime 5 and the consolidation (increasing cost of leverage) regime 6. The first order condition for debt Bit leads to specific conditions on the relative costs of retained earnings with respect to debt for ‘unconstrained’ regimes. For regime 3, debt is not zero and does not hit its ceiling, so that the marginal cost of debt is ξit (Bit). For simplicity, I denote the after tax and after inflation marginal cost of debt as: (28) Table 3.1 Different regimes for debt level Regimes 1 2 3 4 5 6

; + + + + + +

0 0 0 + + +

+ 0 0 + 0 0

0 + 0 0 + 0

Bit

di,t; di,t+1

0

+ + + 0 0 0

0

The marginal condition on debt is: (29) If ever the tax parameter γt does not change on the next accounting year, this condition simply means that the opportunity cost of retained earnings is equal to the marginal after-tax real cost of debt. This condition determines the level of debt in this regime, while the Euler equation determines investment and the flow of funds equation determines the level of dividends. It is a simplified version of the equation (A1) in Bond and Meghir [1994] appendix. Bond and Meghir test this regime as their unconstrained regime, which includes the square term of the debt-capital ratio in the Euler equation for capital.12 ‘Consolidation or financially distressed’ regime In regime 6, the Lagrange multiplier on the debt ceiling is zero ( ) and the Lagrange multiplier on the dividends floor is strictly positive ( ). One can use the marginal condition on debt to eliminate the discount factor in the Euler equation for capital. They substituted the left-hand side of the following equation by its right-hand side (including the marginal cost of debt) in the Euler equation for capital:

36

JEAN-BERNARD CHATELAIN

(30) As the Lagrange multiplier on the debt ceiling is zero, the parameter does not show up in the Euler condition for capital, after substitution of the ‘corrected discount rate’ by ξit(Bit). Therefore, the Euler equation is identical to the one related to the unconstrained regime 3.13 In both regimes, the expected marginal productivity of capital (corrected by marginal effects on adjustment costs) remains equal to the expected marginal cost of capital, when it is financed by debt. But if the first order conditions for capital are identical, it is no longer the case for the first order condition for debt Bt which is affected by in this regime. It implies a recursive dynamic for the expected Lagrange multiplier on dividends: (31) If βitξit (Bit)>1, i.e. the marginal cost of debt is higher than the opportunity cost of retained earnings, then, the Lagrange multiplier on dividends decreases exponentially towards 0. The financial constraint tends to disappear. The marginal condition on investment is also altered by (32) Investment is determined by the dynamics of the Lagrange multiplier on dividends

and the Lagrange multiplier on capital

It is not yet rational to pay dividends to shareholders, because retained earnings are a cheaper means of finance. Marginal productivity of capital is now such that it is worthwhile to get rid of the debt, as, at the margin, the return on investment can become lower than the marginal cost of debt. In this regime, the debt-capital ratio increases with the previous period financial structure and decreases with profits and increases with the interest rate. This goes the opposite way with respect to the endogenous debt ceiling theory. As only a proportion of retained earnings are used for investment (the remaining part being used to get rid of the debt), the sensitivity of investment to profits is lower in the ‘financially distressed’ regime than in the debt ceiling one. It is important to note that a particular consolidation behaviour holds also for an anticipated binding debt ceiling constraint. For firms who do not face a binding debt ceiling on the current date, a first result derived from the marginal condition on debt is that the shadow cost of external finance is zero ( ) if and only if the Lagrange multiplier on dividends (corrected by the tax parameter γt) in the current period is identical to the expectation of the Lagrange multiplier on dividends (corrected by the expectation of the tax parameter γt+1) in the next period, as shown below: (33) Therefore, two regimes where the shadow price of external finance is zero ( ) are possible. In the first one where the firm is expected to pay dividends in the current and the future periods ( ) it boils down to one of the three unconstrained regimes previously described in the previous section. In the second one, the firm does not pay dividends in both periods, with a constant value of the shadow price of negative dividends ( ), although there is no credit rationing in the current period. This regime holds in the case of an anticipated credit rationing on a future date . Solving forward the marginal condition on debt leads to: (34) (35) Therefore, the manager of the firm decides to accumulate current profits instead of providing dividends in order to soften the anticipated constraint in the future. None the less, this regime cannot be a final one, as no shareholders would accept to receive no dividends for ever (they will require capital gains at a finite date). One can notice that the condition for this regime requires only that a firm can face a strictly positive (even very low) probability of being credit constrained in the future (for ), so that it can be related to a non-negligible number of companies. Credit rationing regime: explicit Lagrange multiplier Papers in the field dealing with credit rationing (GHKW), except Jaramillo et al. [1996], used the marginal condition on debt to eliminate the discount factor in the Euler equation for capital. They substituted the left-hand side of the following equation

CREDIT RATIONING VERSUS CONSOLIDATION

37

(denoted for adjusted discount rate in case of credit rationing) by its right-hand side (including the marginal cost of debt) in the Euler equation for capital: (36) One can note that the parameter is not random on date t. One needs to compute the parameter for the composite Lagrange multiplier for credit rationing Euler equation. This means computing directly the modified discount rate explicitly The marginal condition on investment in the credit rationing regime is:

which modifies the i.e. computing (37)

Investment for the firm facing credit rationing may be derived, after parameterization of the adjustment cost function. A structural specification of the adjusted discount rate for credit rationing, that we denote is given by: (38) The marginal costate variable is Tobin’s marginal q in the unconstrained regime (Hayashi [1982]). In this case, the marginal equation on investment implies that it is a linear function of the unconstrained investment regime rate .14

To get the explicit expression of the Lagrange multiplier, one uses the marginal condition on debt:

The explicit Lagrange multiplier suggests specifications closer to the structural one for Euler equation tests of financial constraints. This expression was not known in previous studies.15 From theory to testing For testing, I parameterize the agency cost function as e.g. the lemons premium is a linear form with respect to the debt-capital ratio: (Jaramillo, Schiantarelli and Weiss [1996]; Barran and Peeters [1998]). I specify an adjustment-cost function which is linearly homogeneous in investment and capital with the standard symmetric parameterization (Whited [1992; p. 1434]; Bond and Meghir [1994; p. 207]):16 (39) where υ can be interpreted as the ‘normal’ rate of investment (theoretically, it could be equal to the scrapping rate δ). By differentiating this adjustment cost function with respect to Iit and Kit: (40)

(41) Under the assumption of linear homogeneity of the production function, the following equality holds: (42) In this expression, FN and FK are the marginal productivity of capital at period t−1 and labour at period t, respectively. Using the marginal condition on labour in the preceding equation leads to: (43)

38

JEAN-BERNARD CHATELAIN

Jaramillo et al. [1996] proposed a first test of the two regimes (consolidation versus credit rationing, described by an exogenous ceiling on the debt—capital ratio). As they did not have a precise idea of the Lagrange multiplier λx, they decided to use the marginal condition on debt to eliminate it in the Euler equation for capital, instead of eliminating the rate of time preference and parameterizing the Lagrange multiplier in ad hoc manner (GHKW studies). Then they use the deposit rate as a proxy for the rate of time preference (the opportunity cost of internal funds). This is a clever trick, but unfortunately, it worked only because they did not include the Lagrange multiplier on dividends, although they did include the non-negativity constraint on dividends in their initial programme. Such a constraint cannot be omitted: a binding debt ceiling can be offset by negative dividends as noted by GHKW. Using the marginal conditions on debt, one can eliminate only one parameter (γx or γd) in the marginal condition for capital, but not both. What is more, as seen in previous sections, γd is likely to present quite a different behaviour in the credit rationed regime than in the consolidation regime. Therefore, we cannot replicate the Jaramillo et al. [1996] testing strategy. As we have a more precise idea of the structural variables which should be included in the Lagrange multiplier λx, we prefered to use this new information on the adjusted rate of time preference in the Euler equation for capital. One substitutes into the Euler equation for capital FK, ΨI, ΨK, ∂iit/∂Kit, replacing βit[ ] by Et {βit[ ]} which can be substituted by 1/[ξit(Bit)] (see equation 34) in the consolidation regime or by in the credit rationing regime (1C is a dummy variable equal to one in the credit rationing regime, and else equal to zero), and finally .17 This results in the following Euler equation to be tested:

(44)

I have introduced a stochastic error in place of the expectation operator εi,t+1. The assumption of rational expectations implies that εi,t+1 is a white noise expectation error, uncorrelated with any period t information. In addition, we have allowed for the possibility of fixed firm-specific and time-specific effects, denoted fi and YEARt respectively. One can check the optimal conditions for an exogenous debt ceiling (B≤B*) and that in this particular case, ζП=ζK=1 in the above Euler equation. On the one hand, in order to get a non-fractional (but still non-linear) model to test, a second order Taylor development of the adjusted discount rate for the consolidation regime leads to: (45)

(46) (47)

CREDIT RATIONING VERSUS CONSOLIDATION

39

On the other hand, a first order Taylor development of the adjusted discount rate in the credit rationing case with respect to the adjustment cost parameter α (expected to be small) leads to (for a time invariant statutory income tax τt on firms, which was the case over our estimation period, and a constant relative price of investment with respect to output price): (48) With the given parameterization of the adjustment cost fiunction, it is possible to find the explicit expression of the first difference of investment facing credit rationing, as a solution of a quadratic equation and consider the first term of the Taylor expansion with respect to the adjustment cost parameter α (Appendix 3.3):

Δ is the first difference operator. Explanatory variables of include the explanatory variables of the financially constrained investment (dividends are constrained to their lower bound): the growth or the first difference between the next period and the current one of the expected (cash flow)–(equity) ratio, the growth of the expected (debt)–(equity) ratio. Other variables include the real marginal cost of debt, the discount factor and finally the growth of the shadow cost of capital in the unconstrained regime (any variables which can determine the variation of Tobin’s marginal for a firm which is not financially constrained). For testing, we estimate the unobservable growth factor of the unconstrained Tobin’s q divided by a discount factor, that we denote

The assumption that the debt—capital ceiling may depend on expected profitability implies that the Lagrange multiplier of the debt ceiling constraints affect differently current cash-flow and current investment through the combination of the bankruptcy cost parameter ζΠ and of the Lagrange multiplier . The coefficient of cash-flows divided by the coefficient of current investment is:

(49)

When the firm faces only an increasing cost of leverage (as defined by Jaramillo, Schiantarelli and Weiss [1996] agency cost function) or when it does not face a financial constraint, the coefficient boils downs to a negative value (−1/(1−δ)α). This standard result is acknowledged for the unconstrained model by Bond and Meghir [1994], In the credit rationing case, the sensitivity is modified. Therefore, Kaplan and Zingales variations of investment-cash-flow sensitivities can be accounted for by the Euler equation in our model. Note also that the sensitivities for financially constrained firms are firm specific in the credit rationing regime. Then we tested this Euler equation (including the Lagrange multiplier λx) simultaneously with the flow of funds. It is necessary to add the flow of fund equation in the estimation in the case of financially constrained regimes (with new share issues too costly and dividends hitting the dividend signalling floor). The flow of funds equation relates directly the current level of debt to investment, so that their behaviour are intimately linked. In the credit rationing regime, the debt is determined by the debt ceiling, so that the flow of funds allows us to derive an explicit expression of investment. In the consolidation case, investment is determined by the Euler equation, which allows us to determine the debt level with the help of the flow of funds constraint. More generally, the connection between debt and investment is given by the system of the Euler equation on

40

JEAN-BERNARD CHATELAIN

capital (after substitution of the parameter given by the Euler equation for debt), of the flow of funds equation and of the quantitative constraints.18 The flow of funds can be normalised by the total value of capital, taking into account the parameterization of the adjustment costs function:

(50)

An econometric test of the investment regimes Data The data used comprises annual accounts data extracted from the ‘Centrale des Bilans’ database at the bank of France for 292 firms in the manufacturing sector, continuously present for the period 1989–1994 (six years).19 This sample was done after the elimination of outliers for five variables: the investment/total assets, cash-flows—total assets, average productivity, apparent interest rate (debt service-debt) and the rate of growth of sales, by eliminating the tails of their distributions at the 1 per cent level. In order to use the second and third lags of variables as instrumental variables, estimations deals with the three years 1992–1994. This selection of 292 firms includes only firms which did not provide dividends during these three years. This data set includes smaller firms and non-quoted ones, but over a smaller period than the Fazzari, Hubbard and Petersen’s [1988] sample of quoted companies used by Kaplan and Zingales [1997]. Searching for credit rationed firms Splitting the sample between high and low dividends firms is not sufficient for exploring financial constraint. We investigated two variables to split the sample of zero dividend firms between firms facing credit rationing and firms doing consolidation of their financial structure. (1) We consider firms who experienced a decrease in debt between 1992 to 1994. We split the sample according to whether (Bi,94−Bi,92)/Ai,92 is higher than the first 33 per cent of the distribution of the variable; Ait is total assets, used to normalize the variables. In so doing, we have at least around 100 firms in the first smallest group, which is necessary for the quality of GMM estimation. Firms doing consolidation should be mostly in the first group, while credit rationed firms should be in the second group. (2) In the low dividends firms, it is possible to discriminate firms facing credit rationing as high profitability firms (the interest rate on their current debt does not adjust to their marginal productivity) and firms willing to consolidate their financial structure as low profitability firms (the apparent interest rate on their debt is higher than the marginal productivity of investment). This condition is taken from a simplification of the Euler equation in the particular case (no adjustment cost (α=0) and so on): (51) We take the average productivity of capital as a proxy for the marginal productivity of capital under our assumption of constant returns to scale. We compute the ‘apparent’ interest rate as the ratio of debt service with respect to debt. We constructed a profitability variable as difference between the marginal product of capital and the ‘apparent’ interest rate on current debt minus the current year inflation rate (for expected inflation). It allows us to compare the return of investment inside the firm with respect to the return on consolidation (decrease of the burden of the current debt) for a constant scrapping rate. The sample has been split between two groups of firms according to whether the median of their profitability between 1992 to 1994 is over (high profitability firms) or under (low profitability firms) 66 per cent of the distribution of firms’ median profitability. Table 3.2 give the number of firms in each case for the two dummies. Unfortunately, as for most of the papers in the field, these kinds of selection may be endogenous.20 Some of the sample splitting criteria are likely to be correlated with both the firm-specific and time-invariant component of the error term, as well

CREDIT RATIONING VERSUS CONSOLIDATION

41

as with the idiosyncratic component. None the less, with regard to the short time period of estimation (3 years), it is close to the option of taking contemporaneous observations for splitting the sample, using two lags on endogenous variables as instruments (Schiantarelli [1996] seems to prefer this option than splitting observations with out of sample data). Estimation and results Our analysis is based on a simultaneous regression model of the Euler equation and of the flow of funds equation. All variables are scaled by total nominal assets Ai,t-1 in order to provide more comparable measures between small and large firms, and to reduce the impact of heteroskedasticity on our results. Similar results can also be obtained when scaling by total sales rather Table 3.2 Sample sizes Profitability Low High Total

Total 70 26 96

122 74 196

192 100 292

than total assets. Total nominal assets Ait is taken as a proxy for the capital stock.21 Cash-flows, retained earnings, debt and physical investment are taken from company accounts. The debt—equity ratio appearing in the Lagrange multiplier is approximated by its series development as a finite sum of powers of debt-capital ratios , as it was creating new outliers. The apparent interest rate is computed from interest charges divided by debt. The interest rate variable is usually not taken into account in Euler equation tests, but we think it is important to add this available information using the apparent interest rate for each firm, because it is compared to the marginal product of new investment in the consolidation behaviour. Endogenous variable enters into the combination of the flow of funds equation and the Euler equation to be tested. This econometric model is dynamic, in the sense that the set of explanatory variables include a lagged dependent variable (see Baltagi [1995] for strategies for dynamic panel data). The usual method for estimating random and fixed effect models does not work for dynamic models, unless one has very long time series. Anderson and Hsiao [1981] suggested to taking first differences and to using instruments for the differenced dependent variables. Another reason for including instrumental variables is that the model to be tested is based on rational expectations. Forecasting errors are due to information that was not available on the current date. Thus all variables in the present period must be orthogonal to future residuals. Therefore, the expectation of the error term multiplied by the instrumental variable is unchanged and usually zero. We use instrumental variables with the Generalized Method of Moments (GMM) procedure (Hansen [1982]).22 Before estimating the model, first differences are taken to eliminate the fixed effect fi. Instrumental variables should be correlated with endogenous explanatory variables, but uncorrelated with the error term. In so doing, we will achieve some extra orthogonality conditions, which may help to identify the parameters more effectively. Variables at least one period lagged are (theoretically) valid instruments (as first difference are taken, instruments two periods lagged are valid). Arellano and Bond [1991] suggested using all variables, with all possible lags beyond two to get consistent estimates for linear models. As we are limited by the time length of our data (6 years), lags in the Euler equation and first differences to be taken, we only use lagged three variables. As the model is non-linear, this suggests using powers and cross-products of variables. But one has to remain parsimonious when selecting instrumental variables for GMM estimations: having too many instruments leads the chi-squared test to over-reject the over-identifying restrictions of the model (Kocherlakota [1990]). A strategy is to avoid instruments too correlated with other instruments. We tried to take this into account in selecting among three period lags of variables as instruments, except for one variable which is not lagged in the model, for which its two period lag has been used. The structural parameters are estimated directly. They are the adjustment cost parameter α, the inverse of the mark-up rate μ, the agency cost for the lemon’s premium added in the credit interest rate and the parameters for bankruptcy costs determining the debt ceiling, such as the proportion of expected profits lost ζΠ and the proportion of the next period value of capital lost ζK. Year dummies are added. Other coefficients are parameterized: expected inflation πe by observed inflation, income tax rate by its statutory values, the depreciation rate at 8 per cent, equal to the adjustment cost parameter υ. The parameter ct for the ratio of the discounted expected growth factor of Tobin’s q if the firms were not constrained is set to 1.02. Results are presented in Table 3.3 (the number of observations for each variable is 876) testing with the Euler equation only (columns indexed by i=1) or testing simultaneously the Euler equation and the flow of funds (columns indexed by i=2). Columns (Ai) are related to the model with an increasing cost of borrowing as the degree of leverage increases, where the Lagrange multiplier for credit rationing is not taken into account Columns (Bi) test the credit rationing regime versus the increasing cost of borrowing only for the dummy equal to zero (for firms who exhibit the median of their profitability

42

JEAN-BERNARD CHATELAIN

between 1992 and 1994 below the median of the sample between 1992 and 1994). Columns (Ci) test both the credit rationing regime versus the increasing cost of borrowing only for the dummy equal to zero for firms who exhibit a rise (or fall) of debt between 1992 and 1994 below the median of the sample. Ten instruments have been used for model A: constant, year dummies for 92 and 93, lagged three variables Y/K–1, B/K, (B/ K)2, I/K–1, (I/K–1)2, iB/B, Table 3.3 Estimation results α (−0.14) μ (18.5) ф (0.57) ζK

A1

A2

B1

B2

C1

C2

−0.03 (−0.14) 0.94 (19.2) 0.26 (0.45)

0.04 (−0.30) 0.93 (29.50) 0.18 (0.93)

−0.001 (−0.65) 0.98 (31.5) 0.12 (1.03) 0.62 (0.49) 14.07 (1.47) −0.002 (−0.65) −0.019 (−2.02) 16.3 (6) 0.012

−0.005 (−0.77) 0.97 (32.98) 0.12 (1.01) 0.56 (3.05) 16.71 (0.98) −0.005 (1.15) −0.017 (3.23) 56.63 (19) 1.3. 10–5

−0.006 (−1.12) 1.011 (35.3) 0.14 (1.00) 1.27 (2.87) 3.01 (1.34) 0.007 (1.06) −0.017 (−4.15) 12.45 (6) 0.052

−0.01

(0.54) ζΠ d92 0.001 (0.18) (0.40) d93 −0.006 (−1.16) (−1.31) Sargan 13.86 (DF) (5) p-value 0.016 Note: T statistics in parentheses.

(1.26) 0.002 (−0.23) −0.006 (−2.21) 50.44 (15) 1.0. 10–5

1.006 0.13 1.13 4.18 0.007 −0.020 55.22 (19) 2.1. 10–5

RE/K−1 (RE is after tax retained earnings). For models B and C, three instruments were added: 1 CCF/K−1, 1CB/(K−B), 1CB/K, with CF for cash-flow and the respective dummy 1c used in each model (equal to 0 for model A). The figures in brackets are t statistics. The degrees of freedom (DF) for the Sargan statistics (which follows a X2) are the number of instruments times the number of equations minus the number of parameters to estimate.23 The X2 test of overidentifying restrictions gives the probability of satisfying the orthogonality conditions of the expectation error εit with the chosen instruments. The P-values are given below the degrees of freedom. The overidentifying restrictions are accepted at the 5 per cent level for model C1, while model A1 and B1 are both rejected. Adding the flow of funds equation for a simultaneous test with the Euler equation leads to a rejection of the models A2, B2 and C2, which may be due only to a too large number of instruments.24 Adding this equation does not alter the value of the parameters but increases their students’ t statistics. The parameter on the adjustment cost α is not significant and its value is zero. On the other hand, the inverse of the markup μ, is below one except for model C, and highly significant in all cases. These results are identical to the one obtained by Barran and Peeters [1998] on a sample of Belgium firms (model 2 of their table 3). The agency premium parameter Φ is not significant, but its value is plausible in models B and C (it has to be divided by 4 to give the interest rate premium for a firm having a debt/(total assets) of 50 per cent). It is higher than Estrada and Vallès’ [1995] results on a sample of Spanish firms (they found values close to 0.01). Model (A1) is exactly the model they tested, except on two points: first, we have chosen the individual apparent interest rate instead of the internal rate of return on public debt maturing over two years, second, we estimated it from a first order Taylor expansion of the discount rate, whereas they estimate the fractional form of the Euler equation. The credit rationing regime is related to bankruptcy cost parameters and the inclusion of the variation of cash-flow in the ‘adjusted’ discount rate. These parameters are those which should confirm the heterogeneity of investment-cash-flow sensitivities due to the difference between credit rationing and the regime with an increasing cost of leverage only. ζK is significant for models C, but its value exceeds 1. For model B, around 60 per cent of capital is expected to be lost in case of bankruptcy and is not taken as collateral. ζП is not significant and its estimates are too high. At least two reasons could explain this failure. First, the discounted stream of future profits could be taken as collateral, instead of just the profit in the next period. Second, we set the discounted growth factor of Tobin’s q, which enters into the explicit ‘adjusted discount rate’ for credit rationing, to 1.02, for all firms. Attempts to estimate it lead to high values of ζП (they are available on request). This

CREDIT RATIONING VERSUS CONSOLIDATION

43

parameter is related to the unobservable behaviour of the firm if it were not facing credit rationing. Finding a relevant firm specific proxy for this variable may improve the model. Conclusion This chapter proposed an explanation of the difference between investment— cash-flow sensitivities for low (or zero) dividend firms, presumably facing financial constraints. If firms face an endogenous debt ceiling which depends on the proportion of future profit taken as collateral, then their investment—cash-flow sensitivity may be higher than for other firms. An empirical test confirms this theory, using a new parameterization of the Lagrange multiplier, related to a binding debt ceiling. Two proxies were used for discriminating credit rationed firms from firms facing an increasing cost of leverage only: high profitability, or an increase in the debt ratio between the three years of estimation. They confirm credit rationing and the difference between investment—cash-flow sensitivities. None the less, further work remains to be done to improve the estimation of the lemons premium determining the increasing cost of leverage. Notes 1 The views presented in this chapter do not necessarily reflect those of the Banque de France. 2 This derivation follows Poterba and Summers [1985]. 3 Under an imputation relationship between corporate and personal taxes, the parameter θt=1/(1–st) where st is the rate of imputation. Under a classical relationship, θt is simply unity. 4 Jaramillo, Schiantarelli and Weiss [1996] consider a production function with no lag with respect to capital. 5 In our presentation, adjustment costs are priced at value added price (they are considered as a part of the production function), but an alternative where they are priced at investment price is also possible (adjustment costs are then considered as a part of the cost of investing). 6 A micro-foundation for these transaction charges can be found in Myers and Majluf ([1984]. 7 See Miller and Rock [1985] for a micro-foundation on dividend signalling. As noted by Hubbard, Kashyap and Whited [1995], this approach assumes that 8 GHKW’s assumption is the particular case: firms facing a debt ceiling simultaneously face a new share issues rationing. 9 For example, Yafeh and Yosha [1995] consider funds devoted to activities with scope for managerial moral hazard, such as advertising, research and development and entertainment expenses in Japanese firms’ balance sheets. 10 A rationale for a prohibitive cost of new share issues is noted by Kiyotaki and Moore [1997, p. 218] in their specific setting. They argue that issuing equity is not possible given the specific nature of an entrepreneur’s technology, and the fact that he can withdraw his labor: new holders of equity could not be assured that they would receive a dividend. Debt contracts secured on the entrepreneur’s capital are ‘the only financial instruments investors can rely on’ in this type of asymmetric information problem. 11 Regime 2 describes a very specific and rare regime where the upper limit on the debt ceiling determines an after tax lemon’s premium just equal to the rate of time preference. 12 A particular case holds when there is no agency premium: Whited [1992] and Hubbard, Kashyap and Whited [1995] made the assumption that ‘the equality between β and 1/(1+(1−τ) iit−πe) continues to hold in the presence of finance constraints. In other words, we assume external financial markets uphold this arbitrage condition between the returns on equity and debt, while financial constraints operate on the margin of the firm’s dividend payout.’ Modigliani—Miller theorem with taxation applies. Any combination of means of finance is indifferent for an individual firm for financing the long run capital stock. The square term of the debt—capital ratio do not show up in the Euler equation for capital. 13 Therefore, the dummy for low dividends in the Capital Euler equation test by Bond and Meghir [1994] seems loosely related to their structural model. 14 It can also be derived in solving the linear approximation near the equilibrium of the model and focusing on the saddlepath. 15 As mentioned by Chirinko ([1993], pp. 1902–1904), in GHKW Euler equation test, ‘the borrowing constraint is imposed exogenously and the endogenous variables that parameterize the Lagrange multiplier related to this constraint (such as cash flow and net worth) are not accounted for explicitly in specifying the econometric equation, thus blurring econometric interpretations of the statistical tests. It remains uncertain whether significant liquidity and net worth variables are capturing a structural element heretofore missing in the investment equation or are merely reflecting general misspecification.’ 16 Gertler, Hubbard and Kashyap [1990] and Hubbard and Kashyap [1992] assumed an alternative parameterization of the convex adjustment cost function, which adds a third parameter to estimate:

Hubbard, Kashyap and Whited [1995] assume:

44

JEAN-BERNARD CHATELAIN

The three estimated specifications differ with respect to the parameter interpretation of the discounted constant term. 17 The substitution of

amounts to the neglect of the covariance of the adjusted discount rate with date t+1 variables which are discounted in the Euler equation. Therefore, this covariance is subsumed into the error term. 18 The flow of funds is not exactly an accounting equation as it includes the adjustment cost function. It therefore includes structural parameters on technology (the adjustment costs parameters) which are always estimated in the neo-classical approach. In firms’ accounts, the flow of funds equation (the variation of total assets is equal to the variation of total liabilities) holds for a broad definition of investment as total assets (including inventories and investment in financial assets, net of sales of physical assets) and of debt (including trade credit) on the liabilities side. In accounting terms, the adjustment cost function amounts as modelling the variation of other assets than productive investment (net of the variation of liabilities omitted by the ‘flow of funds equation’) as a quadratic function of productive investment. 19 Many thanks to Elisabeth Kremp for providing access and extremely useful information on the companies account database of the ‘Observatoire des Entreprises’ at Banque de France, hints in SAS programming and for cleaning outliers. I thank as well JeanChristophe Teurlay for tips in SAS programming and numerous comments. 20 The covariance matrix is not corrected by the possible heteroscedasticity induced by the substitution of the selectivity term and its estimate. 21 The capital stock, constructed using the perpetual inventory method and used as a denominator for all variables will create serial correlation and will lower the quality of lagged variables as instruments. What is more, the perpetual inventory method is likely to provide poor proxies for six observations per firm. 22 The idea of the GMM method is to match theoretical moments to sample moments, and in so doing estimate the parameters. If more sample moments are used than theoretical moments (so called ‘identifying restrictions’), one cannot be sure to find parameters that match these extra moments. Therefore, one must use some weighting criterion for the restrictions. This method also allows for heteroscedasticity and serial correlation, although serial correlation restricts the set of legitimate instruments. We implemented the method by using proc model on SAS software 23 See Whited [1992] for a similar computation of degrees of freedom. 24 Proc model on the SAS software does not allow us to specify specific instruments for each equation. The flow of funds equation includes only one parameter to estimate: α. It may require less than 5 instruments.

References Anderson T. and Hsiao C. [1981]. ‘The Estimation of Dynamic Models with Error Components.’ Journal of the American Statistical Association 76, 598–606. Arrelano M. and Bond S. [1991]. ‘Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations.’ Review of Economic Studies 58, 277–297. Baltagi B.H. [1995]. Econometric Analysis of Panel Data. Chichester: John Wiley & Sons. Barran F. and Peeters M. [1998]. ‘Internal Finance and Corporate Investment: Belgian Evidence with Panel Data.’ Economic Modelling 15, 67–89. Bernanke B., Gertler M. and Gilchrist S. [1996]. ‘The Financial Accelerator and the Flight to Quality.’ The Review of Economics and Statistics 78, 1–15. Bond S. and Meghir C. [1994]. ‘Dynamic Investment Models and the Firm’s Financial Policy.’ Review of Economic Studies 61(2), 197–222. Chirinko R.S. [1993]. ‘Business Fixed Investment Spending: A Critical Survey of Modeling Strategies, Empirical Results, and Policy Implications.’ Journal of Economic Literature 31(4), 1875–1911. Estrada A. and Vallès J. [1995]. ‘Investment and Financial Costs: Spanish Evidence with Panel Data.’ Banco de España. Servicio de Estudios, Documento de Trabajo 9506. Fazzari S.M., Hubbard R.G. and Petersen B.C. [1988]. ‘Financing Constraint and Corporate Investment: Response to Kaplan and Zingales.’ NBER working paper 5462. Fazzari S.M., Hubbard R.G. and Petersen B.C. [1996]. ‘Financing Constraint and Corporate Investment.’ Brookings Papers on Economic Activity 1, 141–195. Gertler M., Hubbard G. and Kashyap A. [1990]. ‘Interest Rate Spreads, Credit Constraints, and Investment Fluctuations: An Empirical Investigation.’ NBER working paper 3495. Hansen L.P. [1982]. ‘Large Sample Properties of Generalized Method of moments Estimators.’ Econometrica 50, 1029–1054. Hart O. and Moore J.H. [1994]. ‘A Theory of Debt Based on the Inalienability of Human Capital.’ Quarterly Journal of Economics 109, 841–879. Hayashi F. [1982]. ‘Tobin’s Marginal q and Average q: a Neoclassical Interpretation.’ Econometrica 50(1), 213–224. Hubbard R.G. [1998]. ‘Capital Market Imperfections and Investment.’ Journal of Economic Literature 36, 193–225.

CREDIT RATIONING VERSUS CONSOLIDATION

45

Hubbard R.G. and Kashyap A.K. [1992]. ‘Internal Net Worth and the Investment Process: An Application of U.S. Agriculture.’ Journal of Political Economy 100, 506–534. Hubbard R.G., Kashyap A.K. and Whited T.M. [1995]. Internal Finance and Firm Investment. Journal of Money, Credit and Banking 27 (3), 683–701. Jaramillo F., Schiantarelli F. and Weiss A. [1996]. ‘Capital Market Imperfections before and after Financial Liberalization: An Euler Equation Approach to Panel Data for Ecuadorian firms.’ Journal of Development Economics 51, 367–386. Kaplan S.N. and Zingales L. [1997]. ‘Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints?’ Quarterly Journal of Economics 112, 169–216. Kiyotaki N. and Moore J.H. [1997]. ‘Credit Cycles.’ Journal of Political Economy 105(2), 211–248. Kocherlakota [1990]. ‘On Tests of Representative Consumer Asset Pricing Models.’ Journal of Monetary Economics 25, 285–304. Miller M.H. and Rock K. [1985]. ‘Dividend policy under Asymmetric Information.’ Journal of Finance 40(4), 1031–1051 Myers S.C. and Majluf N.S. [1984]. ‘Corporate Financing and Investment Decisions when Firms have Information that Investors do not Have.’ Journal of Financial Economics 13(2), 187–221. Poterba J.M. and Summers L.H. [1985] ‘The Economic Effect of Dividend Taxation.’In Recent Advances in Corporate Financeedited by E.I. Altman and M.G. Subrahmanyam, pp. 227–84. Homewood, IL: Richard D.Irwin. Schiantarelli F. [1996]. ‘Financial Constraints and Investment: Methodological Issues and International Evidence.’ Oxford Review of Economic Policy 12(2), 70–89. Schiantarelli F. and Georgoutsos D. [1990]. ‘Monopolistic Competition and the Q Theory of Investment’. European Economic Review 34 (5), 1061–78. Stokey N.L., Lucas R.E. and Prescott E.C. [1989]. Recursive Methods in Economic Dynamics. Cambridge, MA: Harvard University Press. Whited T.M. [1992]. ‘Debt, Liquidity Constraints and Corporate Investment: Evidence from Panel Data.’ Joumal of Finance 47(4), 1425–1460. Yafeh Y. and Yosha O. [1995]. ‘Large Shareholders and Banks: Who Monitors and How?’. CEPR, Discussion Paper Series.

4 Investment, uncertainty and industry structure1 Rina Bhattacharya and Paul Hope

Introduction Macroeconomists have long recognised the importance of the role of risk and uncertainty in influencing many (if not most) economic and financial decisions, particularly investment and asset portfolio decisions. As Fischer and Merton (1984) note, the common practice in macroeconomic modelling has been either to assume that investors are risk-neutral and to replace the economic variables assumed to be known with certainty in the formal model structure by their expected values, or alternatively to take account of the existence of risk and uncertainty by adding a constant risk premium to the base risk-free interest rate. At the theoretical level significant advances have been made in recent years in understanding the basic principles of the uncertainty-investment relationship. These relatively new theoretical models have some interesting policy implications. In particular, the literature on irreversible investment suggests that, if one of the government’s objectives were to stimulate investment over the short to medium term, then stability and credibility of macroeconomic policy may be more important than the actual level of policy variables such as tax rates or interest rates. This contrasts with the emphasis of the standard neoclassical models on the cost of capital (or relative factor costs) as the main factor influencing the level of investment, with uncertainty entering the investment decision via a risk premium in the cost of capital. Another interesting feature of the recent theoretical literature is the conclusion that the effect of uncertainty on investment is, in general, ambiguous and depends upon: • • • •

the degree of competition amongst firms in the industry; the extent of irreversibility of investment; the nature of the production technology in the industry; and firms’ attitude toward risk.

Hartman (1972) and Abel (1983) showed, somewhat counter-intuitively, that, under perfect competition, constant-returns-toscale production technology and risk neutrality, an increase in price uncertainty may actually increase investment. However, the combination of important degrees of imperfect competition and asymmetric adjustment costs may reverse the positive correlation between uncertainty and investment. What is crucial here is the role of imperfect competition, where the degree of imperfect competition is defined in terms of the size of the mark-up of price over marginal cost. The size of the mark-up is related in turn to the elasticity of demand for the firm’s product. Caballero (1991) further showed that decreasing returns to scale make a negative investment—uncertainty relationship more likely, and conversely increasing returns to scale make it less likely. Relaxing the assumption of risk neutrality and allowing for risk aversion makes a negative relationship between investment and uncertainty more likely (Craine 1989). These theoretical ambiguities suggest that the relationship between investment and uncertainty is ultimately an empirical issue, and is likely to vary with industry structure. However, empirical estimation of these models, and in particular models of irreversible investment, is riddled with difficulties. This is because, as Dixit and Pindyck (1994) point out, these models ‘do not describe investment per se, but rather the critical threshold required to trigger investment…The models tell us that if volatility increases, the threshold increases. Only to the extent that we can also describe (or make assumptions about) the distribution across firms of the values of potential projects, or of the marginal profitability of capital, can we also derive a structural model that relates volatility to actual investment’ (p. 422). Perhaps because of these difficulties there has been relatively little empirical work looking at the relationship between investment and uncertainty in the UK. Moreover, most existing UK studies have looked at aggregate manufacturing investment, and to the best of our knowledge none have looked in any detail at the issue of whether the effect of uncertainty on investment varies with industry structure. In this chapter we use panel data on a large number (103) of UK production industries over the period 1980 to 1992 to estimate a variety of investment equations and look empirically at the response of

RINA BHATTACHARYA AND PAUL HOPE

47

investment to uncertainty about demand. Given the difficulties of modelling the uncertainty-investment relationship explicitly, we follow the admittedly ad hoc approach of augmenting a conventional investment equation with an uncertainty term. The proxy for uncertainty that we use is sectoral level CBI survey data on demand uncertainty as a constraint on investment. We also use industry concentration ratios to look in more detail at the relationship between investment, uncertainty and industry structure. As far as we are aware this type of analysis has not been undertaken in the UK context. The chapter proceeds as follows. The next section examines the relationship at the theoretical level between investment and uncertainty. This is followed by a brief discussion of the data and of the methodological issues involved in estimating investment equations in a dynamic panel data context. We then present and discuss the empirical results, and a final section concludes. Uncertainty and investment In the standard neoclassical theory of investment uncertainty enters the investment decision as a risk premium in the cost of capital; that is, in the discount rate which firms use in making their net present value calculations. This suggests that the cost of capital measure may be correlated with our proxy variable for uncertainty. However, uncertainty can affect investment through two other channels without directly affecting the cost of capital. Hartman (1972) and Abel (1983) showed that uncertainty increases the expected return from an investment when the marginal revenue product of capital is convex in the uncertainty variable.2 For example, suppose that firms are uncertain about the future cost of capital. In particular, instead of knowing with certainty that future real interest rates will be 2 per cent, firms know that there is an equal probability that they will either be 1 per cent or 3 per cent. The expected future real interest rate is the same in both cases. But a 1 per cent interest rate increases discounted net benefits by more than a 3 per cent rate reduces them. So the expected return from investment is greater when interest rates are uncertain. If firms are risk-neutral this will, ceteris paribus, encourage them to invest more. However, with risk-aversion this will not necessarily be the case. Take the scenario where a 3 per cent real interest rate could lead to substantial losses, and even bankruptcy, for the firm. In such a situation greater uncertainty would make a risk-averse firm more reluctant to invest, even if the expected returns are higher. In another example Abel (1983) showed that, as long as the marginal revenue product of capital is a strictly convex function of the price of output, greater uncertainty about the future output price will tend to increase the expected marginal revenue product of capital. Constant-returns-to-scale and the substitutability of capital with other factors in the Abel-Hartman models ensure that the marginal profitability of capital is convex in output price and input costs. Intuitively, the assumed concavity of the production function means that the costs of having too low a capital stock during periods when the price-cost ratio is high outweigh the costs of having too high a capital stock when the price—cost ratio is low. Consequently, in this theoretical framework greater uncertainty about output price will increase the size of the optimal capital stock. Conversely, the expected return from investment falls when the marginal revenue product of capital is concave in the uncertainty variable. This argument (which is basically an application of Jensen’s inequality) can be illustrated as follows. Figure 4.1 plots the marginal revenue product of capital as a convex function of the uncertainty variable, say, output price. Suppose that the output price in the next period can take one of two values, P1 or P2. The expected price is then Pe, and the expected marginal revenue product of capital is given by η. Now suppose that the level of uncertainty increases, and the price in the next period can be either P1′ P2′. The expected price is unchanged, but the expected marginal revenue product of capital is now higher at η ′. Dixit and Pindyck (1994) showed that, if investment is at least partially irreversible, uncertainty drives a wedge between the required return from investment and the direct cost of investment, and thereby raises the required rate of return above the cost of capital. The intuition is that, because firms find it costly to disinvest, they face the possibility of holding ‘too much’ capital. This possibility in turn means that the required rate of return that justifies investment is higher than the (direct) cost of capital. The amount by which it is higher—the size of the wedge—depends on how likely it is that the firm would ex post like to disinvest, and on how costly it would be to do so. The size of the wedge reflects what is known in the literature as the ‘option value of waiting’. By waiting, a firm might be able to obtain useful information so that it can be more certain of the likely return from its prospective investment. The option of waiting is therefore a valuable one, and the theory of finance shows that options are worth more the greater the degree of uncertainty. Hence greater uncertainty (about costs, output prices or demand) increases the size of the wedge and thereby discourages investment in the short run. In the Dixit—Pindyck framework, for example, a one-time increase in volatility should reduce investment, at least temporarily, because project values that were above or close to what was a lower critical threshold are now below a higher one. Unfortunately, economic theory has, to date, very little to say regarding the effects of uncertainty on the long-run steady–state level of investment, the investment–output ratio or the capital-output ratio. To sum up, the theoretical literature suggests that uncertainty affects investment through three channels: 1 via a risk premium in the discount rate r;

48

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

Figure 4.1 The marginal revenue product of capital and uncertainty.

2 the Abel–Hartman argument and ‘Jensen’s inequality’: is the marginal revenue product of capital convex or concave in the uncertainty variable?; 3 the ‘option value of waiting’, and consequently the size of the wedge between expected and required rates of return, is positively related to the degree of uncertainty regarding the expected returns from the investment project. The Abel-Hartman framework assumes perfect competition, constant returns-to-scale and risk-neutrality. By contrast, models of irreversible investment (e.g. Bertola (1988); Pindyck (1988)) assume either imperfect competition or decreasing returns to scale or both. Caballero (1991) shows that when firms are perfectly competitive, and with constant returns-to-scale production technology, the Abel-Hartman argument holds no matter how asymmetric the adjustment costs.3 That is, investment today affects the future level of profits, but does not affect the marginal profitability of capital in the future. The issue of asymmetric adjustment costs is therefore irrelevant, since the marginal rate of return on capital does not vary with the level of the capital stock. Under these conditions the convexity of the marginal profitability of capital with respect to prices is the dominant factor in determining the sign of the investment-uncertainty relationship, and the Abel-Hartman result holds. However, the combination of significant degrees of imperfect competition and asymmetric adjustment costs may reverse the positive correlation between price uncertainty and investment. Caballero shows that what is crucial here is the role of imperfect competition, where the degree of imperfect competition is defined in terms of the size of the mark-up of price over marginal cost. The size of the mark-up in turn is related to the elasticity of demand for the firm’s product. Under these conditions the level of capital significantly affects the marginal profitability of capital. An increase in investment today makes it more likely that the firm will find its second-period capital too large relative to the desired capital stock. In this situation the symmetry of the investment cost function can become significant: the more asymmetric the adjustment costs, and the stronger the negative relationship between the marginal profitability of capital and the level of the capital stock, the more likely it is that greater uncertainty about price will impact negatively on investment. Caballero further shows that decreasing returns to scale make a negative investment–uncertainty relationship more likely, and conversely increasing returns to scale make it less likely. The argument is similar to the one just presented above: both a downward-sloping demand curve (i.e. demand less than infinitely elastic) and decreasing returns to scale imply that higher investment today lowers the marginal profitability of capital in the future. Consequently, the profit function’s convexity with respect to price declines and the Jensen inequality argument that implies higher investment under greater price uncertainty loses power, and indeed may be reversed. And the more risk-averse the firm, and the greater the sunk costs of investment, the more likely it is for the sign of the investment-price-uncertainty relationship to be negative. What about uncertainty about demand? Note that, under perfect competition, the issue of uncertainty about demand does not arise for any individual firm in the industry. This is because, by assumption, each firm in the industry can sell any amount it likes at the prevailing market price: the demand curve for each individual firm is horizontal. The only uncertainty (from the point of view of each individual firm) is the price at which it will be able to sell its output. This is not the case under imperfect competition, where each firm faces a downward-sloping demand curve (and consequently a downward-sloping marginal revenue curve). Nickell (1978) shows that, in general, it is not clear whether the marginal revenue product of capital

RINA BHATTACHARYA AND PAUL HOPE

49

is concave or convex in demand. Consequently, the imperfectly competitive firm’s response to an increase in demand uncertainty is not clear a priori, even if it is risk neutral. Nickell further shows that, with constant elasticity or linear or quadratic demand curves, the marginal revenue product of capital is concave in output (demand), implying that investment is inversely related to the level of uncertainty about demand. And, when we allow for risk-aversion on the part of the firm, the level of investment is a declining function of the degree of risk aversion, making a negative investment-demand uncertainty relationship more likely. In general under imperfect competition it is very difficult to predict on theoretical grounds the sign of the investment– uncertainty relationship. On the one hand uncertainty and irreversibility imply an option value of waiting for the reasons mentioned above, and therefore greater hesitancy in undertaking major investment expenditures. However, fear of preemption by a rival and the consequent need to act quickly play a more important role in oligopolistic industries. For example, by investing immediately a firm may be able to erect barriers to entry in a particular market. This may counteract the desire to wait. Hence, as Ghosal and Loungani (1996) note, without specifying the nature of strategic interaction among the oligopolists it is difficult to predict the sign of the investment—uncertainty relationship under imperfect competition. Take the case of investment in a new product or process technology. Tirole (1988) points out that patent laws provide obvious pre-emptive motives for investing in a new technology. At the same time, the possibility of imitation makes the innovation more of a public good and introduces reverse incentives (see Katz and Shapiro (1984)). The cheaper and quicker it is to imitate a new technology the greater the incentive for each firm in the industry to delay its own investment plans and wait until a rival firm has adopted the new technology. In other words, the more uncertain the returns from investing in a new technology, and the cheaper and quicker it is to imitate it, the greater is the ‘option value of waiting’. And if the cost of imitation/adoption is expected to fall over time the option value of waiting will be higher still. This implies a negative relationship between investment and uncertainty, particularly if patent laws are weak or non-existent and if the expected returns from investing are not very high if rival firms follow suit. On the other hand suppose that patent protection is strong, the cost of imitation is high, and that, through loss of competitiveness, the firm may incur heavy losses in the future by not investing—it may even become bankrupt. In this situation the relationship between investment and uncertainty is likely to be positive, particularly if the firm is risk-averse. Therefore it is not clear a priori whether the ‘imitation’ factor or the ‘preemption’ factor dominates. In short, the relationship between investment and uncertainty under imperfect competition is an empirical issue, to which we now turn. Methodological and data issues As mentioned in the introduction, empirical estimation of theoretical models of irreversible investment and uncertainty poses considerable problems. This is because these models do not describe investment per se, but rather the critical threshold required to trigger investment, and the latter is not directly observable. Only to the extent that we can also describe (or make assumptions about) the distribution across firms of the values of potential projects, or of the marginal profitability of capital, can we derive a structural model that relates volatility to actual investment. Pindyck and Solimano (1993) and Caballero and Pindyck (1996) attempt instead to estimate equations for the required rate of return of capital. Since the latter is unobservable, they obtain proxies for this variable by using extreme values of estimates of the marginal profitability of capital. However, as Eberly points out, the proxies for the threshold that they use ‘make it impossible to distinguish a structural relationship from a purely statistical one; this method cannot test the implications of the model unless a reliable proxy for the unobserved threshold can be identified’ (Eberly (1993), pp. 312). It is difficult to obtain good proxies for the (unobserved) investment threshold or required rate of return. Hence most studies (including this one) have instead adopted the admittedly ad hoc approach of augmenting a conventional investment equation with an uncertainty term and attempting to estimate it. Indeed, perhaps because of the difficulties mentioned above, there has been very little empirical work looking at the relationship between investment and uncertainty. In the context of the UK economy Driver and Moreton (1991, 1992) constructed series for inflation uncertainty and output uncertainty by taking the standard deviation of one-year ahead forecasts of the inflation rate and the GDP growth rate by twelve forecasting organisations. They went on to estimate error-correction models for plant and machinery investment in the manufacturing sector, and found that both growth uncertainty and inflation uncertainty have a negative and significant (at the 5 per cent level) impact on investment in the short run. However, only growth uncertainty has a significant negative impact on investment in the long run. Hurn and Wright (1994) estimated discrete-time hazard models of the time lag between the discovery of an oil field and the decision to develop the field. The uncertainty measure that they used was the variance of the oil price. They found that this had no significant effect on the appraisal duration of a field, or on the production start-up lag (the lag length between the decision to invest and the field coming on stream). Price (1995) looked at the effect of uncertainty on UK manufacturing investment. In this paper, uncertainty was measured as the conditional variance of (the log of) GDP using a GARCH estimator. Price found a significant and negative effect of

50

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

uncertainty on investment: on average, uncertainty reduced investment by about 5 per cent, but in particular periods the effect was much larger (as in 1974, when the estimated effect peaked at 48 per cent). More recently, Price (1996) set up a non-linear model in which investment was determined by the degree of capacity utilisation and the level of uncertainty in the economy. In this paper, a GARCH estimate of the conditional variance of manufacturing output was used as a proxy for uncertainty. Price estimated an error correction model for investment in which not just the long-run level of investment, but also the speed of adjustment towards equilibrium, was a function of the uncertainty variable. He found that uncertainty had a large and significant effect on both the rate of adjustment to, and the steady state level of, investment. Beaudry et al. (1996) adopted a different approach to test whether monetary instability reduces the informational content of market signals and thereby results in a less efficient allocation of investment. They developed a theoretical model to show that, as the money supply process becomes more predictable, the firm’s own relative price, and hence its profit opportunities, become easier to forecast. They went on to argue that this should cause the cross-sectional distribution of investment to widen, since better information allows firms to channel investment towards the most profitable opportunities. As support for this hypothesis they found that the cross-sectional variance of the investment—capital stock ratio across 963 quoted UK companies was inversely related with the conditional variance of inflation over the period 1972 to 1990. They also found a negative relationship between the cross-sectional variance of the investment rate and the conditional variance of money growth, but the relationship was not statistically significant. Also consistent with their hypothesis, they found a significant negative correlation between the variance of the investment rate and the variance of (the log of) the profit rate. Apart from Beaudry et al. (1996), all of these studies looked at aggregate manufacturing investment, and none have analysed how industry structure influences the effect of uncertainty on investment. However, as discussed above, the theoretical literature suggests that the impact of uncertainty on investment is likely to vary with the level of competition in the industry. This seems to us to be an important omission in the existing empirical literature on the investment—uncertainty relationship in the context of the UK economy. Turning now to the US economy, the only study that we are aware of that has looked empirically at this issue is Ghosal and Loungani (1996). In this chapter we follow their approach and use panel data on a large number (103) of UK production industries over the period 1980 to 1992 to estimate a number of investment equations. The equations that we estimate are of the form (1) (2) where ln(I/Y)i,t is the investment—output ratio for industry i at time t, Xi,t are the other explanatory variables in the investment equation, and εi,t is the error term which includes an industry-specific constant μj (a fixed effects term to capture unobservable industry-specific factors influencing the investment-output ratio). Ghosal and Loungani (1996), in their empirical study of US manufacturing industry, used as their dependent variable the ratio of gross industry investment scaled by the beginning-ofperiod capital stock. We chose instead to use the investment—output ratio as our dependent variable. This is in part due to the poor quality of the available data on the UK capital stock. Moreover, industry-specific capital stock data is only available at the 2-digit SIC 1980 level, whereas annual data on investment and output are available at the 3-digit SIC 1980 level from the Census of Production (see Appendix 4.1). Among the explanatory variables included in the investment equation are sector-specific measures of capacity utilisation and liquidity constraints and a proxy for demand uncertainty. An aggregate cost of capital measure is also included as a regressor. The capacity utilisation term is based on Question 4 of the CBI Industrial Trends Survey, which asks respondents ‘Is your present level of output below capacity (i.e. are you working below a satisfactory full rate of operation?)’.4 This measure is used to capture cyclical effects, which may differ across industries: some industries are more vulnerable to cyclical factors than others, as reflected in differing income elasticities of demand across products within manufacturing. Our liquidity constraint measures and our proxy for demand uncertainty are also based on the CBI Industrial Trends Survey. Question 16c of the survey asks: What factors are likely to limit (wholly or partly) your capital expenditure authorisations over the next twelve months? (i) inadequate net return on proposed investment; (ii) shortage of internal finance; (iii) inability to raise external finance; (iv) cost of finance; (v) uncertainty about demand; (vi) shortage of labour including managerial and technical staff;

RINA BHATTACHARYA AND PAUL HOPE

51

(vii other. ) Our industry-specific measure of CBI demand uncertainty is the percentage of firms in each industry responding positively to (v). One problem with this proxy is that it may reflect concern about low levels of demand rather than uncertainty about demand, but it is the only sector-specific measure of uncertainty that is available. Moreover, cyclical effects would presumably be captured by the capacity utilisation term. It also has the advantage that, being based directly on survey responses, it is not subject to the generated regressor problems which arise when fitted values from an estimated (and possibly misspecified) equation are used as an explanatory variable to estimate another equation. ‘Uncertainty’—whether about demand, or inflation, or future profits from investment—is not directly observable. Consequently, the CBI demand uncertainty measure that we use is only a proxy for uncertainty, and may be subject to significant measurement error. If this is the case then OLS estimates of the coefficient on the CBI demand uncertainty term will biased. If the uncertainty term were the only explanatory variable in the investment equation the direction of the bias would be unambiguously downwards; if other explanatory variables are included in the equation the direction of bias is more difficult to determine (see Levi (1973)). Because of this measurement error problem, lagged values of the uncertainty variables were used as instruments in the panel data estimation (even though it could be argued that ‘uncertainty’ is an exogenous variable as far as each individual industry is concerned). The standard neoclassical model of investment assumes that firms have access to unlimited sources of investment finance at an exogenously given cost. However, there is a considerable literature on capital market imperfections which suggests that there may be firms that are liquidity constrained. One reason for this is that informational asymmetries between lenders and investors can lead to both adverse selection and moral hazard problems (see Stiglitz and Weiss (1984)). This increases the relative cost of external finance and can result in credit rationing, with some firms unable to raise external finance even for financially viable investment projects. Consequently, the investment expenditure of these firms is likely to be constrained by the availability of internally generated funds, typically retained profits. We attempt to take this into account by including as an explanatory variable LILIQ—for each industry the proportion of CBI respondents to Question 16c reporting that their investment expenditure is constrained by a shortage of internal finance. We also experimented with an alternative measure of liquidity constraints, LIQ, by adding up the percentage of firms responding positively to 16c (ii) and (iii): that is, LIQ is equal to LILIQ plus the percentage balance of firms responding positively to 16 (iii). This might seem rather odd, since capital market imperfections imply that external funds—debt and new equity finance —are a more costly substitute for internally generated funds (typically retained earnings). Hence firms that have difficulty in raising external finance might also be expected to face a shortage of internally generated investment funds. A further complication is that firms are allowed to respond to both and so there could be some overlap between (ii) and (iii). However, as Woods (1995) points out, the fact that the extent of internal financial constraints is quantitively much larger suggests that the overlap issue is not very important. And in any case, both measures of liquidity follow similar patterns, at least at the aggregate level, and yield very similar empirical results. Also included among the explanatory variables is an aggregate cost of capital measure (see Appendix 4.1). It is debatable whether the cost of capital can be taken to be exogenous, at least in an aggregate investment equation. Assume that the cost of capital adjusts to equate the demand for and supply of investment funds, and that different sources of finance for investment are not perfect substitutes for each other. Then any variable (such as macro-economic uncertainty) that affects investment, and thereby the demand for investment funds, will also affect the cost of capital. In turn, this implies that the cost of capital term is likely to be correlated with the error term in an investment equation estimated using OLS. However, in the context of panel data estimation, it can be assumed that any change in the demand for investment funds by an individual industry will have only a marginal effect on the economy-wide demand for investment funds and consequently on real interest rates. Thus, the cost of capital may be taken to be exogenous to that industry. Moreover, even if the cost of capital is endogenous we can get around this problem by using instrumental variables. In what follows we present two sets of results depending on whether the cost of capital is assumed to be exogenous or endogenous. Empirical results One problem with estimating a dynamic model like equation (3) is that, since In (I/Y)i,t is a function of μi, In (I/Y)i,t−1 is also a function of μi (where μi is the industry-specific effect). Therefore In (I/Y)i,t−1,a right-hand regressor in the equation, is correlated with the error term. This in turn implies that OLS estimates of the parameters in equation (3) are biased and inconsistent, even if the υi,t are not serially correlated. To get consistent estimates equation (3) was estimated in first differences. For most of the explanatory variables a Generalised Method of Moments (GMM) estimator of the kind developed by Arellano and Bond (1991) was used to obtain efficient parameter estimates. However, with regard to the cost of capital and the demand uncertainty term, the DPD programme would not invert the matrices when GMM was applied to these variables.

52

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

This was probably due to collinearity in the instrument set when GMM is used. In particular, current levels of the cost of capital appear to be quite highly correlated with their own lagged values. Hence only the second and third lags of the levels of the cost of capital (when this was taken to be endogenous) and of the uncertainty variables were used as instruments; for all the other variables, GMM using levels lagged two to seven periods were used as the instrument set. All of the equations reported in this paper reject the Wald test that all the dependent variables are jointly insignificant. Furthermore, all of them pass the m2 test for no second-order serial correlation5 in the estimated equations, and also the Sargan test for validity of the instrument set. It is relevant to note, however, that the Sargan test6 is biased towards accepting the null hypothesis—that the instrument set is uncorrelated with the error term—because the estimated standard errors of the two-step estimates have a strong downward bias. Table 4.1 presents results using just the (log of the) level term in the uncertainty variable. All the equations give quite plausible parameter estimates. Moreover, the estimated coefficients do not vary much depending on whether the cost of capital is assumed to be exogenous or endogenous. The coefficient on the CBI demand uncertainty variable is negative and significant at the 5 per cent level when the cost of capital is taken to be exogenous, but becomes less significant when the cost of capital is taken to be endogenous. Two different approaches were used to see if the impact of uncertainty on investment varies with industry structure. HDUM is a dummy variable which takes a value of 1 for the nine industries in our sample with a concentration ratio greater than 80 and a value of 0 otherwise. Conversely, LDUM is a dummy variable which takes a value of 1 for the eleven industries in our sample with a concentration ratio less than 15 and a value of 0 otherwise. These dummy variables were multiplied by the (log of) the CBI demand uncertainty term to create two additional explanatory variables to add to our investment equation. The results are presented in Table 4.2. The estimated coefficients on the uncertainty term are higher than those reported in Table 4.1 and in all cases are statistically significant at the 1 per cent level. The coefficients on the other explanatory variables do not differ much from those in Table 4.1. However, the estimated coefficients on our dummy Table 4.1 Basic equation Independent variables

Eq. 1 COK exogenous

Eq. 2

Eq. 3 COK endogenous

Eq. 4

ln(I/Y)i,t−1 (6.346)** ln (COK)t−1 (−2.082)* ln(CAPU)i,t (4.297)** In (CAPU)i,t−1 (−4.420)** ln(LILIQ)i,t (−1.869) ln(LIQ)i,t

0.5487 (6.193)** −0.3764 (−2.170)* 0.2535 (4.134)** −0.2235 (−4.463)** −0.0684

0.5306 (6.528)** −0.3970 (−1.996)* 0.2448 (4.167)** −0.2352 (−4.1866)**

0.5626 (6.4072)** −0.3612 (−2.103)* 0.2486 (3.978)** −0.2050 (−4.266)** −0.0638

0.5474

(−1.785) −0.0953 −2.490)* −0.1451 −2.201)* 834 254.30

−0.3818 0.2376 −0.2121

−0.0893 (−2.381)* −0.1128 (−1.803) 834 242.32

ln (σCBIU)i,t−1 −0.1624 −0.1243 (−2.024)* (−1.667) No. of observations 834 834 Wald test of joint 295.62 288.83 significance (0.000) (0.000) (0.000) (0.000) m2 test for second order −0.346 −0.422 −0.290 −0.358 serial correlation (0.729) (0.673) (0.772) (0.721) Sargan test forvalidity of 102.85 103.24 102.99 102.98 instrument set (0.993) (0.993) (0.994) (0.994) Notes: All variables in first differences; results reported are one-step estimates with robust test statistics (except for the Sargan tests, which are based on two-step estimates). Dependent variable: ln(I/Y)i,t; sample period: 1984–1992 (103 industries). Figures in brackets are t-ratios on the estimated coefficients for the dependent variables, and p-values for the test statistics. * Significant at the 5 per cent level; ** Significant at the 1 per cent level.

RINA BHATTACHARYA AND PAUL HOPE

53

variable terms are always statistically insignificant, implying that there is no difference across industries in the response of investment to uncertainty. This is of course assuming that any variation in the response of investment to uncertainty would show up most clearly at the end points. These results contrast sharply with those found by Ghosal and Loungani (1996) for the US. They found that, for the US manufacturing sector as a whole, relative price uncertainty has no measurable impact on current investment. However, for industries with a high degree of competition the estimated impact is negative, reasonably large and statistically significant. For relatively non-competitive industries, by contrast, the impact is always small and not significantly different from zero. Their results are thus consistent with the argument that, at least in the US, industrial structure has an important influence on the response of investment to relative price uncertainty. Table 4.2 Impact of uncertainty Independent variables

Eq. 5 COK exogenous

Eq. 6

Eq. 7 COK endogenous

Eq. 8

ln (I/Y)i,t−1 (6.124)** ln (COK)t−1 (−2.001)* ln (CAPU)i,t (4.149)** ln (CAPU)i,t−1 (−4.878)** ln (LILIQ)i,t (−1.664) ln (LIQ)i,t

0.5474 (6.028)** −0.3708 (−2.127)* 0.2502 (3.959)** −0.2290 (−4.887)** −0.0633

0.5306 (6.277)** −0.3947 (−1.915) 0.2410 (4.035)** −0.2396 (−4.767)**

0.5593 (6.216)** −0.3562 (−2.048)* 0.2458 (3.826)** −0.2133 (−4.807)** −0.0582

0.5457

ln (σCBIU)i,t−1 (−2.995)** HDUM* ln (σCBIU)i,t−1 (1.180) LDUM* ln (σCBIU)i,t−1 (−0.166) No. of observations Wald test: joint significance m2 test: second order serial correlation

(−1.569) −0.0885 (−2.144)* −0.1857 (−3.025)** 0.2903 (0.963) −0.0636 (−0.171) 834 362.61 (0.000) −0.332

−0.3794 0.2344 −0.2192

−0.0817

−0.1970 (−2.760)** 0.2496 (1.502) −0.0659 (−0.123) 834 423.53

(−2.027)* −0.1670 (−2.742)** 0.3340 (1.266) −0.0461 (−0.144) 834 337.53

834 399.10

(0.000) −0.410

(0.000) −0.298

(0.000) −0.369

−0.1721 0.2960 −0.0541

(0.740) (0.682) (0.766) (0.712) Sargan test: validity of 103.11 103.49 102.26 102.88 instrument set (0.995) (0.995) (0.997) (0.996) Notes: All variables in first differences; results reported are one-step estimates with robust test statistics (except for the Sargan tests, which are based on two-step estimates). Dependent variable: ln (I/Y)i,t; sample period: 1984–1992 (103 industries). Figures in brackets are t-ratios on the estimated coefficients for the dependent variables, and p-values for the test statistics. * Significant at the 5 per cent level; * Significant at the 1 per cent level.

Theory tells us little about exactly how the elasticity of investment to uncertainty should vary with industry structure—that is, the functional form that it should take. Hence, in our second approach to looking at how the impact of uncertainty on investment varies with industry structure, we experimented with different functional forms. The results presented in Table 4.3 look at the case where the elasticity of investment to demand uncertainty is linear in the (log of the) concentration ratio, while the results in Table 4.4 look at the case where the elasticity is quadratic in the (log of the) concentration ratio.7 Once again all of the estimated equations give quite sensible parameter estimates. The results in Table 4.3 show that both the level and linear terms are statistically significant and suggest a negative

54

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

Table 4.3 Impact of uncertainty Independent variables

Eq. 9 COK exogenous

Eq. 10

Eq. 11 COK endogenous

Eq. 12

ln(I/Y)i,t−1 (5.395)** ln(COK)t−1 (−1.978)* ln(CAPU)i,t (4.257)** ln(CAFU)i,t−1 (−4.352)** ln(LILIQ)i,t (−2.483)* ln(LIQ)i,t

0.4523 (5.340)** −0.3783 (−2.030)* 0.2420 (4.191)** −0.1969 (−4.407)** −0.0801

0.4409 (5.440)** −0.3922 (−1.901) 0.2371 (4.132)** −0.2081 (−4.257)**

0.4606 (5.422)** −0.3664 (−1.966)* 0.2375 (4.042)** −0.1810 (−4.373)** −0.0767

0.4517

(−2.410)* −0.1039

−0.3802 0.2305 −0.1881

−0.0993

(−3.070)** −1.0934 (−4.426)** 0.2539 (3.892)** 834 224.30

(−2.973)** −1.0930 (−4.416)** 0.2607 (4.049)** 834 221.27

ln(σCBIU)i,t−1 −1.0530 −1.0507 (−4.424)** (−4.424)** ln(CR)i,t−1** 0.2407 0.2483 ln(σCBIU)i,t−1 (3.812)** (3.992)** No. of observations 834 834 Wald test: joint 261.86 258.80 significance (0.000) (0.000) (0.000) (0.000) m2 test: second order −0.163 −0.222 −0.133 −0.186 serial correlation (0.871) (0.825) (0.895) (0.852) Sargan test: validity of 102.71 102.88 102.81 103.01 instrument set (0.995) (0.994) (0.995) (0.995) Notes: All variables in first differences; results reported are one-step estimates with robust test statistics (except for the Sargan tests, which are based on two-step estimates). Dependent variable: ln(I/Y)i,t; sample period: 1984–1992 (103 industries). Figures in brackets are t-ratios on the estimated coefficients for the dependent variables, and p-values for the test statistics. * Significant at the 5 per cent level; ** Significant at the 1 per cent level.

elasticity of investment to uncertainty up to a concentration ratio of around 80, and a positive elasticity thereafter. Given the distribution of the concentration ratio across industries (see Table 4.6) this implies a negative elasticity for over 90 per cent of the industries in our sample. Along similar lines, the estimated equations in Table 4.4 imply a negative elasticity of investment to the CBI demand uncertainty measure for industries with a concentration ratio of less than about 70, and a positive elasticity thereafter. That is, the estimated elasticity is negative for about 90 of our 103 industries. The results also imply that the elasticity of investment to uncertainty reaches a minimum value at a concentration ratio of around 8 or 9. Note also that results with the quadratic functional form suggest that the elasticity of investment to uncertainty follows a U-shaped pattern in the (log of the) concentration ratio, but there are no observations on the Table 4.4 Impact of uncertainty Independent variables Eq. 13 COK exogenous

Eq. 14

Eq. 15

Eq. 16 COK endogenous

Eq. 17

ln(I/Y)i,t−1 (5.578)** ln(COK)t−1 (−2.132)* ln(GAPU)i,t (4.158)** ln(CAPU)i,t−1

0.4644 (5.528)** −0.3935 (−2.164)* 0.2319 (4.064)** −0.1998

0.4516 (5.678)** −0.4065 (−2.051)* 0.2268 (3.998)** −0.2105

0.4737 (5.626)** −0.3828 (−2.111)* 0.2260 (3.910)** −0.1825

0.4629

0.4731 (5.612)** −0.3968 (−2.119)* 0.2331 (4.126)** −0.2029

−0.3959 0.2189 −0.1896

RINA BHATTACHARYA AND PAUL HOPE

Independent variables Eq. 13 COK exogenous

Eq. 14

Eq. 15

Eq. 16 COK endogenous

(−4.619)** ln(LILIQ)i,t (−2.613)** ln(LIQ)i,t

(−4.611)** −0.0848

(−4.469)**

(−4.595)** −0.0815

(−4.545)** −0.0835 (−2.664)**

55

Eq. 17

(−2.581)** −0.1079

−0.1035

(−3.261)** (−3.150)** ln(σCBIU)i,t−1 0.7477 (1.269) ln(CR)it−1* −0.8059 −0.3847 −0.3808 −0.3767 −0.3714 CBIU ln(σ )i,t−1 (−2.411)* (−5.090)** (−5.174)** (−4.984)** (−5.050)** (ln (CR)i,t−1)2* 0.1471 0.0894 0.0877 0.0892 0.0874 ln(σCBIU)i,t−1 (3.128)** (4.887)** (4.959)** (4.886)** (4.952)** No. of observations 834 834 834 834 834 Wald test: joint 244.02 239.52 272.82 235.62 269.74 significance (0.000) (0.000) (0.000) (0.000) (0.000) m2 test: second order −0.102 −0.125 −0.193 −0.095 −0.159 serial correlation (0.918) (0.901) (0.847) (0.924) (0.873) Sargan test: validity 102.74 103.38 103.29 103.28 102.00 of instrument set (0.995) (0.996) (0.996) (0.996) (0.997) Notes: All variables in first differences; results reported are one-step estimates with robust test statistics (except for the Sargan tests, which are based on two-step estimates). Dependent variable: ln(I/Y)i,t; sample period: 1984–1992 (103 industries). Figures in brackets are t-ratios on the estimated coefficients for the dependent variables, and p-values for the test statistics. * Significant at the 5 per cent level; ** Significant at the 1 per cent level.

downward-sloping segment on the U-curve—that is, there are no industries in our sample with a concentration ratio of less than 10. Thus both sets of results—with the linear and quadratic functional forms—suggest that the (negative) response of investment to demand uncertainty is highest (in absolute value) for the least concentrated industries. What can explain our empirical result that the response of investment to uncertainty is negative at low concentration ratios, but becomes positive at high concentration ratios? It was mentioned earlier that Caballero (1991) has shown that decreasing returns to scale make a negative relationship between uncertainty and investment more likely, whereas increasing returns Table 4.5 Decomposition of percentage change in the investment-output ratio, 1984–1992 average Annual average growth rate Dependent variable: Δln(I/Y)i,t Independent variables Δln(I/Y)i,t−1 Δln(COK)t−1 Δln(CAPU)i,t Δln(CAPU)i,t−1 Δln(LILIQ)i,t Δln(σCBIU)i,t−1 Δ[ln(CR)i,t−1* ln(σCBIU)t,-i] Δ[(ln(CR)it−1)2* ln(σCBIU)i,t-1]

Eq. 1

Eq. 9

Eq. 14

−3.416 −0.140 0.124 −1.343 −0.140 −0.021

−2.816 −0.141 0.119 −1.183 −0.164 −0.157 −0.753

−2.891 −0.147 0.114 −1.200 −0.174

−7.013 −6.226 0.373 0.491 6.007 2.051 0.144 −2.965 −21.489

−4.936

−0.910 −5.096

1.141 −1.921 −0.780 −5.079

56

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

Table 4.6 Distribution of concentration ratios, 1980–1992 average Concentration ratio

No. of industries

0–0.99 10–19.99 20–29.99 30–39.99 40–49.99 50–59.99 60–69.99 70–79.99 80–89.99 90–100 Total

0 14 21 17 14 13 11 4 6 3 103

to scale offset imperfect competition, bringing the uncertainty—investment relationship closer to the results of Abel and Hartman. It is plausible that decreasing returns to scale are more likely in less concentrated sectors, and increasing returns to scale more likely in highly concentrated industries. This could provide a (partial) explanation for the empirical relationship found between investment, demand uncertainty and industry structure. Over our estimation period (1984–1992) the investment-output ratio fell, on average, by around 7 per cent. We can use our results to calculate how much of this was due to uncertainty. Column 2 of Table 4.5 sets out the average growth rates of the relevant variables for the panel dataset over this period. It shows that CBI demand uncertainty grew on average by around 0. 14 per cent per annum. Over the same period the average concentration ratio fell slightly, from 43.2 in 1984 to 42.3 in 1992. To evaluate the relative contributions of the different explanatory variables towards the fall in the investment-output ratio we multiplied their average growth rates over this period by the estimated coefficients in equations 1, 9 and 14 respectively. The results are shown in Columns 3, 4 and 5 of Table 4.5. They suggest that our estimated equations can explain about 5 of the 7 percentage point fall in the investment-output ratio over this period. The results with just the (log of the) level term of the CBI demand uncertainty variable suggest that the increase in uncertainty over this period made only a marginal contribution towards this fall, just slightly over 0.02 percentage points. However, our results with the linear and quadratic functional forms suggest that the increase in demand uncertainty, together with the fall in the concentration ratio, contributed between 0.75 to 1. 00 percentage points to the 7 percentage point fall in the investment-output ratio between 1984 and 1992. Conclusion Theory is ambiguous about the impact of uncertainty on investment. The degree of competition in the industry, the extent of irreversibility of investment and firms’ attitude toward risk all play a part in determining the sign of the investmentuncertainty relationship. Moreover, theory has little concrete to say about how the impact of uncertainty on investment should vary with industry structure. However, Caballero (1991) has shown that decreasing returns to scale make a negative relationship between uncertainty and investment more likely, whereas increasing returns to scale offset imperfect competition, bringing the uncertainty-investment relationship closer to the results of Abel and Hartman. But, ultimately, the relationship between investment, uncertainty and industry structure is an empirical issue. Using CBI data at the sectoral level, the results presented in this chapter suggest that demand uncertainty has, on balance, a negative impact on investment for UK production industries as a whole. Our results also suggest that the elasticity of investment to demand uncertainty is negative for most industries, but becomes positive at very high concentration ratios. It is plausible that decreasing returns to scale are more likely in less concentrated sectors, and increasing returns to scale more likely in highly concentrated industries. This could provide a (partial) explanation for the empirical relationship found between investment, demand uncertainty and industry structure. Finally, our results looking at the cases where the elasticity of investment to uncertainty is linear or quadratic in the (log of the) concentration ratio suggest that the increase in demand uncertainty, when combined with the effects of the fall in the average concentration ratio across industries, contributed between 0.75 to 1.00 percentage points to the 7 percentage point fall in the investment—output ratio between 1984 and 1992. List of variables I/Y:

Industry-specific investment-output ratio.

RINA BHATTACHARYA AND PAUL HOPE

COK: CAPU: LILIQ: LIQ: σCBIU: CR: HDUM:

LDUM:

57

Aggregate cost of capital measure (see Appendix 4.1). Sector-specific measure of capacity utilisation based on the CBI Industrial Trends Survey. An increase in CAPU implies an increase in capacity utilisation. Balance of firms facing internal liquidity constraints on investment expenditure. Sector-specific measure based on the CBI Industrial Trends Survey. LILIQ plus balance of firms facing external liquidity constraints on investment expenditure. Sector-specific measure based on the CBI Industrial Trends Survey. CBI sector-specific measure of demand uncertainty as a constraint on investment (see Appendix 4.1). Concentration ratio. Dummy variable=1 for (9) high concentration industries with a concentration ratio >80 =0 otherwise. Dummy variable=1 for (11) low concentration industries with a concentration ratio<15 =0 otherwise. Appendix 4.1: data Investment, output and industry concentration data

We compiled annual data on investment, output and industry concentration at the 3-digit SIC 1980 level from the Census of Production. Our data set covers 103 industries (listed in Table 4.3) over the period 1980–1992. Output and investment were measured gross8 in nominal terms. The output figures were deflated using producer output prices. Unfortunately, for confidentiality reasons, output prices were not available at the 3-digit level for all industries in all years. Wherever possible, we used 2-digit price data instead, but for some industries we were unable to obtain a complete time series. Nominal investment was deflated using aggregate price series for the extraction and manufacturing industries respectively, derived from tables 13.6 and 13.7 in the Blue Book. Industry concentration was defined as the proportion of total sales and work done accounted for by the five largest enterprises in the industry. CBI data Sectoral level data on capacity utilisation and liquidity were drawn from the CBI’s quarterly Industrial Trends Survey. The capacity series used was the proportion of positive responses to the question ‘Is your present level of output below capacity?’9 For liquidity, we used the proportion of firms reporting a ‘shortage of internal finance’ and ‘inability to raise external finance’ as factors ‘likely to limit capital expenditure authorisations over the next twelve months’ (question 16c). Another of the factors listed as a constraint on investment is ‘uncertainty about demand’, and we used responses to this as one of our measures of uncertainty. Data in the Industrial Trends Survey are at a level of aggregation similar to those in the Census of Production, but they do not match exactly. Moreover, the CBI data were collected on a SIC 1968 basis until January 1984; from April 1984 they were collected on a SIC 1980 basis. We matched the CBI data to the census data as closely as possible, taking averages to produce annual series. The cost of capital measure The cost of capital measure that we use is based on Jorgenson’s (1963) real user cost of capital measure and is calculated as

where A=present discounted value of capital allowances associated with £1 spending on capital goods τ=rate of mainstream corporation tax on undistributed profits Pk=price of capital goods = price deflator for investment expenditure (excluding housing) Py= Producer Output Price Index (excluding food, drink and tobacco) for manufacturing

58

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

=long-term interest rate =short-term interest rate WL=Market value of ICCs’ bond liabilities as a share of market value of ICCs’ bond liabilities+ICCs’ bank borrowing WS=1−WL ECGt=expected capital gains. A backward-looking measure was used, given by ECGt=[(1−At)Pk,t−(1−At−1)Pk,t−1]/(1−At−1)Pk,t−1 δ=depreciation rate Table 4.7 Industries in sample 221000 222000 223000 224000 231000 233000 241000 242000 243000 244000 245000 246000 247000 248000 251000 255000 256000 257000 258000 259000 260000 311000 312000 313000 314000 316000 320001 321000 322000 323000 324000 325000 326000 327000 328000 329000 330001 341000 342000 343000 344000 345000

Iron and steel industry Steel tubes Steel drawing, cold rolling and cold forming Non-ferrous metals industry Extraction of stone, clay, sand and gravel Salt extraction and refining Structural clay products Cement, lime and plaster Concrete, cement or plaster products Asbestos goods Working of stone and other non-metallic minerals, NES Abrasive products Glass and glassware Refractory and ceramic goods Basic industrial chemicals Paint, varnish and printing ink Special industrial chemicals and agricultural chemicals Pharmaceutical products Soap and toilet preparations Household and office chemicals Production of man-made fibres Foundries Forging, pressing and stamping Bolts, nuts, springs, chains, metal treatment Metal doors, windows, etc. Handtools and finished metal goods Industrial plant and steelwork Agric. machinery and tractors Metal-working machine tools and engineers’ tools Textile machinery Food, chemical, etc. machinery; process eng. contractors Mining machinery, construction and mechanical handling eqpt Mechanical power transmission equipment Misc. industrial machinery Other machinery and mechanical equipment Ordnance, small arms and ammunition Office machinery and data processing equipment Insulated wires and cables Basic electrical equipment Industrial electrical equipment, batteries and accumulators Electronic components, capital goods, telecoms. eqpt, etc. Other electronic equipment

RINA BHATTACHARYA AND PAUL HOPE

346000 347000 351000 352000 353000 361000 362000 363000 364000 365000

Domestic electric appliances Electric lamps and lighting eqpt Motor vehicles and engines Vehicle bodies, trailers and caravans Motor vehicle parts Shipbuilding and repairing Railway and tramway vehicles Cycles and motor cycles Aerospace eqpt manufacturing and repairing Other vehicles

371000 372000 373000 374000 411000 412000 413000 414700 415000 416000 419000 420000 421000 422000 423900 424000 426100 427000 428300 429000 431000 432000 433600 434000 435000 436000 437000 438000 439000 441000 442000 451000 453000 455000 456000 461000 462000 463000 464000 465000

Measuring, checking and precision instruments and apparatus Medical and surgical equipment and orthopaedic appliances Optical precision instruments and photographic equipment Clocks and other timing devices Organic oils and fats other than crude animal fats Animal slaughtering, meat and animal by-products Milk and milk products Fruit and vegetables processing Fish processing Grain milling Bread, biscuits and flour confectionery Sugar and sugar by-products Ice cream, cocoa, chocolate and sugar confectionery Animal feeding stuffs Miscellaneous foods Spirit distilling and compounding Wines, cider and sherry Brewing and malting Soft drinks Tobacco industry Woollen and worsted industry Cotton and silk industries Throwing, texturing, etc. of continuous filament yarn Flax, etc. spinning and weaving Jute and polypropylene yarns and fabrics Hosiery and other knitted goods Textile finishing Carpets etc. Miscellaneous textiles Leather (tanning and dressing) and fellmongery Leather goods Footwear Clothing, hats and gloves Household and made-up textiles Fur goods Sawmilling, planing, etc. of wood Wood semi-manufactures and treatment Builders’ carpentry and joinery Wooden containers Other wooden articles (except furniture)

59

60

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

466000 467000 471000 472000 475000 481000 483000 491000 492000 493000 494000 495000

Cork etc.; brushes and brooms Wooden and upholstered furniture and shop and office fittings Pulp, paper and board Conversion of paper and board Printing and publishing Rubber products Plastics processing Jewellery and coins Musical instruments Photographic laboratories Toys and sports goods Misc. manufacturing industries

Kelly and Owen (1985) approximate straightline depreciation by a proportional rate defined by the asset life and show that δ≈2/L, where L is the (assumed) asset life: Wp, Wb and Wp are the weights for plant and machinery, buildings and vehicles in the production sector. These are 0.80, 0.15 and 0.05 respectively. Lp=assumed asset life for plant and machinery. Lb=assumed asset life for commercial and industrial buildings. Lv=assumed asset life for vehicles. Asset lives

1979Q1–1981Q4 1982Q1+

Lp

Lb

Lv

20 17

60 60

10 10

The reduction in the assumed asset life of plant and machinery is a proxy for premature scrapping following oil shocks and rapid changes in computing technology. Appendix 4.2: the Dixit—Pindyck model The ideas presented in the second part of the chapter can be illustrated more formally using the model of irreversible investment under diminishing returns to capital set out in chapter 11 of Dixit and Pindyck (1994). We assume that each unit of capital costs K, and investment in the unit is irreversible. The firm faces uncertain demand given by P=Y.D(Q) where Q is output, P is price and Y is a shift variable that follows a geometric Brownian motion given by (1) where dz is the increment of a Weiner process, α is the drift parameter and σ is the variance parameter. Thus the change in Y (denoted by ΔY) over any time interval Δt is normally distributed and has an expected value and variance The firm’s optimisation problem is to choose the path of its capital stock Kt in the face of uncertain demand so as to maximise the expected present value of its operating profits, net of the cost of investing. Given a production function G(K), and assuming for simplicity that there are no variable costs, the firm has a profit flow π given by (2) and the marginal revenue product of capital is given by YHK(K). We assume diminishing returns to capital so that the marginal revenue product is decreasing in K, or the revenue function is concave in K, which means that HKK(K)<0. This could reflect diminishing physical returns to capital [GKK(K)<0], or a downward-sloping demand curve [DQ(Q)<0] or some combination of the two. Setting up the problem as a dynamic program where ρ is the rate at which future profits are discounted and ε is the expectations operator, the maximised value function W(K,Y) is defined by the Bellman equation (3)

RINA BHATTACHARYA AND PAUL HOPE

61

Dixit and Pindyck show that, since we have assumed that H(K) is concave in K, the Bellman function W(K,Y) must also be concave in K. We now go on to analyse the relationship between W(K,Y) and the shift variable Y. Expanding the right-hand side of (3) by Ito’s lemma and keeping terms to order dt, we find that W(K,Y) satisfies the second order differential equation (4) which Dixit and Pindyck call the fundamental quadratic equation. The general solution to (4) is given by (5) where β1 and β2 are, respectively, the positive and negative roots of the fundamental quadratic (6) and B1 and B2 are the constants of integration (to be determined). Optimality requires that W(K,Y) satisfies the following boundary conditions: (7) (8) (9) where (as mentioned earlier) κ is the price of a unit of capital. Condition (7) is required to keep W(K,Y) finite as the limit of Y goes to zero (that is, the value of the firm will be zero when Y=0). To ensure this we must leave out the negative power of Y in the solution by setting B2(K)=0. Equation (8) is the valuematching condition, which implies that the value of a change in the capital stock equals the cost. Equation (9) is the smoothpasting condition from dynamic programming. Equation (5) now becomes (5′) Note that W(K,Y) is non-linear in Y and is thereby consistent with risk aversion and the existence of a risk premium (since (5′) does not preclude W(K,Y) being concave in the uncertainty variable Y). The value-matching and smooth-pasting conditions can be used to solve for the constant of integration B1. In equation (5′) the term YH(K)/(ρ−α) is the expected present discounted value of the profits the firm would get if it kept its capital stock constant at the initial level K0 forever. The other term is the current value of the firm’s future option to expand capacity. The solution is characterized by the property that a critical threshold for Y exists, such that if Y exceeds this value, the firm will decide to go ahead and invest. This threshold is defined as (10) where (11) and δ=ρ–α. Equation (10) can be interpreted as follows. Suppose we install a marginal unit of capital dK when the existing stock is K0 and the stochastic shock has the value Y. The contribution of this marginal unit to the profit flow is Y Hk(K) dK. Since Y is expected to grow at the rate α, and future profits are discounted at the rate ρ, the expected present value of this contribution is given by . The cost of installing the marginal unit is κ dK. Then equation (10) says that the marginal addition to the firm’s capital stock is justified when the expected present value exceeds the cost by the multiple β1/(β1−1), reflecting the option value of waiting. In other words the expected present value of the project’s marginal contribution must be a multiple β1/(β1−1) of the cost of investment where β1 is the usual root of the fun-damental quadratic that involves the discount rate and the drift and volatility coefficients of the price process. The value-matching condition can be written as (8′) where

denotes the derivative of B1(K) with respect to K.

This condition can be interpreted as saying that, at the threshold that justifies the incremental investment dK, its expected contribution to the capitalised profit flow YHĸ(K)/(ρ−α) should equal the direct cost of investment K plus the opportunity cost of the option to wait

62

INVESTMENT, UNCERTAINTY AND INDUSTRY STRUCTURE

So how does uncertainty affect the investment threshold? The most obvious channel is through the discount rate ρ. In the capital asset pricing model (CAPM) from the theory of finance the risk-adjusted discount rate ρ is modelled as (12) where r is the risk-free interest rate, Φ is the market price of risk, and ΓYM is the correlation coefficient between the asset that tracks Y perfectly and the market portfolio. Here it is assumed that stochastic changes in Y are spanned by existing assets in the economy: capital markets are sufficiently ‘complete’ so that (at least in principle) one could find an asset or construct a dynamic portfolio of assets (a portfolio whose holdings are adjusted continuously as asset prices change), the price of which is perfectly correlated with Y. The effect of an increase in α depends on what else is held constant. Note that any marginal contribution to the firm’s profit flow is discounted at the rate δ=ρ−α. If r and α are fundamental exogenous constants, then an increase in α must increase ρ and therefore δ. On the other hand, if r and δ are basic constants, then α must adjust to offset the effect of a higher σ on ρ. In the first case the threshold will be affected, in the second case it will remain unchanged. In the Dixit-Pindyck framework, the second channel through which uncertainty affects investment is via the wedge term β1/ (β1−1). If σ increases, it can be shown that the root β1 decreases and the option value multiple (the wedge) increases. This raises the threshold of Y for any given K. In this sense greater uncertainty serves to discourage investment. As Dixit and Pindyck (1994) point out, it is important to be clear about exactly what this means. It says merely that some values of Y that would have justified a given increment of capacity for a low value of σ may be insufficient for a higher σ. If uncertainty rises, for a particular firm prices must rise more than hitherto to trigger a decision. But as the variance of the price has increased, any given value will be exceeded more often, so investment may occur at more frequent intervals. To find the overall impact on investment of increasing uncertainty, the effect of raising the threshold has to be balanced against the effect of raising the volatility of movements in Y. As Price (1995) points out, aggregate investment during any discrete interval may or may not increase, but the dynamics of the process will become more lumpy. There is a third effect of uncertainty that does not appear in our formulation so far, because we have made the profit flow function linear in Y. More generally, we could have stipulated a profit function . The marginal effect of capacity expansion on the profit flow is therefore given by . If the marginal profit flow is a convex function of γ, greater uncertainty raises the expected marginal returns from investment (the Abel—Hartman argument) and lowers the threshold value of Y, implying that an increase in uncertainty encourages investment. In terms of the value-matching condition the expression in equation (8′) will be replaced by a term which is a convex (instead of a linear) function of Y. Following on from this it can be shown that greater uncertainty raises the discounted present value of the marginal profit flow from a given investment project and so implies a lower threshold value of Y (for any given capital stock K). Notes 1 We thank Mark Cornelius, Nigel Jenkinson, James Proudman, Stephen Redding and Tony Yates for helpful comments and advice on earlier drafts of this chapter. Special thanks to Ian Small, both for his comments and for his assistance with the computer programmes to carry out the empirical analysis. Any views expressed in this chapter are solely those of the authors. 2 In turn, this also reduces the threshold which triggers investment, as we explain below. 3 Here asymmetric adjustment costs refer to the case where it is more expensive to adjust the capital stock downwards than to adjust it upwards. Irreversibility is a special case of asymmetric adjustment costs where the resale price of capital goods is zero. 4 Our measure of capacity utilisation is 100 minus the percentage of firms reporting that they were working below capacity. 5 Based on Equation 8 of the Arellano-Bond (1991) paper. 6 Based on Equation 10 of the Arellano-Bond (1991) paper. 7 Using the log of the concentration ratio gave much better results—in terms of the statistical significance of the estimated coefficients —than using the concentration ratio itself. 8 Investment is net of disposals, but not of depreciation. 9 In estimation, our measure of capacity utilisation was 100 minus the proportion of responses to this question (number four in the Industrial Trends Survey).

References Abel, A.B., ‘Optimal Investment under Uncertainty’, American Economic Review, 73, 1, 288–283, 1983. Arellano, M. and S.Bond, ‘Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations’, Review of Economic Studies, 58, 277–297, 1991. Beaudry, P.M. Caglayan and F.Schiantarelli, ‘Monetary Instability, the Predictability of Prices and the Allocation of Investment: An Empirical Investigation using UK Panel Data’, September 1996. Bertola, G., ‘Adjustment Costs and Dynamic Factor Demands: Investment and Employment under Uncertainty’, Ph.D. dissertation (Chapter 2), MIT, June 1988.

RINA BHATTACHARYA AND PAUL HOPE

63

Caballero, R.J., ‘On the Sign of the Investment-Uncertainty Relationship’, American Economic Review, 81, 279–288, 1991. Caballero, R.J. and R.S.Pindyck, ‘Uncertainty, Investment and Industry Evolution’, International Economic Review, 37, 3, 641–662, August 1996. Craine, R. ‘Risky Business: the Allocation of Capital’, Journal of Monetary Economics, 23, 201–218, 1989. Dixit, A.K. and R.S.Pindyck, Investment under Uncertainty, Princeton University Press, Princeton, 1994. Driver, C. and D.Moreton, ‘The Influence of Uncertainty on UK Investment’, Economic Journal, 101, 1452–1459, November 1991. Driver, C. and D.Moreton, Investment, Expectations and Uncertainty, Basil Blackwell, Oxford, 1992. Eberly, J.C., ‘Comment’in R.S.Pindyck and A.Solimano, ‘Economic Instability and Aggregate Investment’, National Bureau of Economic Research Macroeconomics Annual, 1993. Fischer, S. and R.C.Merton, ‘Macroeconomics and Finance’, Carnegie-Rochester Conference Series on Public Policy, 21, 1984. Ghosal, V. and P.Loungani, ‘Product Market Competition and the Impact of Price Uncertainty on Investment: Some Evidence from U.S. Manufacturing Industries’, Journal of Industrial Economics, 44, 2, 217–218, June 1996. Hartman, R., ‘The Effects of Price and Cost Uncertainty on Investment’, Journal of Economic Theory, 5, October, 258–266, 1972. Hurn, A.S. and R.E.Wright, ‘Geology or Economics? Testing Models of Irreversible Investment Using North Sea Oil Data’, Economic Journal, 104, 363–371, March 1994. Jorgenson, D.W., ‘Capital Theory and Investment Behavior’, American Economic Review, 53, 2, 247–259, May 1963. Katz, M. and C.Shapiro, Terfect Equilibrium in a Development Game with Licensing or Imitation’, Discussion Paper 85, Woodrow Wilson School, Princeton University, 1984. Levi, M.D., ‘Errors in the Variables Bias in the Presence of Correctly Measured Variables’, Econometrica, 41, 5, 985–986, September 1973. Nickell, S.J., The Investment Decisions of Firms, Cambridge University Press, Cambridge, 1978. Pindyck. R.S., ‘Irreversible Investment, Capacity Choice and the Value of the Firm’, American Economic Review, 78, 969–985, December 1988. Pindyck, R.S. and A.Solimano, ‘Economic Instability and Aggregate Investment’, National Bureau of Economic Research Macroeconomics Annual, 1993. Price, S. ‘Aggregate Uncertainty, Capacity Utilisation, and UK Manufacturing Investment’, Applied Economics, 27, 147–154, 1995. Stiglitz, J. and A.Weiss, ‘Credit Rationing in Markets with Imperfect Information’, American Economic Review, 71, 3, 393–410, June 1981. Tirole, J., The Theory of Industrial Organisation, MIT Press, Cambridge, Massachusetts, 1988. Woods, R., ‘Econometric Models of Business Investment: The Role of Factor Prices, Q and Financial Constraints’, Government Economic Service Working Paper 127, June 1995.

5 Corporate governance, investment and economic performance Simon Peck and Paul Temple

We want forms of governance that encourage companies to invest and grow —to take the long-term strategic view and a broad view of their obligations. President of the Board of Trade, Margaret Beckett in a speech to the CBI, 1997 Introduction The long standing debate about the relative merits of public and private ownership has, since the 1980s, been superseded by a debate about the merits of the various forms that private ownership actually takes. Issues of corporate governance—the mechanisms through which firms are organised and controlled—have come to the fore as increasing international integration has highlighted the very real differences that exist in these mechanisms. How far do different governance structures impact upon economic performance of the corporate sector and hence upon the income and welfare of a country’s citizens? The question is as important for the so-called major industrial economies as it is for the transition economies, where fundamental choices have had to be made. In the UK and US there is currently much discussion of the efficacy of corporate governance arrangements. While the highly influential review of such arrangements in the UK, the Cadbury Committee on the Financial Aspects of Corporate Governance, concluded that the UK’s system was ‘basically sound’, other commentators are however, more cautious regarding this assessment (e.g. Jensen 1993). As we see below, while the UK system shares important economic characteristics with the US, contrasts can be made with prevailing governance arrangements in Japan, Germany and much of Western Europe. Here also, as Mayer (1997) notes, there is a concern that existing systems of governance are stifling innovation and growth. Although the term corporate governance has only been in common usage for just over a decade, some of the issues can be traced back to the seminal observations of Berle and Means (1932) on the implications of the separation of ownership from management in the largest, and economically pre-eminent, corporations of America. Nevertheless, there is still often considerable disagreement relating to the scope of the concept itself. Some commentators limit corporate governance to dealing with ‘the ways in which suppliers of finance to corporations assure themselves of getting a return on their investment’ (Shleifer and Vishny 1997, p. 737); more broadly conceived, however, it refers to the mechanisms by which companies are controlled, directed and made accountable (Conyon and Peck, 1998). At the very broadest level the debate concerns an attempt to place the concept of corporate governance within a wider ‘stakeholder society’ in which the corporate sector is held accountable to a wider constituency than just that of its owners (Hutton, 1995) A considerable literature on the stylised distinctions between different systems of corporate governance now exists. However, the links between the functioning of these systems and their role in the determination of economic performance is not transparent. This is often because it is difficult to make assertions regarding corporate behaviour on the basis of economybased, or macro, characteristics. It is, perhaps, more useful to consider the ways in which these differences in governance arrangements can determine specific firm behaviour by considering the various incentives and potential constraints they produce. The precise ways in which the corporate governance arrangements interact with other economic variables to determine the performance of companies is still the subject of considerable academic and policy debate. Indeed as Mayer (1997) notes it is an area where ‘opinion has drowned fact’ (p. 152). In the spirit of producing more focus to the debate, the purpose of this chapter is a relatively modest one: to consider one specific channel in which corporate governance may impact on corporate performance, namely through investment behaviour. Three strands of the debate stand out. The first emerges on the demand side and is that managers may possess discretionary powers and have interests which conflict with owners. This means that we can not necessarily assume that the demand for investment is determined by profit maximising behaviour. The second relates to the provision of finance and is that alternative governance mechanisms produce quite different incentives to generate and share information with investors. The

SIMON PECK AND PAUL TEMPLE

65

final strand concerns the impact of so-called ‘short-termism’ (or myopia) on investment decisions, which is dependent upon specific governance relationships and which straddles both the demand and the supply side of the investment decision. The chapter looks at each of these in turn after the consideration of a basic typology of alternative governance mechanisms. Reference is also made throughout to what is known on these issues, though robust empirical evidence is often scarce. The final section concludes. Alternative governance structures Why governance matters In traditional economic theory, there is little scope for issues of corporate governance and performance for two main reasons. First, the objectives of the firm are rarely seen as problematic. Ideally, the owner and manager, as in the Marshallian entrepreneurial firm, are one and the same person. While a variety of objectives for these owners are conceivable, the standard assumption of profit maximisation is preferred, not least because it is the only assumption consistent with survival under strongly competitive conditions. Second, since the constraints under which these firms operate include a production function— a mechanical translation of inputs into outputs— the firm’s internal workings are deemed largely irrelevant. In the corporate governance literature, the scope of economic theory is extended in both directions, creating the need for an explicit theory of the firm. The firm’s goals become problematic while the efficiency properties of different forms of organisation—goal compliance—becomes a central aim for analysis, opening up the black box of the production function. As Hart (1995) notes, the problem with the traditional approach to the firm emerges under two conditions. First, that there is an ‘agency’ problem —owners and managers are not the same people and the managers (agents), who make the executive decisions care about the outcome of the firm’s operations. Second, at least some relevant information is privately held by the agent (i.e. there is an ‘information asymmetry’). In typical models, chance affects outcomes, and so managerial effort cannot be directly observed; hence compensation cannot be tied to effort. Instead, managerial rewards can be tied to actual profit outcomes. However this generates a trade-off between incentives and risk sharing; managerial compensation can be set (at the extreme) so that all the increase in profit arising from extra effort accrues to the manager, with a lump-sum payment for the owner. However this loads all the risk on to management. This is certainly inefficient, given that management is likely to be risk averse, while owners, who can diversify their own risk by holding shares in many companies, are risk neutral. In fact, Hart points out that this still does not specify a role for governance structures where contracts can be made specifying all states of the world. Instead the question of corporate governance and the effectiveness of alternative corporate governance structures arises in situations where contracts, because of a variety of transactions costs, are incomplete. As a corollary of these general considerations, the governance debate has been chiefly concerned, not with the entrepreneurial firm where goal compliance problems are (to some extent) attenuated, but the large corporation and especially those firms for which titles to ownership are actively traded on a stock market. Today the listed joint stock form of enterprise dominates large swathes of modern economies, albeit to an extent showing significant variation.1 The investment (and other) decisions of many joint stock companies (‘policies’) are effectively made by managers who, conceptually, are quite separate from the owners. The latter group bear title to the residual (or surplus) income generated by firms and have the power to dismiss management but have no direct control over policy. Management may or may not be owners themselves and may act in a self-interested manner, which is divergent from that of the owners. In these companies, managers possess sufficient power over information flows to satisfy both conditions outlined by Hart. In addition, the complexity of the world ensures that there is no way that managerial contracts can be written which specify rewards that could enforce owners’ interests (i.e. contracts are said to be incomplete). The relevance of governance relates to the effectiveness of alternative structures in achieving goal compliance and to the nature of those goals themselves. A basic taxonomy If any degree of consensus exists in the governance debate, it lies in the basic taxonomy of structures and how alternative systems are distributed across the advanced economies. This observation arises from the fact that goal compliance issues at the level of top management involve two basic variables: first, the extent to which there is overlap between ownership and management and second the extent to which ownership titles are dispersed. In general, highly dispersed ownership structures are more likely to be found in the Anglo-Saxon economies, with little overlap between ownership and management. Elsewhere, in Japan and Europe, ownership structures tend to be more concentrated. The importance of the degree of dispersion of share ownership stems from a simple consideration of the costs and benefits, in a world where information is costly to acquire, of monitoring management and enforcing particular policies. Where ownership is highly concentrated, there is a very powerful incentive to monitor management actions because the benefits of this

66

CORPORATE GOVERNANCE AND PERFORMANCE

costly action accrues in the main to the same group of individuals. This incentive (and the fact of concentrated ownership generally), results in owners having some degree of direct representation on the Board of Directors, which (in all the systems examined here) acts as the legal representative of the owners. By contrast, where ownership is diffuse, and where owners tend to reduce risk by maintaining a portfolio of equity claims, it may be in no individual’s interest to incur the costs of monitoring management. The problem is compounded by public good free-rider effects, since the benefits of monitoring accrue indiscriminately to owners as a group. It follows from this that concentrated ownership tends to be associated with some means of directly influencing management policies. In the modern literature this is referred to as ‘inside control’. Where ownership is diffuse and the individual shareholder has little direct influence over policy we have a situation referred to as ‘outside’ or sometimes as ‘market control’. From a governance perspective the difference between inside and outside control is fundamental. That large managerial firms are likely to be in one or the other camp, but not both is based on economic considerations, frequently reinforced by legal and regulatory ones. Large shareholders who do not have a commensurate degree of influence over policy face a ‘lock in’ problem when confronted by evidence of poor performance, because the size of their holding effectively prevents them from selling their shares. Regulations such as those relating to so-called ‘insider-trading’ information, frequently reinforce this polarisation. The above analysis suggests (following for example Franks and Mayer (1992) or Morris 1998) that a four way classification of governance types is useful, based on the extent of overlap between management and ownership and on the degree of ownership dispersion, as shown in Table 5.1. While it may be thought that therefore concentrated ownership patterns possessed some clear economic advantages, on public goods grounds, the conclusion is far from being clear cut. Small shareholdings generally allow greater liquidity and offer greater potential for diversification. This may allow economies, where dispersed ownership is widespread, to adopt an investment profile which is rather riskier in general and this may be closer to a social welfare optimum. In any event it should be noted that large shareholdings do not offer the same incentives as 100 per cent ownership. Moreover, there are alternative disciplinary mechanisms to direct monitoring by owners and these may be just as, or indeed more, effective. These are generally recognised as: the Board of Directors and proxy battles; management compensation contracts; the threat of hostile take-over; and the structure of finance. To make the discussion more concrete, it is helpful to discuss these mechanisms in the light of governance practice in four advanced economies. Governance structures in four economies Many of the ideas can be illustrated with reference to practice in just four economies—the US, UK, Japan and Germany. Our attention is primarily focused on corporations whose shares are listed on a stock exchange. The relative importance of these listed companies varies substantially. Basic data on the extent of stock market capitalisation (in relation to GDP) in each of the four economies is presented in Table 5.2. It should be noted that, for Germany and Japan, such data are hard to interpret because of the extent of cross-ownership of shares. Table 5.3 looks at the structure of beneficial ownership across four economies. It can be seen that only in the US are individuals the most important type of owner. Non-financial corporations are important in Japan Table 5.1 A taxonomy of governance structures Degree of ownership dispersion? Executive decisions? Owners Management

Low

High

Entrepreneurial control Inside control

Partnership Outside control

and especially in Germany, where cross-holdings of company shares are extremely common. The data in Table 5.3 tell us nothing about the concentration of ownership, nor whether shares are held for portfolio or for corporate control purposes. What we need to know is whether large shareholders typically engage in monitoring. Here the differences are more striking. Both Porter (1992) and Prowse (1994) adopt a method for reformulating the basic data in Table 5.3—splitting shares held by financial institutions into those for which shares are held as an agent for other investors (e.g. pension funds) and those which are held for control purposes. Prowse reckons that the vast proportion of shares held by financial institutions are held as agents for portfolio purposes in the US and UK, with only 2.0 per cent and 0.7 per cent respectively of outstanding stock held for long term control purposes. Typically financial institutions are reluctant to take

SIMON PECK AND PAUL TEMPLE

67

large stakes in individual companies in these countries. In the UK for example there is generally a limit of 5 per cent on pension funds’ or unit trust holdings of equity stakes in individual companies. Large holdings are also hindered by insider trading laws in both countries. Conversely, Prowse estimates that 38.5 per cent of outstanding shares are held in Japan by financial institutions for control purposes, and 33 per cent in Germany. The latter figure includes substantial holdings of shares by individuals for whom a bank exercises proxy rights. On concentration of ownership, Franks and Mayer (1992) showed that patterns of ownership in Germany were much more concentrated than in the UK. Of the 200 largest companies in Germany nearly 90 per cent had Table 5.2 Stock market capitalisations, 1997 Country

Stock market size (£billion)

No. of companies

% GDP

UK US Germany Japan Source: Datastream.

1,250 6,490 502 1,289

2,465 7,724 700 1,805

180 150 36 41

Table 5.3 Ownership of common stock in 1990 (percentage of outstanding shares owned) All corporations Of which: financial institutions Non-financial corporations Individuals Foreign Government Source: Prowse (1994).

US

UK

Japan

Germany

44.5 30.4 14.1 50.2 5.4 0

62.9 52.8 10.1 28.0 6.5 2.5

72.9 48.0 24.9 22.4 4.0 0.7

64.0 22.0 42.0 17.0 14.0 5.0

one shareholder with 25 per cent or more of the equity issued. In the UK, in two-thirds of the largest 200 companies no investor held 10 per cent or more of the equity. A similar lack of concentration of ownership has been noted for the US (see Cho 1998). While the essence of the distinction between inside and outside control resides in the existence of owners with sufficient incentive to monitor management and the power to do it effectively, there are, as we have seen, other mechanisms through which governance may be exercised. We consider these in turn. Legally, the Board of Directors constitutes the elected representatives of the shareholders. As Jensen (1993) remarks ‘the board, at the apex of the internal control system, has the final responsibility for the functioning of the firm’ (p. 862). In extremis, of course, the shareholders have the power to dismiss the Board in a proxy fight organised by one of more dissident shareholders. However, since the costs of mounting such a battle may be considerable there is clearly a similar free-rider problem here to that we saw in the monitoring role. In external control situations, reference is sometimes made of the ability of independent, or non-executive directors to undertake informed monitoring of management on behalf of shareholders. These are directors with no, or little, direct financial stake in the company. Their role may be important because it seems reasonable to assume that executive (inside) directors will not self-monitor, or monitor effectively the performance of the Chief Executive Officer (CEO). The career of the executive director, after all, is closely tied to the incumbent CEO and so they do not possess sufficient incentives to remove them or to, say, restrict their compensation growth (Hart 1995; Jensen 1993). There are, however, ample reasons for believing that non-executives have insufficient incentives or power to monitor management. The company equity holdings by nonexecutives are typically low and so is the willingness to correct managerial mistakes. In addition, the compensation received by outside directors, along with their chances of re-selection as a non-executive director, are influenced by the CEO. So, as Nickell (1995) aptly remarks ‘why should they make a fuss rather than keep quiet and collect their fees?’ Second, there seems to be a clear asymmetry of information between executives, as full time employees, and non-executive directors. Typically, non-executive directors spend only a fraction of their time at the company and, indeed, can be executive directors at other companies. This seems to build into the internal governance system an information bias in favour of the executive directors. Board structures differ in the German model of corporate governance which embodies the principle of co-determination. Whereas board independence is considered paramount in the unitary Anglo-Saxon structure, the German counterpart, the supervisory board, or Aufsichtsrat, is selected from a constituency of those with interests in the firm, not least the German banking sector. The 9 to 21 members of the Aufsichstrat appoint a management board, or Vorstand consisting of 5 to 15

68

CORPORATE GOVERNANCE AND PERFORMANCE

members. No overlap in membership is allowed between the two boards. In practice however, it seems that management board chairs wield significantly more influence than do the supervisory board members (see Conyon and Peck 1998a). Japanese unitary Board structures stand in a certain relief to practices, not only to both the US and UK, but also to Germany. A majority of directors are internally promoted, many of whom will have been with the company since university graduation; however, the length of tenure of the CEO in the US, or the Managing Director in the UK, are not apparently that much shorter than the president in Japan (Odagiri 1992). More significant perhaps is that the background of the directors tends to be in production, technology or marketing, with a more firm-specific focus, than in the UK or US, where financial, business school backgrounds are more common. Moreover, Japanese board structures also tend to have more representation from outside companies and banks, representing both the importance of business groupings and the generally greater importance of bank lending for the corporate balance sheet. If the Board does not constitute, in practice, an effective means of monitoring management, managerial compensation contracts may offer a different route for aligning the interests of owners and management. The link between managerial compensation and corporate performance has been the subject of much empirical testing (for a review of UK evidence, see Conyon and Peck 1998). Current evidence points towards a statistically significant though typically small relationship between corporate performance and managerial compensation. The effect of company size is always a more significant determinant of managerial pay. There is a paucity of evidence exploring the relationship in other economies, particularly in those economies characterised by inside control. However one recent study by Conyon and Schwalbach (1997) using UK and German data finds that the pay–performance relationship in Germany is not significantly different from the UK. The governance mechanism which shows the greatest degree of disparity across the internal-external control divide, is the potential, among listed companies, for hostile take-over. In theory, one management team, aware of slack in another company, can gain a controlling interest; installing a new management team can remove the inefficiency and may yield a capital gain on the enhanced market value of the company. However, in addition to the substantial sunk costs involved in mounting a take over, once again there is a problem of appropriability (Grossman and Hart 1980). As far as small shareholders (who are likely to believe that their actions will not influence the probability of a take-over) are concerned, the interest of a ‘raider’ in a poorly managed company may act as a signal of under-performance. By refusing to accept the tender offer, they may appropriate some of the capital gain for themselves if the take-over goes ahead. Two other reasons which may reduce the gains from a take-over are noted by Hart (1995). First, the existence of more than one raider, competing for the gains and creating a ‘bidding war’ may reduce ex post returns. Second, incumbent management may change policy to similar effect. There have been many studies of the economic impact of take-overs but of course, motivation for take-over are various, and what matters in the current context is the impact of so-called hostile take-overs (i.e. they are opposed by the incumbent management). So while many empirical studies have indicated that firms that have been taken over are not especially unprofitable (see for example the survey in Morris 1998), Edwards and Fischer (1994) note that only a minority of studies actually distinguish between hostile and friendly take-overs. They refer to studies of US hostile take-overs (e.g. Morck, Shleifer and Vishny 1988; Ravenscraft and Scherer 1987) which do provide some tentative evidence that firms who are the object of hostile take-overs were performing poorly; they also note however, that the same studies do not indicate any postmerger improvement. Interpretation of any such evidence is any case fraught with difficulty, since it is the (largely unobservable) threat of take-over, which is important in constraining managerial behaviour. Whatever their impact in the US or UK, hostile take-overs are, in contrast to these economies, extremely rare in both Japan and Germany. In Germany, and probably more important than any legal restriction, is the existence of large shareholders. If the purchase of these large shareholdings is necessary to achieve control, then the raider may not be able to purchase shares at a price low enough to make any gain on changing the management. In the case of Japan where ownership concentration is not as high as in Germany, Odagiri (1992) points to the many idiosyncratic personnel and other management practices at the level of the firm, and which may dramatically raise the cost of managerial reorganisation. As a final possibility for a corporate governance mechanism, a number of writers have suggested the possible importance of financial structure as a means of signalling the quality of management. By committing itself to making regular interest payments and the repayment of debt, management limits the degree to which it can be inefficient or pay too much for acquiring assets. The strength of this mechanism clearly depends upon what penalties ensue when there is a default on repayments. In many cases of course, termination of employment may impose a considerable penalty on management. Quite large changes in the debt–equity ratio may form part of owners’ strategies to constrain managerial behaviour (e.g. Jensen 1986). The role of banks A question that naturally emerges in studying the comparative structures of governance across the four economies is whether the role of banks in Germany and Japan have especial significance. As both debt and equity holders, banks may have

SIMON PECK AND PAUL TEMPLE

69

incentives to engage in monitoring activity, although as debt holders they may be more interested in the bottom end of possible profit outcomes from investment projects. For Germany, where banks have a powerful position not only as significant owners, and as representatives of other owners, but also as providers of a wide range of financial services, there is a powerful presumption that agency problems will be reduced and longer-term investment stimulated. In Japan the case is less compelling: Japanese banks are in fact restricted by law to a maximum of 5 per cent equity holding in any one firm, so although the average equity holdings by banks are perhaps three times larger (probably around half that of the total holdings of financial institutions [Odagiri 1992]), this must be held by several banks. Useful empirical work on the influence of banks on investment is surprisingly thin. For Germany, the classic study of the superiority of the German system, was that of Cable (1985) who based his study of company profitability of 48 of the largest 100 German Aktiengesellschaft or AGs (public companies as opposed to the private Gesellschaft mit beschrankter Haftung or GmbH). He distinguishes in his study between: 1 the ‘internal capital market (ICM) hypothesis’ (in which the banks are able, through supervisory board representation, to impose profit maximising behaviour); 2 a possible market power effect (banks helping to establish cartels); and 3 a simple financial expertise effect (i.e. that banks can supply expertise through their involvement in a manner which is not marketable). Cable concluded that only the capital market hypothesis is consistent with his econometric results. This view has been challenged more recently in an important study by Edwards and Fischer (1994). They point out that his evidence is consistent with the more general hypothesis that it is the existence of large shareholding concentrations which improves profitability, with no special role for banks. Their own investigation, based upon a detailed examination of external finance in Germany criticises much of the conventional wisdom. For example, German banks do not provide a larger proportion of external finance for investment than in the UK (see also Mayer 1990 on this). Moreover this overwhelmingly takes the form of loans rather than equity (which we would not expect if there were real informational advantages from their position on supervisory boards). Edwards and Fischer also raise another important question, that of ‘who monitors the monitor’? The fact that the big three German banks are actually AGs with a rather dispersed share ownership begs the question of what precisely are the incentives for the apparent monitors to exercise their power in favour of owners. Prowse (1994) also makes a similar point: because of cross-holdings of shares by firms, the degree of dispersion in ultimate (as opposed to direct) ownership, is difficult to unravel. He cites evidence from Schreyrogg and Steinman (1981) that, on the basis of ultimate ownership, although 90 per cent of their sample of 300 large German firms could be classified as owners controlled on the basis of direct ownership, less than half the sample could be classified as such under ultimate ownership. In a more recent study, Chirinko and Elston (1997) look specifically at the impact of bank influence in Germany. They examine the cross-sectional characteristics of firms according to the extent of bank influence, which they measure in two ways: by concentrations of bank ownership and by ‘voting power’, which includes banks’ proxy voting rights. They do not find that bank influenced firms have lower financing costs nor are they more profitable. They conclude that bank influence is just a substitute governance mechanism, rather than a unique (and more effective) one. In Japan similar considerations apply as in Germany. As we have seen, financial institutions figure heavily as owners, but their own ownership is rather diffuse. One commonly held view is that in Japan, the so-called ‘main bank’ acts to provide an efficient ICM (e.g. Sheard [1989]; Hoshi et al. [1991]). As in the case of Germany, there are misgivings about this role. Odagiri (1992) doubts whether the role of banks is much different in Japan, from that in the US or the UK. However, he emphasises the role that stability in bank-firm relationships may have in allowing a build up of assets which are highly firm specific. As debt holders, banks are of course especially concerned with ensuring that firms avoid financial distress and a slightly different hypothesis to the ICM is that banks in Germany and/or Japan are in better position to help firms in difficulty. This possibility may arise from economies of scope. In Japan for example, Aoki (1990) notes that the ‘main bank’ may be able to detect problems at an early stage through information gained from its range of financial services as well as other more or less explicit monitoring activities. Certainly it may be in the bank’s interest to monitor firms closely in the context of poor profit states, since firms close to bankruptcy may adopt excessively risky policies. In Germany, similar hypotheses have been advanced about the importance of economies of scope for the main bank. A priori, it might be supposed that supervisory Board representation might lead to earlier identification of problems. This view is once again challenged in Edwards and Fischer, who argue that banks in fact rarely get involved in management activities in circumstances of distress, although (largely for reasons of ‘image’) they may involve themselves in re-organisation attempts among larger firms. Their incentive to involve themselves in management activities may be smaller than commonly supposed because their loans are generally well secured.

70

CORPORATE GOVERNANCE AND PERFORMANCE

Implications for investment behaviour (1): managerial discretion The above analysis gives ample grounds for supposing that alternative governance relationships may need to be taken into account when modelling investment behaviour. If governance matters, then two questions stand out: when governance is weak, how will managerial discretion be used? And, second, how do the informational characteristics of alternative governance relationships impact upon investment? In the neo-classical economic parables, and by implication in the vast majority of theorising about investment behaviour, the firm itself is reduced to a black box—the production function—which acts as a constraint on profit maximisation. In many models the conventional behavioural assumption—that of profit maximisation—is largely unnecessary, since it is enforced through the assumption of perfectly competitive product markets. Although even with perfect competition, institutional details may matter in modelling the investment decision (in modelling for example the impact of the costs of firm insolvency), it is when we move to imperfectly competitive product markets that the question of governance moves to the fore. However, for governance to matter, it is also important that management must have goals which diverge from shareholders. Despite its importance, in recent times the question of managerial goals has not perhaps received the prominence it deserves. Two literatures on the question of managerial discretion can be discerned. An earlier literature, stemming ultimately from the work of Berle and Means, considered the broader motivation of management in a ‘corporate society’. In his seminal contribution, Marris (1964) suggested that economic, psychological and sociological contributions to the question of motivation tended to point to the importance, of salary, status, power and security. In a managerial context, these first three motives may give rise to a desire for size, in addition to profits. Certainly, as we saw above, a consideration of the narrow economic incentives resulting from managerial compensation tended to the view that size mattered. The question then might be posed as to whether it was size or the rate of change of size that mattered—the two only amounting to the same if the initial capital stock were given. Marris himself was inclined to the view that it was the rate of change of size, i.e. the growth of corporate assets, which provides the correct starting point in the determination of corporate policy. At least after a certain point, growth conflicts with profitability, and raises the probability of a take-over and so increases insecurity. The class of models that emerges from this literature is instructive from a governance perspective, and it may be helpful to examine some of the salient features. Corporate policies refer to the choice of alternative steady states typically described by a constant growth rate for the capital stock, g=dK(t)/K(t) and a constant profit rate, p=Π(t)/K(t), where profits are defined as operating profits, gross of any investments incurred to secure the growth of demand. The choice over profit rates and growth rates is conditional upon the ‘Penrose’ effect (Penrose 1959; Uzawa 1969; Marris 1964; Odagiri 1981), according to which, at least beyond a certain point, there are costs attached, not to an increased scale of operations, but to their growth, arising from limitations of managerial capacity. The Penrose effect stresses the importance of firm specific investments in human capital, which mean that new managers are not immediately as productive as existing managers. In general, if we suppose that a manager’s productivity is related to length of time with the firm, then higher rates of growth are associated with management teams of lower average experience, and lower productivity. Thus, the ratio of investment to capital, Ψ(g)=I(t)/K(t) is a function dependent upon the rate of growth (but independent of time). Other factors, from the demand side, such as advertising expenditures intended to raise sales, may also produce similar results. In such a world, growth has a double-edged effect. On the one hand it causes future profits to rise, but only at the expense of net profits (after deducting the investment necessary for growth, which may take the form of managerial effort). This negative impact eventually dominates, and creates a trade-off between the value of the firm and economic growth Specifically, it can be shown that the market value of the firm divided by the capital stock, v (or Tobin’s Q) is (with discount rate i) given by:2 This trade-off describes the so-called ‘growth-valuation frontier’. Although the Penrose effect eventually dominates, i.e. Ψ′(g) >0, it may well be that for small g, the growth valuation frontier may be rising, implying that a zero growth rate for the firm may involve missing out on profitable investments. Completion of the model requires a managerial utility function U=U(g, v), as well as a specification for the supply of finance. With only capital market constraints on managerial discretion in position, the valuation ratio acts as a proxy for insecurity (the probability of being taken over). Much of the managerial discretion literature itself is based upon how individual corporations might choose between policies located on this frontier. In Marris’s original view, the classical owner-controlled firm will locate where v is maximised (the dynamic analogue of static profit maximisation). Managerially controlled firms would, he argued, pursue faster growth at the expense of the firms’ valuation: investment will be pushed beyond that indicated by shareholders’ opportunity cost of capital. At the theoretical level there are a number of problems associated with the use of the firm’s growth rate in the utility function, if we regard the initial level of the capital stock, not just as historically given, but also as corporate policy variable (Solow 1971). In the light of recent experience, but contrary to the terms of the original debate, it is clear that large and discontinuous changes can be made to corporate size. This is not just in an upward direction (through price reductions or merger and acquisition), but also in a downward direction, where sales of physical assets have been accompanied by share

SIMON PECK AND PAUL TEMPLE

71

buybacks. With the current period capital stock potentially a policy variable, the use of the corporate growth rate in a utility function may seem rather odd. Solow also points out, in a number of comparative static exercises, that the qualitative differences (in their response for example to changes in policy) between managerially motivated and traditional profit maximisers may be small. A more recent approach (see Marris 1991) examines the properties of a system which allows firms to choose an adjustment period to a steady state growth path, allowing for joint optimisation of the adjustment period with the steady state growth path. However, optimisation with a discounted value of employment as one of the arguments of the objective function is allowed for. A large number of attempts have been made to test empirically the managerial hypothesis and determine whether ownership structure does impact upon performance whether measured by Q, or by some accounting rate of profit. A useful survey is provided by Short (1994) who highlights some of the difficulties in this work. These include determining whether or not a particular firm is owner controlled (see in particular Cubbin and Leech 1983), and the problems in controlling for other conditions affecting a firm’s profitability. These will include such factors as organisational design, the degree of product market competition, or the extent to which firms are subject to threat of take-over. In an extension to Marris’s basic hypothesis, Mueller (1972) observed that there might be a life cycle effect at work. Younger firms are more likely to be owner controlled and associated with earlier stages of the product cycle, and may therefore be both more profitable and grow faster. Most empirical work relates to heterogeneous performance across companies within the same economy. There is perhaps only one major study which looks to differences in corporate governance as key to differences in both micro- and macroeconomic performance across economies. This is due to Odagiri’s development of work by Marris in Odagiri 1981, 1992; see also Marris 1972, 1991). Odagiri (1981) introduces a model in which individual firm’s R&D decisions have an impact on the economy’s natural rate of growth and so it is possible to introduce a macro-economic analogue to the micro growth-valuation transformation frontier. Odagiri uses this device to argue that the degree of managerial preference for growth can explain differences in macro-economic performance between Japan and the Anglo-Saxon economies. These differences include not only the higher rates of investment and growth observed in Japan in the post-war period (at least up to the 1990s), but also the comparatively low recorded profit rates. A key component in this argument is that Japanese performance was helped by rather weak corporate governance—explained in particular by absence of hostile take-over activity. The frontier provides one possible explanation for the co-existence of high growth rates and low profit rates, as illustrated in Figure 5.1 for a sample of companies in the four economies.3 Measures to make agents behave more as if they owned the company and give them a greater financial stake have been promoted as a way of ensuring goal congruence between principals and agents. Recent evidence on the role of ownership in investment performance is contained in Cho (1998). This study explicitly allows for the potential endogeneity of ownership, investment and corporate performance, and finds for a sample of US firms, it is investment that drives firm value that affects ownership structure (i.e. concentration of ownership), but not the reverse; this suggests that concentrated ownership does not necessarily induce managers to make value maximising investment decisions. The policy implication of this study is that measures to give managers greater ownership stakes in the firm (e.g. through the use of stock options) may not automatically deliver better quality decision making regarding investment. Implications for investment behaviour (2): asymmetric information and the cost of capital In a basic neo-classical investment model, the capital market is frictionless. Armed with full information, and where investment is reversible, entrepreneurs face a constant cost of capital—basically the real market rate of interest adjusted for the risk involved for that particular class of investment projects. Associated with this approach is the Modigliani—Miller ‘irrelevance’ hypothesis according to which (and in the absence of transactions costs and differential taxation of alternative forms of finance) entrepreneurs should be indifferent to the source of finance whether from retained profits, from new share issues, or from external debt. That this view of the world is overly simplified can be seen by noting two things. First that, as a stylised empirical fact, retained earnings are generally the favoured form of finance for investment in all advanced economies (see for example Mayer 1990). This is despite the fact that most tax systems (by making debt interest payments tax deductible) provide an incentive for debt over retained earnings. Second, this class of models, given the dependence of investment upon expected profitability and the cost of capital only, seems incapable of explaining the extreme macro-economic volatility of investment typically observed. With these limitations of the neo-classical in mind, economists have been exploring, for a long time, the implications of information asymmetries for investment. A typical example might be where the outcome of an investment project depends upon not only physical capital, but also upon managerial effort, and chance factors. Understanding the temptation that exists for lowering effort, investors will demand a ‘lemons premium’ for providing external finance. This drives a wedge between the cost of capital from external sources and internal sources (retained earnings). If the investment opportunities of an individual firm are such that not all the required funds can be met from profits (or net worth), then the firm can be described as being finance constrained. For those firms that face such a constraint we would expect to find that investment depends

72

CORPORATE GOVERNANCE AND PERFORMANCE

Figure 5.1 Return on capital employed. Source: Datastream.

upon net worth. Examples of such models applying to loan markets are those of Jaffee and Russell (1976) and Stiglitz and Weiss (1981); Myers and Majluf (1984) have investigated the consequences for equity markets. Note that the approach seems able to explain not only the observed preference of firms for internal finance, but also some of the observed volatility of investment, since profits are themselves volatile and dependent, inter alia, upon capacity utilisation. Empirical investigation of the role of information related financing constraints is hampered by the fact that current cash flow (which usually proxies for net worth) are likely to be related to investment opportunities: ways must therefore be found to proxy for investment opportunities (Hubbard 1997). So-called Q models attempt to provide for this by assuming that stock market valuations of a firm’s capital stock provide a control for investment opportunities. In fact of course it is marginal Q— the impact on a firm’s present value of an additional unit of capital—that matters, and not the average Q which is observable from stock market and accounting data. Most empirical implementations of Q models are based on the idea of a representative firm and hence cannot distinguish between finance constrained and non-finance constrained firms. Few of these studies then include a proxy for financial constraints. More relevant are panel studies of firms which allow for a differential impact of net worth depending upon whether a particular firm is, or is not, cash constrained. A number of contributions based on the work of Fazzari et al. (1988) use a priori sample splits of firms in the US. For example, Fazzari et al. use a split based upon firms’ pay-out ratios, on the assumption that firms able to pay substantial dividends are less likely to be faced with financial constraints. For given investment opportunities, they found a substantially larger impact of cash flow for the low pay-out firms. Studies similar to that of Fazzari et al. have shown the importance of financial constraints for groups of firms in: Japan (Hoshi et al. 1990); the UK (Devereux and Schianterelli 1990; Blundell et al. 1992; Bond and Meghir 1994; and for Germany, Elston and Albach 1995). The study by Hoshi et al. has interest in the current context since it deals explicitly with the role of banks in Japan, and the impact of financial deregulation in the early 1980s. Prior to the reforms, bank ties to most firms in Japan were very close, and historically, banks provided a high proportion of external financing. Reform included the abolition of interest rate ceilings on corporate bonds which, by 1983, could be issued on an un-collateralised basis and with warrants to purchase equity. As a result, bank borrowing on a long-term basis declined dramatically (Hoshi et al., 1990). Econometric examination of a panel of firms reveals that those firms who weakened ties with banks faced financial constraints, while those maintaining ties faced no such problem. The authors interpret this as evidence that bank monitoring is an effective means of overcoming information problems, while acknowledging that this poses the question of why some firms severed links with their banks, i.e. what are the costs involved with bank finance? Does a positive cash flow coefficient on an investment equation provide solid evidence for the role of asymmetric information? Kaplan and Zingales (1997) show that the theoretical basis for assuming that the size of the coefficient on cash

SIMON PECK AND PAUL TEMPLE

73

flow is increasing in the degree of constraint may be slim. Indeed, in their own empirical estimations on US company data, they find that less financially constrained firms display greater cash sensitivities. From the current perspective, however, it is clear that a positive coefficient is also consistent with managerial discretion, but as Kathuria and Mueller (1995) point out the two hypotheses are directly opposed: according to the managerial discretion model, we would expect to see the marginal return on investment pushed below that of the shareholders’ opportunity cost of capital while in a world where asymmetric information rules, we would expect to see the marginal return on investment above the cost of capital. Studies of short-termism What is short-termism? In the 1980s, the governance debate manifested itself in popular discussion in the concept of ‘short-termism’, a term which actually captures a variety of related ideas. The broader debate concerned the ways in which the various institutions of economies could interact in practice to produce a generalised ‘institutional myopia’ (Sharp 1998). More narrowly interpreted, it was the assertion that the institutions associated with outside control forms of governance, and in particular, the stock exchange and the big institutional shareholders, could be accused of taking an excessively short-term view which reacted unfavourably on various kinds of investment. This could either involve a systematic underestimation of longer term cash flows, or the employment of higher discount rates in the longer term. However, two further sources of myopia need to be considered. One arises not from stock market evaluations, which are basically neutral between shorter and longer term flows, but from ‘managerial myopia’. This may involve acting irrationally in (for example) using sub-optimal investment appraisal strategies, measures of managerial performance, or remuneration schemes (e.g. Marsh 1990). The other source of myopia arises, not from any irrationality emerging from management, nor from any failure of the financial markets, but from the outside system of governance itself. However important managerial failure may be in practice, the most important theoretical contributions concern the other sources of myopia. We consider these before turning to the empirical evidence. Models of short-termism The most prominent theoretical contributions concern the problems of outside control where information asymmetries are important. The most frequently referred to is that of ‘signal jamming’ (Stein 1988). Dispersion of ownership ensures that owners have little incentive to gather detailed information on any particular company, and hence are relatively ignorant regarding the true state of the firm. Firms therefore have to engage in some form of ‘signalling’ to the market their true state, in particular the growth of accounting variables such as profit, or dividends. Given ‘outside’ ownership, shareholders unaware of the true value of the firm would be willing to accept an offer of take-over from a better informed rival. This implies that share prices will be responsive to such signals. When the ability to pay dividends is linked to firm performance, a separating equilibrium is established between ‘good state’ and ‘bad state’ firms. An important implication of the model is that it is in those very areas where managers’ informational advantage is at its greatest, namely investment in intangibles (R&D, supplier links, training etc.), that can be ‘raided’ in order to raise current dividends. The problem illuminated in Stein’s model is a co-ordination failure across firms. As Stein points out, if managers could coordinate their activities such that none behaved in a short-term manner, then this would also be a situation investors would prefer. However, managers are subject to a prisoner’s dilemma, in that divergent behaviour by any of them (i.e. not short-term) would be perceived as less effective than all others (and their shares marked down by the market). A number of other models explicitly recognise the conflict of interests that characterise the owner-manager relationship. Dickerson and Gibson (1995) look at the firm’s investment decision as a dynamic game between shareholders and managers. Managers have two objectives, first, the rate of growth of the firm, and second, job security (i.e. the firm not going bankrupt, or being taken over). The primary source of growth in their model is via the investment of retained earnings. The constraint placed upon this investment is that distributed earnings or dividends must be sufficient to satisfy shareholders’ demands, and high enough to ensure a share price that will deter corporate raiders. Although managers can propose dividend levels, shareholders have the right to vote on this, and ultimately on the tenure of managers. Managers do have almost complete discretion over the direction of retentions towards investment, but future returns are uncertain in the sense that they are dependent on the future level of dividends, and on the manager still being employed by the firm. Shareholders face the problem that they can accept lower dividends now in the expectation of higher investment and higher dividend levels in the future. This though requires them to trust managers to invest the retained earnings not distributed as profits, and not spend it on managerial perquisites. The problem is the classic agency one, coupled with the fact that in the presence of a large number of disparate shareholders the incentives to engage in costly monitoring in order to reduce the agency problem is limited. As Dickerson et al. (1995) point out, both shareholders and managers are faced with an inter-temporal consumption problem with

74

CORPORATE GOVERNANCE AND PERFORMANCE

uncertain future gains due to the discretionary behaviour of the other party. As they show, where managers and shareholders act independently, the firm will invest sub-optimally, whereas a co-operative solution would lead to higher investment and joint income. This result they put down to the interaction of the principal–agent problem, the characteristics of the financial system, and the existence of a market for corporate control. Sub-optimal levels of investment in the models discussed above may be viewed as stemming from an outside control system as such. As we mentioned, popular conception tends to place the financial institutions alone in the dock, with the institutions preventing management from taking a longer term view. In fact, one of the popular reasons for subscribing for this view of short-termism is the belief that the pressures on fund managers arising from performance appraisal (for example that their performance is assessed quarterly) can be dismissed as naïve. This is because it assumes that current share prices do not reflect all the available information whether about the shorter term or the longer-term profitability of firms (Marsh 1990). However, there is some evidence that this method of appraisal may not be altogether harmless. For example, fund managers apparently may engage in sub-optimal ‘churning’ of equities, in order to ‘lock in’ any performance which happens to be ahead of the general stock exchange index (Lakonishok et al. 1991; see also Nickell 1995). Some theoretical considerations point to the importance of stock market evaluation as such. When there are discrepancies between current share prices and current information, can we be sure that arbitrage operations are neutral between information which relates to shorter or longer time horizons? This is the basis of the mis-pricing arbitrage model of Schleifer and Vishny (1990). In a capital market in which the interest rate equals the rate of return on assets, but where some mis-pricing of assets may occur, arbitrageurs who can borrow at the prevailing interest can attempt to profit. According to Schleifer and Vishny, long term assets are more likely to be mis-priced because the arbitrage cost of long term assets is systematically higher than for short-term assets. This may be due to the fact that the arbitrageur may need to liquidate the assets before price correction takes effect, or because over a longer time horizon, the fundamental value of the company may deteriorate. In an important contribution in a complementary vein, Morris (1998) specifically considers the information content of different types of asset, introducing the role of both sunk costs and externalities. In this regard he distinguishes between four types of return from investment: first, the simple excess of revenue over cost in the short term as a result of undertaking the investment. Second, there may be subsequent long-term profitability, resulting from maintaining the investment, as higher investment is associated with a lower average age of the capital stock, incorporating more recent technological advance. Third, any investment will typically create new opportunities for further investment. Thus, rather than thinking of high investment firms as moving down some investment demand schedule, the very act of investment may transform the nature of the investment returns that the firm faces. One could imagine this effect happening as a result of new cost or demand conditions. Finally, there is the scope for further profit opportunities due to improvements in the firm’s strategic position, i.e. access to new technologies, markets, skills which do not accrue directly because of the initial investment, but would have be unattainable if it had not been made. The problem of assessment and quantification obviously becomes more acute as one moves down this list. Indeed it is unlikely that standard methods of investment appraisal would even consider anything beyond the immediately quantifiable. But the returns from these ‘external’ benefits are potentially only likely to become significantly positive when, argues Morris (1998), the present value of projects using traditional methods of evaluation have fallen. Also the higher the level of sunk cost investment, the longer the period until later elements of the profit profile appear. However, this leaves the firm open to negative shocks that may render it vulnerable to losses, and thus a falling share price and take-over. Empirical evidence So what empirical evidence is there for the existence of short-termism with an impact on investment behaviour? First of all, what evidence is there that stock exchange valuations matter when firms’ make investment decisions? Here we might expect a clear difference between the principal governance forms. In a specific study of the impact of the stock exchange on investment, Mullins and Wadwhani (1989) cite survey evidence (Abbegglen and Stalk, 1985) that managers pay less attention to the stock market in Japan or Germany than in the Anglo-Saxon economies, where hostile take-overs are much more likely. They provide econometric evidence to back this up using a standard neo-classical capital stock adjustment model, modified to allow for bankruptcy costs; they find that the stock market has a greater influence on investment behaviour in the UK and US than in Germany or Japan. However, the stock exchange’s influence occurs via its impact on the debt–equity ratio. Morris (1998) argues that this probably reflects the information content of this variable in outside control structures. In fact the ratio of the market value of equity to asset replacement cost (i.e. Tobin’s Q) has no additional explanatory power. Despite its prominence in much of the writings on economic performance, the issue of short-termism has come in for relatively little formal attention by applied economists. Three distinct methodologies can be distinguished: surveys; studies which show the impact of particular announcements on equity prices; and formal econometric techniques. Some of the available evidence is summarised in Table 5.4.

SIMON PECK AND PAUL TEMPLE

75

The CBI surveys are well known, and in their questionnaires the issue of ‘constraints’ on particular actions is well known. In particular, the issue of constraints on investment is explored. The CBI (1987) survey suggested that the company’s share price was not considered a significant constraint on long-term investment by some 60 per cent of Chief Executive Officers (CEOs). However, whether one considered 40 per cent a large number or not, Marsh (1990) suggests that it is the CEOs’ reluctance to invest that makes them lowly rated (though the CBI survey does not test for whether these CEOs are, indeed, from low rated firms). Event-type studies have tended to provide negative evidence for the shorttermist hypothesis. Typical perhaps was that of McConnell and Muscarella (1985) who found that announcements of planned investment increases were associated with share price increases, with the converse for planned reductions. Evidence of such a kind must be regarded as inconclusive if only because it tells us nothing about possible and profitable investment projects, which never go forward because of fears that the reaction of the financial markets may be unfavourable. Even if we had evidence regarding such projects it would not of course prove that financial markets were at the root of the problem, since an irrational belief in the short-termism of the markets may be sufficient. Important UK evidence is presented in Miles (1993). He estimates equations for 477 UK company share prices over the period 1975–89 on the basis that, apart from random deviations, they correctly anticipate both future dividends and the price of the share when it is sold. An important part of this study was that Miles was able to introduce firm specific risk premia as an element in the discount rate. On various definitions of shorttermism, including higher discount rates in the longer term, or undervaluing longer term cash flows, he found significant evidence for short-termism by Table 5.4 Empirical studies of short-termism Study

Method

Sample

Results

Demirag (1994)

Attitudinal survey

116 R&D directors of UK manufacturing companies

Marston and Craven (1994)

Attitudinal survey

UK finance directors

Miles (1993)

Econometric methods; test of GAPM

447 UK firms over the period 1980–1987

Nickell and Wadhwani (1987)

Econometric test of share prices

UK stock market firms

Mayer and Alexander (1990)

Empirical study of firm performance

Matched sample of quoted and unquoted UK firms

CBI (1987)

Survey

UK CEOs

Office of the Chief Economist, SEC (1985)

Econometric

US firms 1980–1987

Coates, Davies and Stacey (1994)

Survey

UK, US and German TNCs

Grinyer, Russell and Collison (1998)

Survey

Finance directors of Times 1000

Directors perceived short-term pressure from capital markets. Thus they act in a shortterminist manner. Bias against long-term R in favour of D No unanimous view on shortterm pressures by finance directors Stock market valuations of firms are short-term; long-term cash flows are underestimated by an estimated 40% Find relative weight attached to current dividends relative to future ones is too high Quoted companies outperform unquoted companies; more heavily concentrated in R&D intensive industries Weakness in share price cited as a major constraint by 7% of respondents, as significant by 34% on long term investment No evidence of either size of institutional stock holdings, vulnerability to take-over, or lower R&D Authors argue that managerial incentive schemes reinforce short-termist bias in UK and US enterprises Short-termism amongst finance directors associated with their perception of importance of short-term indicators to the stock market

76

CORPORATE GOVERNANCE AND PERFORMANCE

Study

Method

Sample

Results

McConnel and Muscarella (1985)

Econometric

‘Announcements’ of investment plans/R&D and share price of US companies

Except in regulated industries, higher investment associated with higher stock prices

financial markets. For example, he estimates that cash flows six months in the future are underestimated by 5 per cent relative to rational expectations but cash flows which do not accrue for five years are systematically underestimated by almost 40 per cent. Miles’s study lends support to both the mispricing arbitrage and the Morris model of short-termism. One difficulty with the study is the assumption that the firm-specific risk premia are constant over time. However the theoretical case for supposing that these rise with time is not compelling. There are also an increasing number of studies that are attempting to measure managerial perceptions of short-termism, often to integrate them with external sources such as stock markets, and the subsequent affect of these pressures on action and performance. For example, Demirag (1994) comes to the conclusion that managers can behave in a short-term way simply because of their beliefs about the short-term nature of capital markets, even if there is no actual evidence that markets have a preference for the short term over the long run. This largely arises from the fact that if managers believe markets to be shorttermist, then they are unlikely to divulge information based on the long-term prospects of the firms if they believe markets could penalise them for doing this. Demirag (1995) also finds that management (finance directors) that perceive markets to be short-termist tend to be highly sensitive to internal measures (such as sales or profits) in the setting of their R&D budgets. Grinyer et al. (1998) drawing evidence from an attitudinal survey, note a reluctance amongst finance directors to engage in ‘revenue investment’, i.e. investment that should have future beneficial effects, but only at the expense of current profits—due to the level of emphasis placed on current earnings by the market. This allows them to come to the conclusion that ‘many finance directors of large UK companies are shorttermist in their perceptions’ (p. 13). Suimming up The terms of the corporate governance debate have shifted completely in recent years. The failure of the Japanese economy to grow, and the huge increases in unemployment seen in Germany, have suggested to many commentators that far from being the villain, Anglo-Saxon market based systems of governance may provide the more efficient set of vehicles for investment and growth. In this concluding section, we consider the implications of some of the micro-economic considerations we have been addressing—looking at the wider implications for economic performance. The significance of corporate governance for national economic performance stems from the interaction between two sets of considerations. The extent to which managerial motivation diverges significantly from those of owners, and the extent to which information asymmetries allow any such divergence to impact upon actual behaviour. The importance of either set of factors is not an immutable constant, but is likely to be constantly changing and evolving. On a priori grounds however, we might expect the importance of informational problems to be growing in the advanced economies as a greater proportion of investment is based on technological opportunity. On the other hand, growing product market competition may be reducing the scope for managerial discretion. Looking at what may happen to the pattern of investment, we cannot expect the impact of either of these factors to be neutral with respect to the governance patterns that are in place. Moreover, considering investment, they may impact on either its quantity or its quality. In our review of the relevant literature we have seen that there are clear differences between alternative governance structures across the four economies upon which we have focused. The most important consideration, as emphasised originally by Berle and Means, was the extent of dispersion of ownership in the modern corporation. The central problem in a dispersed ownership structure resides in the lack of incentive for owners to provide what is essentially a public good—the effective monitoring of management. By contrast the major alternative in the Anglo-Saxon economies—that the possibility of gain through hostile take-over acts as an effective monitoring device—has been shown to be both theoretically problematic and empirically hard to sustain. This is because of the substantial costs involved in any actual take-over. However, we should note that the threat of take-over could be highly effective in constraining managerial behaviour (if difficult to demonstrate); and here there is certainly some evidence that stock-market valuations are important for investment in the Anglo-Saxon economies. On these grounds it might be thought that the prevalence of concentrated ownership structures—most clearly exemplified in Germany, but also to a lesser extent in Japan—would constitute a clear advantage. However, we noted that there is a problem in unravelling the ownership structure in these economies because of the extent of cross-ownership. Monitoring activities may therefore be less effective in these economies than commonly supposed. Moreover, the widely held view of the special role of the banks in Germany and Japan (inter alia as particularly effective monitors) has recently been challenged. This means that there is no presumption that these economies are any less prone to the implications of managerial discretion. But

SIMON PECK AND PAUL TEMPLE

77

what do we really know of managerial motivation and the implications of managerial discretion? A number of writers have suggested that it may may entail an attempt by companies to grow faster which will favour more rapid macro growth rates. Moreover, as Morris (1998) for one has observed—if social rates of return to investment are above private rates of return, then managerial discretion may lead an economy closer to a welfare optimum. However, the recent experience of the Japanese economy (which has stagnated in the 1990s) suggests that this result can not be assumed to hold indefinitely. In particular, it may have depended upon the importance of ‘inter-temporal’ spillovers, where buoyant investment by a firm tends to create more favourable investment opportunities in the next. In any event what is less open to dispute is that the German and Japanese systems, because of the tendency for owners to be ‘locked in’ to their investments, will favour investments where the gains accrue further into the future and which tend to involve higher degrees of firm specificity. The question of whether this involves excessive long-termism or whether AngloSaxon governance systems generate excessive short-termism is less clear, although of course both may be possible. The debate has not produced clear cut conclusions here, although there is evidence for at least some degree of short-termism in stock markets. Higher levels of firm specific investment behaviour in Germany and Japan also tend to encourage internal, organic growth of the firm, rather than growth through merger and acquisition. This will favour the competitiveness of firms in their own particular sectors. In general, in a world where retained earnings are particularly important for the resources made available for investment purposes, this may lead to a diminished capability of the capital market to shift resources across sectors and into new innovative ones. While such a case could be made out for Germany, it would be less easy to make for Japan, which did manage to shift resources into more dynamic sectors (such as electronics) after the oil shocks of the 1970s. Ultimately however what is important from a welfare perspective is the impact that alternative governance forms have on the macro picture. This cannot be assessed appropriately at the micro level, comparing firms with given governance characteristics, since the choice of governance system is at least partly endogenous. In Germany for example, the fact that we do not observe more AGs may be precisely because the co-determination system is unpopular. At the macro level, in comparing the four economies, a number of empirical features stand out. As far as the quantity of investment is concerned, Germany and Japan have devoted consistently higher shares of GDP to investment. But the question of the quality of investment needs also to be addressed. One way of looking at this question in the context of the governance debate is to look at the ability of economies to generate ‘information intensive’ investments, i.e. additions to the capital stock which are prone to the kind of information asymmetry described above. A helpful distinction was made in this regard between ‘defensive’ investments on the one hand, and ‘enterprise’ investment on the other by Lamfalussy (1959). He distinguished between investments which were primarily aimed at maintaining existing markets through cost cutting and productivity enhancement (and which were comparatively free from information related problems), and those which were aimed at creating new markets through innovation and product development. Such investments frequently involve substantial sunk costs. Moreover, there is evidence that, in relation to Anglo-Saxon systems of governance, both the quantity and the quality of investment may be suboptimal because of the inability of firms to convey strategic information. On the other hand, perhaps in the US case rather than in the UK, the ability of owners to diversify risk may help in the spawning of new technology based sectors. Something similar may have happened in the Japanese economy in the 1970s and 1980s, aided by the ability of firms to co-operate on projects across industries. In Europe, on the other hand, precisely how the profits of mature industries can be made available for investment projects in newer, technology based industries may constitute a problem. The point remains, however, that both major types of governance system are able to produce means of compensating for their deficiencies, e.g. cooperative research ventures in the case of Germany and Japan and venture capital firms in the case of the US or UK. Perhaps herein is the main lesson for policy-makers in a world where wholesale change in corporate governance may be difficult to implement. Notes 1 It is worth noting however that the general question of goal compliance does not disappear in the entrepreneurial firm, where the owner manager has to engage employees. The question of ‘control loss’ within an organisation, and the efficiency properties of different employment regimes, are not within the scope of this chapter, except in so far as employee motivations impinge upon managerial motivation, as we shall see below. 2 Formally, let V(t) be the value of the firm at time t, which is given by the future discounted value of net profits.

In steady state conditions, and with an arbitrary initial capital stock, K(t)=K(0)e(gt), so that

78

CORPORATE GOVERNANCE AND PERFORMANCE

On integrating 3 The sample sizes were approximately 500 companies for the UK and Germany, 1000 for Japan, and 1500 for the US.

References Abbegglen, J. and Stalk, G. (1985) Kaisha, the Japanese Corporation, New York: Basic Books. Aoki, M. (1990) ‘Toward a Japanese model of the firm’, Joumal of Economic Literature, 28, March, 1–27. Berle, A.A. and Means, G.C. (1932) The Modern Corporation and Private Property, New York: Macmillan. Blundell, R., Bond, S., Devereux, M. and Schianterelli, F. (1992) ‘Investment and Tobin’s Q: evidence from company panel data’, Journal of Econometrics, 51(1), 233–258. Bond, S. and Meghir, C. (1994) ‘Dynamic investment models and the firm’s financial policy’, Review of Economic Studies, 61(2), 197–222. Borio, C.E.V. (1990) ‘Leverage and financing of non-financial companies: and international perspective’, Bank for International Settlements, working paper 27, Basle. Cable, J.R. (1985) ‘Capital market information and industrial performance: the role of West German banks’, Economic Journal, 95, 118–132. Chirinko, R.S. and Elston, J.A. (1997) ‘Finance, control, and profitability: the influence of German banks’, mimeo Berlin: Berlin Institute for Economic Studies. Cho, M. (1998) ‘Ownership structure, investment and the corporate value: an empirical analysis’, Journal of Financial Economics, 47, 103–121. Coates, J., Davies, E.W. and Stacey, R.J. (1994) ‘Managerial short-termism in multinational companies: the role of distributed corporate performance measurement system’, paper presented to 17th annual conference of the European Accounting Association. Confederation of British Industry (1987) Investing for Britain’s Future: Report of the City/Industry Task Force. London. Conyon, M.J. and Peck, S.I. (1998) ‘Recent developments in UK corporate governance’, in Buxton, T., Chapman, P. and Temple, P. (eds) Britain’s Economic Performance, 2nd edn, London: Routledge. Conyon, M.J. and Peck, S.I. (1998a) ‘Board size and corporate performance: evidence from European countries’, European Journal of Finance, 4(3), 291–304. Conyon, M.J. and Schwalbach, J. (1997) ‘European governance: executive pay and performance in Europe’, in Carpenter, J. and Yermack, D. (eds) Executive Compensation and Shareholder Value in Europe, Berlin: Kluwer. Cubbin, J. and Leech, D. (1983) ‘The effect of shareholding dispersion on the degree of control in British companies: theory and evidence’, Economic Journal, 93, 351–369. Demirag, I.S. (1994) ‘An empirical study of research and development managers’ perceptions of short-term pressures from capital markets in the UK’, Critec Discussion Paper, Sheffield University Management School 8. Demirag, I.S. (1995) ‘Assessing short-term perceptions of group finance directors of UK Companies’, British Accounting Review, 27, 247–281. Devereux, M. and Schianterelli, F. (1990) ‘Investment, financial factors and cash flow’, in R.Glenn Hubbard (ed.) Asymmetric Information, Corporate Finance, and Investment, Chicago: University of Chicago Press. Dickerson, A.P. and Gibson, H.D. (1995) ‘Short-termism and underinvestment: the influence of financial systems’, Manchester School, 351–367. Edwards, J. and Fischer, K. (1994) Banks, Finance, and Investment in Germany, Cambridge: Cambridge University Press. Elston, J.A. and Albach, H. (1995) ‘Firm ownership structure and investment: evidence from German manufacturing’, IFO Studien zeitschrift fur Empirische Wirtschaftforschung, Heft 1. Fazzari, S.M., Hubbard, R.G. and Petersen, B.C. (1988) ‘Financing constraints and corporate investment’, Brookings Papers on Economic Activity, 1, 141–195. Franks, J. and Mayer, C. (1992) ‘Corporate control: a synthesis of the international evidence’, London Business School, IFA working paper no 165–92. Grinyer, J., Russell, A. and Collison, D. (1998) ‘Evidence of managerial shorttermism in the UK’, British Journal of Management, 9, 13–22. Grossman, S. and Hart, O.E. (1977) ‘On value maximisation and alternative objectives of the firm’, Journal of Finance, 32, 389–440. Grossman, S. and Hart, O.E. (1980) ‘The costs and benefits of ownership: a theory of vertical and lateral integration’, Journal of Political Economy, 94, 691–719. Hart, O.E. (1995) ‘Corporate governance: some theory and implications’, Economic Journal, 105, 678–689.

SIMON PECK AND PAUL TEMPLE

79

Hoshi, T., Kashyap, A. and Scharfstein, D. (1990) ‘Bank monitoring and investment: evidence from the changing structure of Japanese corporate banking relationships’, in Glenn Hubbard, R. (ed.) Asymmetric Information, Corporate Finance, and Investment, Chicago: University of Chicago Press. Hoshi, T., Kashyap, A. and Scharfstein, D. (1991) ‘Corporate structure, liquidity, and investment: evidence from Japanese industrial groups’, Quarterly Journal of Economics, 106, 33–60. Hubbard, R.G. (1997) ‘Capital-market imperfections and investment’, National Bureau of Economic Research Working Paper 5996, May. Hutton, W. (1995) The State We’re In, London: Jonathan Cape. Jaffee, D. and Russell, T. (1976) ‘Imperfect information, uncertainty, and credit rationing’, Quarterly Journal of Economics, 91(4), 651–666. Jensen, M. (1986) ‘Agency costs of free cash flow, corporate finance and take-overs’, American Economic Review, 76, 323–329. Jensen, M.C. (1993) ‘The modern industrial revolution, exit, and the failure of internal control systems’, The Journal of Finance, 48, 831–880. Kaplan, S.N. and Zingales, L. (1997) ‘Do investment-cash flow sensitivities provide useful measures of financing constraints’, Quarterly Journal of Economics, 112(1), 169–215. Kathuria, R. and Mueller, D.C. (1995) ‘Investment and cash flow: asymmetric information and managerial discretion’, Empirica, 22, 211–234. Kester, W.C. (1992) ‘Industrial groups as systems of contractual governance’, Oxford Review of Economic Policy, 8(3), 24–44. Lakonishok, J., Schleifer A. and Vishny, R.W. (1991) ‘Window dressing by pension fund mangers’, American Economic Review, Papers and Proceedings, 81, 227–231. Lamfalussy, A. (1959) Investment and Growth in Mature Economies, London: Macmillan McConnell, J. and Muscarella, C. (1985) ‘Corporate capital expenditure decisions and the market value of the firm’, Journal of Financial Economics, 14, 399–422. Marris, R.L. (1964) The Economic Theory of Managerial Capitalism, London: Macmillan. Marris, R.L. (1972) ‘Why economics needs a theory of the firm’, Economic Journal, special issue in honour of E.A.G.Robinson. Marris, R.L. (1991) Reconstructing Keynesian Economics with Imperfect Competition, London: Edward Elgar. Marsh, P. (1990) Short-termism on Trial, Institutional Fund Managers Association, London. Marston, C.L. and Craven B.M. (1994) ‘An empirical study of corporate perceptions of short termism in large UK corporations’, paper presented to 17th Annual Congress of the European Accounting Association, Venice, 1–19 April. Mayer, C. (1990) ‘Financial systems, corporate finance, and economic development’, in Glenn Hubbard, R. (ed.) Asymmetric Information, Corporate Finance, and Investment, Chicago: University of Chicago Press. Mayer, C. (1997) ‘Corporate governance, competition and performance’, Journal of Law and Society, 24, 152–176. Mayer, C. and Alexander, I. (1990) ‘Banks and securities markets: corporate financing in Germany and in the UK’, CEPR discussion paper. Miles, D. (1993) ‘Testing for short-termism in the UK stock market’, Economic Journal, 103, 1379–1396 Morck, R., Shleifer, A. and Vishny, R. (1988) ‘Management ownership and market valuation: an empirical analysis’, Journal of Financial Economics, 20, 293–315. Morris, D.J. (1998) ‘The stock market and problems of corporate control in the UK’, in Buxton, T., Chapman, P. and Temple, P. (eds) Britain’s Economic Performance, London: Routledge. Mueller, D.C. (1972) ‘A life cycle theory of the firm’, Journal of Industrial Economics, 21, 199–219. Mullins, M. and Wadhwani, S. (1989) ‘The effects of the stock market on investment’, European Economic Review, 33, 939–961. Myers, S.C. and Majluf, N. (1984) ‘Corporate financing and investment decisions when firms have information that investors do not have’, Journal of Financial Economics, 13(2), 187–221. Nickell, S.J. (1995) The Performance of Companies: The Relationship between the External Environment, Management Strategies and Corporate Performance, Oxford: Blackwell. Nickell, S. and Wadhwani, S. (1987) ‘Myopia, the dividend puzzle and share prices’, Centre for Labour Economics discussion paper 272, LSE. Odagiri, H. (1981) Theory of Growth in a Corporate Economy: Management Preference, Research and Devlopment and Economic Growth, Cambridge: Cambridge University Press. Odagiri, H. (1992) Growth Through Competition, Competition Through Growth: Strategic Management and the Economy in Japan, Oxford: Clarendon Press. Office of the Chief Economist, Securities and Exchange Commission (1985) Institutional Ownership, Tender Offers and Long Term Investment, New York. Peck, S. and Temple, P. (1996) ‘Understanding competitiveness: the determinants of industrial performance in the OECD economies, 1973– 1992’, London Business School, CBS Working Paper Series 173. Penrose, E. (1959) The Theory of the Growth of the Firm, Oxford: Basil Blackwell. Porter, M.E. (1990) The Competitive Advantage of Nations, London: Macmillan. Porter, M.E. (1992) Capital Changes: Changing the Way America Invests in Busmess, Washington, DC: Council on Competitiveness; Boston: Harvard Business School Press. Prowse, S. (1994) ‘Corporate governance in an international perspective’, Bank for International Settlements Economic Papers, 41, July. Ravenscraft, D. and Scherer, F.Mergers, Sell offs and Economic Efficiency., Brookings Institution. Schreyogg, G. and Steinman, H. (1981) ‘Zur Trennung von Elgentum und Verfugungsgewalt—Eine Empirische Analyse der Beteilgungsverhaltnisse in Deutscher Grossunternchuen’, zeitschrift fur Betriebs-Wirtschaft, 51, 533–556.

80

CORPORATE GOVERNANCE AND PERFORMANCE

Sharp, M. (1998) ‘Technology policy’, in Buxton, T., Chapman, P. and Temple, P. (eds) Britain’s Economic Performance, London: Routledge. Sheard, P. (1989) ‘The Main Bank system and corporate monitoring and control in Japan’, Journal of Economic Behaviour and Organisation, 11(3), 399–422. Shleifer, A. and Vishny, R. (1990) ‘Equilibrium horizons of investors and firms’, American Economic Review, Papers and proceedings, 89, 148–153. Short, H. (1994) ‘Ownership, control, financial structure and the performance of firms’, Journal of Economic Surveys, 8(3), 203–225. Solow, R.M. (1971) ‘Some implications of alternative criteria for the firm’, in Marris, R. and Wood, A. (eds) The Corporate Economy, London: Macmillan. Stein, J.C. (1988) ‘Takeover threats and managerial myopia’, Journal of Political Economy, 96(1), 61–80. Stiglitz, J.E. and Weiss, A. (1981) ‘Credit rationing in markets with imperfect information’, American Economic Review, 71(3), 393–410. Uzawa, H. (1969) ‘Time preference and the Penrose effect in a two class model of economic growth’, Journal of Political Economy, 77(4), 628–652.

6 Uncertainty, macroeconomic volatility1 and investment in new technology Otto Toivanen, Paul Stoneman and Paul Diederen

The central contribution of this chapter is an empirical analysis of the impact of macroeconomic volatility on the diffusion of industrial robotics using a panel data set encompassing 16 countries over the period from 1981 to 1993. There are very few international comparative studies of any kind in the diffusion field and even fewer concerned with the impact of macroeconomic volatility (although Davies (1979) does explore the impact of the business cycle on the diffusion process). However, rather than being limited to just an analysis of diffusion phenomenon we see this chapter as a contribution to a much more extensive academic and policy debate—that concerning the effects of uncertainty on investment. In the policy arena most industrialised nations since the 1980s have emphasised the need to generate within their economies a low and stable rate of inflation. To quote one policy maker (UK Chancellor Kenneth Clarke in his 1994 Budget speech, as quoted in the Financial Times 30.11.1994): ‘low inflation creates a climate of stability which encourages savings and investment…’ and ‘[in] that way [of low inflation] lies the virtuous circle of improved competitiveness, rising productivity, economic growth, low inflation, and more jobs’. The mechanism that is implicit in the argument is that price stability reduces uncertainty. In the academic field there is a large and growing theoretical literature on investment under uncertainty (for a survey, see Dixit and Pindyck (1994)) that shows that under plausible circumstances, increases in uncertainty in the economic environment will lead to, at least, a temporary reduction in investment. There is, however, very little empirical work testing the theoretical propositions (examples are Driver and Moreton (1991) and Pindyck and Solimano (1993)) and thus validating the policy stance. This chapter is a new empirical contribution but differs from the main body of the literature in exploring the relationship between uncertainty and investment in a particular new technology. Concentrating on a specific new technology rather than aggregating investment has two virtues. First, as investment in a new technology such as robots will probably entail technological uncertainty to a higher degree than investment in old and trusted technologies, it follows that the prediction of the policy doctrine (and the theoretical models) on the sign of the investmentuncertainty relationship should hold more strongly in such a case than for investment in general. Second, though important, robot adoption is very likely to have a negligible impact on the current levels of macroeconomic variables. This means that current and past values of macroeconomic variables can be considered predetermined and safely used as explanatory variables in econometric analysis. What one cannot pick up in analysing investment in a particular technology is those aspects of the investment decision that reflect competition between technologies. We do not however consider that to be a significant issue in this context. The chapter proceeds as follows: in the next section, we discuss theoretical underpinnings and previous empirical studies. We then discuss the measurement of uncertainty and the relationship between uncertainty and macro-economic volatility. The next section is devoted to the specification of an appropriate diffusion model. We discuss how to allow for the two different routes through which uncertainty can affect the diffusion path: that is, by affecting the equilibrium2 level of the robot stock or the adjustment speed towards this level. This is an important specification issue which the previous diffusion (and investment) literatures seem to have ignored. The data is then presented along with results, and we reach our conclusions and discuss policy implications. The existing literature Driver and Moreton (1991) argue that the traditional literature on the impact of uncertainty on investment can be reduced to three main approaches (a fourth refers to the real options literature) dealing respectively with: attitudes to risk, technological nonlinearities and the marginal cost of uncertainty. Each of these approaches predicts a negative investment-uncertainty relationship. Measuring uncertainty by the standard deviation of 12 forecasts of GDP growth and inflation and using cointegration techniques, they find that growth uncertainty has a negative impact on UK aggregate manufacturing investment both in the short and long run, but inflation uncertainty has a negative impact only in the short run. The real options literature concentrates on investments that are irreversible: once the option is exercised, it cannot be reversed.3 Although the general investment literature considers asymmetric adjustment costs, irreversible investment is

82

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

asymmetric to the extreme: costs of dismantling are infinity whereas costs of increasing capacity are merely positive (and possibly stochastic). As a result, for investment to take place in real options models the discounted net revenue has not only to exceed discounted costs but also the cost of waiting. This leads to hurdle rates that can be significantly higher than those predicted by traditional investment theories.4 In such real options models it also matters whether the uncertainty the firms in an industry face is idiosyncratic or aggregate. Idiosyncratic shocks cause an asymmetry if the marginal profitability of capital is convex in the stochastic variable, whereas aggregate shocks always lead to asymmetry. The reason for the latter effect is that positive shocks cause more entry, and hence negative shocks will decrease profitability more than positive shocks will increase it. Real options models do not usually, however, allow straightforward predictions to be derived on the impact that uncertainty has on the long-run equilibrium levels of the capital stock (see Pindyck and Solimano (1993) for a discussion of this). They do however predict that an increase in uncertainty will result in a temporary decrease in investment. Translated into a model of technology adoption an increase in uncertainty should thus slow down the diffusion of new technology. Pindyck and Solimano (1993) attempt to estimate the effects of uncertainty on (aggregate) investment thresholds using a 30 country panel encompassing 16 OECD countries and 14 LDCs. Their empirical approach departs radically from the one adopted by Driver and Moreton (1991) using measures of marginal profitability and its volatility to reflect uncertainty. They find the effect of the latter on investment thresholds to be ‘negative but moderate in size’ (Pindyck and Solimano 1993, p. 297). For a critique of their empirics see Eberly (1993) and Hall (1993). Caballero (1991) also studies (theoretically) the sign of the investment— uncertainty relationship. He concludes that in general the sign is ambiguous: imperfect competition, decreasing returns to scale and asymmetric adjustment costs lead to a negative relationship. However none of these is sufficient alone. The literature on the adoption of new technology under uncertainty (surveyed by Reinganum (1989) and Stoneman and Karshenas (1995)) shows that under plausible circumstances, increases in uncertainty in the economic environment will lead to at least a temporary reduction in investment in new technologies. For example, in many models of the diffusion of new technology, firms are allowed to search for more information and thereby to gain a better understanding of whether or not they will profit from adoption. The poorer the initial information and the slower the process of information gathering, the slower will be diffusion (although diffusion itself by generating new information will reduce uncertainty). Jensen (1982) concludes that the duration of diffusion will be longer in an industry where initial price beliefs are more diverse. It seems natural to assume that a more uncertain environment would lead to ‘more diverse’ initial beliefs and to signals of poorer quality. Such models thus give a prediction that there will be a negative relationship between uncertainty and diffusion. The broad thrust of the existing theoretical literature is thus that investment, either in aggregate or in individual new technologies, should be negatively related to uncertainty. There is some limited empirical support for this proposition as it relates to aggregate investment but to date no empirical results at the level of individual technologies. Uncertainty and volatility A key issue in the empirical analysis of the relationship between investment and uncertainty is the measurement of uncertainty and in particular the relationship between uncertainty and volatility (i.e. the variance of one or many macroeconomic indicators). Although most of the theoretical literature concentrates on uncertainty, when it comes to empirical work uncertainty is a difficult concept to measure per se. Instead the empirical work tends to concentrate on volatility in the macroeconomic environment. This is consistent with both the real options literature and microeconomic theory (Rothschild and Stiglitz (1970)) where increases in volatility will lead to greater uncertainty.5 From the point of view of validating policy stances however, it would appear that the major issue is whether volatility per se affects investment. The policy approach is very much one that emphasises volatility as a determinant of uncertainty and thus the volatility of the macroeconomic environment itself would be the appropriate variable to measure. This immediately raises two further issues: what types of volatility matter and over what period. Policy pronouncements tend to emphasise price volatility but in the limit the important variable for the investor is the expected return on investment. It is thus necessary to allow for as many different sources of volatility in the expected return as possible, especially when, as in our case, the calculation of the volatility of the profitability of the investment is beyond the data at hand. The period over which volatility should be measured depends crucially upon the extent to which the past is used to predict the future and this is an empirical issue. Driver and Moreton (1991) measure uncertainty by the standard deviation of 12 forecasts of GDP growth and inflation. By using the standard deviation of forecasts they avoid issues of using the past to predict the future. Pindyck and Solimano (1993) use the volatility (of marginal profitability) to measure uncertainty. However we have neither consistent forecasts for all countries in our data set nor data on the exact application of robots and are thus unable to use either of these two approaches specifically. In empirical macroeconomics, researchers have used uncertainty variables that are based on surveys (Davis and Kanago (1996)) however no such (consistent) survey data is available for all countries in our sample.

TOIVANEN, STONEMAN AND DIEDEREN

83

An alternative approach is to argue that volatility itself is not in fact a good measure of uncertainty, but, rather, a measure of uncertainty should reflect the extent to which the path of macroeconomic variables cannot be predicted. If volatility can be forecast precisely there is no uncertainty. The empirical macroeconomic literature has incorporated these ideas in ARCH models (see, for example, Engle (1982) and Bollerslev (1986)). This approach produces forecasts for given variables and an estimate of the variance of the forecast. It is this estimate of the variance of the forecast that can be used as a measure of uncertainty. In the empirical work reported upon below we allow for as many different sources of uncertainty as possible by including a variety of different macroeconomic indicators. For each we derive indicators of uncertainty experimenting with two approaches. The first is the ARCH model approach. The second proxies uncertainty more directly by volatility measures and as shown above there is some track record for the use of volatility in this way. Specifically for each indicator (except for output prices where we measure volatility by the actual inflation rate in order to get as close as possible to the policy doctrine) we use the standard deviation of the variable over the previous T years divided by the absolute value of its average, calculated over the same period.6 This measure of volatility captures the variation in relation to the absolute size of the variable. As an example, if the long-run average real growth rate in country A has been 10 per cent and in country B 2 per cent with a standard deviation of 2 in both, then our measure of volatility ensures that volatility in country B is many times higher than in country A. One should also note that such measures are always non-negative with upward and downward movements in a variable being considered equally important in measuring volatility. Specification of the empirical model To empirically explore effectively the relationship between uncertainty or macroeconomic volatility and investment one needs either a long run of single country data or a cross-country data set. Although internationally comparative data on the diffusion of new technologies is in fact quite rare, the UN and the International Federation of Robotics (IFR) compile international statistics on the use of industrial robots covering 28 countries and it is this data that is used here. The data is available only at the macro (country) level and thus the empirical analysis must proceed at this level. Although there is a growing literature on the specification of micro level diffusion models (see, for example, Stoneman and Karshenas (1995)) there is only a limited variety of models that have been specified to encompass macro diffusion phenomena, these being primarily of the ‘epidemic’ type (see for early examples, Griliches (1957), Mansfield (1968), Chow (1967) and Dixon (1980)). After some experimentation we have chosen to concentrate on the Gompertz7 variant of the basic model. Defining St as the stock of installed robots at the end of period t and as the equilibrium stock of robots in time t, the Gompertz diffusion curve is specified as equation (1) where the coefficient g can be interpreted as the adjustment speed at which the actual stock approaches the equilibrium stock. (1) This specification is closely related to stock adjustment models of investment (see Nickell (1978)) and can in fact be reproduced by assuming quadratic adjustment costs in such models. The variable ln St on the right hand side is usually proxied by the log of the lagged stock and we follow this practice. The key concern of this chapter is whether uncertainty (volatility) affects the diffusion process. Uncertainty can potentially impact upon (1) in two ways, either upon the equilibrium stock ( ) or upon the adjustment speed g. The literature gives little guidance as to which is to be preferred either theoretically or empirically. The most general specification would allow all variables to affect both the adjustment speed and the equilibrium stock. The size of our data set prevents the estimation of such a model. Similarly the size of our data set prevents us estimating a model in which only the uncertainty variables are introduced into both. We thus proceed by including the uncertainty (volatility) variables only as determinants of g (as opposed to ) and then proceed to test for the validity of this procedure. Robots are used in several industries, and perform a variety of tasks. World-wide, the traditional ‘vehicle’ for diffusion of robots has been the transport equipment industry (especially the motor vehicle industry), but lately, e.g. in Japan, the electrical machinery industry has adopted more robots. The major application areas are welding, machining and assembly, with the leading application area varying over countries. We thus assume that is positively related to manufacturing output, reflecting the total potential area of application. We also allow that is positively related to GDP growth (reflecting general macroeconomic conditions) and negatively related to the price of robots and the real interest rate (reflecting the cost of robots),8 and thus specify that , where Xt is a vector of the aforementioned variables. The hypothesis that uncertainty (volatility) affects the adjustment speed is equal to assuming that g is a function of different uncertainty (volatility) variables i.e. g=g(VOL), where VOL is a vector of uncertainty (volatility) variables. Specifically, in line with arguments above favouring as wide a spread as possible of potential uncertainty indicators, we include uncertainty

84

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

(volatility) indicators relating to five variables—the general price level, the rate of inflation, GDP growth, the real interest rate9 and robot prices. The general price level (i.e. the inflation rate) is included in this list largely to enable us to test the prevailing policy doctrine. We assume that g is a multiplicative function but because the inflation rate (the one volatility variable not measured as other such variables) at times takes negative values, we enter its exponential rather than its level in order to preclude the log of a negative variable entering the estimating equation. We also assume that all variables enter S* in a multiplicative fashion, however as the real interest rate and GDP growth at times take negative values we again enter their exponentials rather than levels.10 Using subscript c for country and subscript t for time, and defining α1 as constants, we can then rewrite equation (1) as (1′) (1′) Using the lagged robot stock as a proxy for Sc,t and assuming an additive i.i.d. error term εc,t yields the following estimating equation. (1′′) where for simplicity (but in slight abuse of notation) the vectors X and VOL contain both linear and logarithmic terms (as explained above). Estimating (1′′) with OLS yields consistent and efficient estimates. To achieve a consistent measure over countries, manufacturing output and robot price levels are measured as indices and thus to correct for differences across countries in the bases of the indices we allow the constant term α1α2 to differ across countries, i.e. α1α2=α3+αc. These country specific effects will also pick up other country specific omitted variables (e.g. different institutional settings). A test of the hypothesis that the uncertainty (volatility) variables affect the adjustment speed rather than the equilibrium stock consists of two parts: the first is a test on the joint significance of the coefficients of the interaction terms between the volatility and level variables and between the volatility variables and the lagged robot stock, the second a series of tests of coefficient restrictions. Using such a highly structured model leads to a number of such restrictions, as is evident from equation (1′′). These are of two types. First, there are 20 restrictions that involve the coefficients of the variables in vector X, the coefficient of the interaction terms between the lagged stock and variables in the VOL vector, the coefficient of the lagged stock (which is equal in absolute value to the constant term of the adjustment speed function g) and the interactions between the variables in the two vectors of explanatory variables. These are of the form: (2) Second, there are five restrictions that involve the constant term, the coefficients of the interaction terms between the lagged stock and variables in the VOL vector and the coefficient of lagged stock and the coefficients of the variables in the VOL vector. These take the form: (3) In the empirical analysis below we test the validity of these restrictions. Data sources and data description The UN and International Federation of Robotics (IFR) statistics cover 28 countries, 16 of which were included in our sample.11 As different countries started adopting robots at different dates, the panel is unbalanced. The sixteen countries are listed together with summary statistics in Table 6.1. IFR has created an international standard (ISO TR 8373) that was used in compiling the statistics.12 Robots are customarily classified into ‘standard’ and ‘advanced’ and by application area. However in the period 1981 to 1993 only aggregate data is available. It is estimated (the statistics are not watertight) that there were some 610,000 robots in use world-wide at the end of 1993. The rate of growth of this stock has been fast, ranging from 16–23 per cent p.a. at the beginning of the 1980s to 6–8 per cent p.a. in the early 1990s. Quantitatively robot adoption is extensive, but differs from aggregate manufacturing investment. In 1993 robot investment amounted to 12 per cent of total machine tool investments (not manufacturing investments) in the US, to 11 per cent in both Germany and the UK and to 6 per cent in France. Japan is by far the largest user of robots whether measured by absolute (some 60 per cent of world stock) or relative numbers (in 1993 Japan had 264 advanced robots per 10,000 employees in

TOIVANEN, STONEMAN AND DIEDEREN

85

manufacturing when the country with the second highest density (Singapore) had 61). Although it would certainly be beneficial to have more detailed country-level data on the composition of the robot stock, and its use, such country-level idiosyncrasies are to a great extent constant over time, and will thus be captured by the country (fixed) effects included in our econometric specification. Robot price data are not available on an individual country basis but are available for both Germany and Italy. The results that we present are based on the German data. These prices were converted into real domestic prices for each country using the respective consumer price index13 and the yearly (average) exchange rates as reported in the IMF’s International Financial Statistics γearbook. It should be noted that this conversion implies that our measure of the volatility of robot prices also captures exchange rate volatility. Data on manufacturing output, GDP growth, the real interest rate, inflation, and wages in manufacturing were sourced from the IFS statistics with any gaps being filled via the use of alternative sources.14 Although we have for good reasons attempted to use many macro indicators of uncertainty it is possible that these may be endogenous to one another. However in our sample, the correlations between the different variables and between measures of their volatility turned out to be small.15 The uncertainty (volatility) variables were calculated according to the principles discussed earlier, i.e. from both ARCH models and more directly as VOLi,t=STDi,t…, t- indicators of uncertainty. It is possible that these latter may be endogenous to one another. However in our sample, the correlations between the different variables and between measures of their volatility turned out to be small.16 Table 6.1 Country-wise descriptive statistics Country

γears

γearly change in robot stock

Robot price (P) 1978=100

Volatility of P Manufacturing output 1978=100

Growth of GDP (DG)%

Volatility

Australia Austria Denmark Finland France Germany Italy Japan Norway Singapore Spain Sweden Switzerland Taiwan UK USA

1985–93 1982–93 1982–93 1982–93 1988–93 1982–93 1982–93 1982–93 1983–93 1982–93 1983–93 1982–93 1983–93 1983–93 1982–93 1982–93

153.22 154.25 47.33 93.75 1236.50 3451.25 1593.17 28921.17 41.27 319.67 354.1 306.42 220.64 247.18 623.00 3666.67

351.54 214.67 301.51 260.03 331.79 220.67 371.95 171.35 308.21 186.31 380.61 334.8 213.53 192.12 267.37 224.79

0.16 0.12 0.14 0.16 0.17 0.12 0.17 0.11 0.16 0.21 0.17 0.17 0.12 0.15 0.16 0.15

2.88 1.97 2.15 1.55 2.08 2.04 2.06 5.12 2.79 7.01 2.85 1.00 1.72 8.47 2.22 2.37

0.43 0.57 1.09 2.51 0.41 6.68 0.49 0.22 1.11 0.54 0.38 0.89 1.14 0.34 3.24 1.59

95.41 119.27 131.21 143.97 116.83 114.00 117.36 144.25 163.9 191.16 114.24 116.55 119.17 193.86 108.09 121.73

Country

γears

Real interest rate (RR)%

Volatility of RR

Inflation (I)

Volatility of I

Robot stock

Australia Austria Denmark Finland France Germany Italy Japan Norway Singapore Spain Sweden Switzerland

1985–93 1982–93 1982–93 1982–93 1988–93 1982–93 1982–93 1982–93 1983–93 1982–93 1983–93 1982–93 1983–93

5.92 4.56 7.00 6.99 5.85 3.81 5.56 3.60 5.96 5.57 6.90 6.39 0.96

0.25 0.19 0.17 0.60 0.08 0.25 0.75 0.18 4.29 0.78 0.47 18.27 4.55

0.06 0.03 0.04 0.05 0.03 0.03 0.09 0.02 0.05 0.00 0.07 0.05 0.0.4

0.32 0.29 0.20 0.24 0.17 0.58 0.17 0.52 0.21 2.49 0.17 0.86 0.28

1304.67 771.00 327.92 549.75 8949.33 19715.00 8572.92 184026.42 438.91 1060.25 1875.3 2949.67 1016.82

86

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

Country

γears

Real interest rate (RR)%

Volatility of RR

Inflation (I)

Volatility of I

Robot stock

Taiwan 1983–93 7.24 1.24 −0.01 1.36 980.55 UK 1982–93 4.91 0.88 0.05 0.29 4625.50 USA 1982–93 4.50 1.91 0.04 0.27 29383.33 Note: All numbers are country-wise averages of the respective variables where the number of observations=no. of years the country is in the sample (i.e. max.=11, 1982–1993).

The uncertainty (volatility) variables were calculated according to the principles discussed, i.e. from both ARCH models and more directly as

Table 6.1 reveals that there is considerable variation in the macrovariables across countries. The mean of GDP growth varies between 8.04 per cent (Taiwan) and 1.00 per cent (Sweden). The mean of the real interest rate varies between 0.96 per cent (Switzerland) and 7.24 per cent (Taiwan) whereas the lowest mean inflation rate is found in Taiwan (–1 per cent) and the highest in Spain (7 per cent). The average yearly change in the robot stock varies between almost 30,000 in Japan to 41 in Norway. There are also considerable inter-country differences in the measures of volatility of the different variables, but the measures do not necessarily move in parallel. Empirical results Our basic model (1″) has a total of 35 explanatory variables when all interaction terms are taken into account, and 15 country dummies to control for country-specific omitted variables. Using the full sample, we have 179 observations. Our discussion above has led us to consider two basic uncertainty indicators. Those derived from ARCH models and those that more directly measure volatility. Although there exists a large literature on ARCH models and macrovariables, a consistent set of estimates of such models for all our variables of interest over all sample countries was not available. To assess this approach we thus gathered data for all the macrovariables of interest over as long a period as consistent data was available (often outside our sample period, the longest observation period that resulted was 1949 to 1994) and fitted ARCH models to each of the time series separately for each country. Details are provided in the appendix. We found using this data that less than 25 per cent of the forecasts generated exhibited a time-varying variance (i.e. conditional heteroscedasticity) the rest having homoscedastic error terms. Using the variance of the forecast as an indicator of uncertainty thus generates variables the larger majority of which do not vary over time within a country. As a result, the correlations between these uncertainty measures themselves and between them and the country fixed effects are high. Indeed, when estimating a simple diffusion model, we found that all the uncertainty terms based on ARCH models carried insignificant coefficients. We conclude that although theoretically such uncertainty measures may be relevant, our data is unable to separate their effect from country specific effects. The rest of our estimates are thus based upon the use of the more direct volatility measures of uncertainty. Given the above findings the country specific effects in this model must be interpreted (inter alia) as including and controlling for any effects of ARCH model uncertainty. The explanatory variables used and the expected signs of coefficients in these estimates are presented in Table 6.2. One problem with our approach is that we have to assume that the slope coefficients are the same across all countries. The large number of Table 6.2 Variable definitions Abbreviation

Expected sign Definition

LP RR LIND DG I

− − + + −

LS LIV, LRRV, LGV, LPV

−

Log of robot price index (1978=100) Real interest rate Log of manufacturing output index 1978=100 Annual percentage change in GDP Inflation rate measured as the annual percentage change in the consumer price index Log of the lagged robot stock Logs of the 3–year standard deviation of respective level variables divided by the absolute value of the 3–year average level

TOIVANEN, STONEMAN AND DIEDEREN

Abbreviation

Expected sign Definition

PI, PIV, PRRV, PGV, PPV RRI, RRIV, RRRRV, RRGV, RRPV INDI, INDIIV, INDRRV, INDGV, INDPV GI, GIV, GRRV, GGV, GPV SI, SIV, SGV, SRRV, SPY

? ? ? ? +

87

Interactions between robot price LP and volatility variables (I, …,LVP) Interactions between RR and volatility variables Interactions between LIND and volatility variables Interactions between DG and volatility variables Interactions between LS and volatility variables

explanatory variables and the small number of country-wise observations precluded the usual ways of controlling for this, namely: (1) running separate regressions for each country; (2) introducing country/country-group dummies and interactions; (3) estimating a random coefficients model. We have thus investigated the robustness of our results by using different data sets, excluding some countries from the sample and then comparing the results to those obtained using the full sample. An obvious outlier is Japan which has over 60 per cent of the world robot stock. Another potential group of outliers are the AngloSaxon countries (UK and USA) which have a different institutional (e.g. financial markets) environment to most other countries. Thus in Table 6.3 we present three sets of results: (1) for the full sample; (2) from a sample excluding Japan; and (3) from a sample excluding the UK and the USA. All estimations have been produced using OLS via LIMDEP 7. In general, the estimations work well in terms of overall performance and diagnostics. As can be seen from Table 6.3, R2’s are high throughout. Autocorrelation (of 1st order) is absent from all regressions as the Durbin— Watson tests show, but as is evident from the Breusch-Pagan heteroscedasticity tests, the data is heteroscedastic. We thus use a heteroscedasticity consistent variance–covariance matrix throughout. In all cases, we can reject the hypothesis that our macrovariables (i.e. all but LP, LS and their interactions) are jointly insignificant. Likewise, for all three estimations, the hypothesis that volatility variables affect the equilibrium stock rather than the adjustment speed is rejected. The interaction terms are always (jointly) significant. We tested restrictions (2) and (3) imposed by our specification by calculating Wald tests for each of the restrictions separately (the results of these tests are presented in the Appendix). We reject the null hypothesis for only 3 (4 when the UK and the USA are excluded) of the 25 restrictions. In addition, when testing the restrictions jointly, we are unable to reject the null for any of the three samples. We therefore accept the null hypotheses that the restrictions imposed by our model are not violated. The results are also very consistent over the different samples. Comparisons of the three sets of results reveal that only two coefficients in the estimation excluding the US and the UK, and four in the estimation excluding Japan have different signs than those obtained using the whole sample, and all of these are insignificant. The significance levels of coefficients also vary remarkably little over different samples: when comparing the full sample to that excluding the US and the UK, one significant coefficient turns insignificant, and four insignificant turn significant (at the 5 per cent level). The same applies to a comparison of the full sample with that excluding Japan. The coefficient on the lagged robot stock (LS) gives a direct estimate of the adjustment speed in a situation where there is no volatility as measured by our volatility variables. The coefficient is 0.198 for the full sample, 0.201 when excluding the USA and the UK, and 0.130 when excluding Japan. Table 6.3 Three sets of results Variable

(1) Full sample

(2) UK and US excluded

(3) Japan excluded

Constant LP RR LIND DG LS I LIV LRRV LGV LPV PI PIV PRRV PGV

1.936*** (0.096) −0.237* (0.1 35) −0.025*** (0.010) 0.180** (0.092) 0.016 (0.013) −0.198*** (0.026) −9.249 (5.745) 0.077 (0.425) 0.235 (0.252) −0.263 (0.248) −0.131 (0.250) 1.254 (1.073) 0.023 (0.039) −0.063 (0.028) 0.014 (0.028)

2.211*** (0.132) −0.293** (0.148) −0.020** (0.010) 0.179** (0.084) 0.008 (0.014) −0.201*** (0.028) −12.267** (6.152) −0.134 (0.487) 0.349 (0.300) −0.339 (0.303) −0.051 (0.256) 1.728 (1.158) 0.039 (0.046) −0.070** (0.029) 0.166 (0.031)

2.190*** (0.097) −0.404*** (0.147) −0.020** (0.010) 0.175** (0.084) 0.018 (0.014) −0.130*** (0.032) −11.262** (5.477) 0.002 (0.394) 0.219 (0.234) −0.245 (0.234) 0.084 (0.247) 2.080* (1.108) 0.008 (0.034) −0.048* (0.026) −0.019 (0.034)

88

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

Variable

(1) Full sample

(2) UK and US excluded

(3) Japan excluded

PPV −0.002 (0.051) −0.014 (0.052) −0.060 (0.051) RRI −0.052 (0.056) −0.068 (0.058) −0.079 (0.056) RRIV −0.010** (0.004) −0.013** (0.005) −0.011** (0.004) RRRRV 0.005* (0.003) 0.009** (0.003) 0.006* (0.003) RRGV 0.004 (0.003) 0.005 (0.003) 0.003 (0.003) RRPV −0.010** (0.005) −0.008 (0.005) −0.008 (0.005) INDI −0.240 (0.158) −0.259 (0.159) −0.257* (0.151) INDIV −0.025 (0.059) 0.010 (0.065) −0.005 (0.061) INDRRV 0.061 (0.049) 0.040 (0.052) 0.051 (0.044) INDGV 0.004 (0.050) 0.014 (0.054) 0.027 (0.052) INDPV 0.020 (0.028) 0.016 (0.028) 0.010 (0.027) GI 0.173 (0.135) 0.193 (0.142) 0.207 (0.133) GIV −0.001 (0.004) −0.001 (0.005) 0.001 (0.004) GRRV −0.009** (0.005) −0.012** (0.005) −0.008** (0.004) GGV −0.002 (0.003) −0.002 (0.003) −0.0003 (0.003) GPV 0.010*** (0.003) 0.009*** (0.004) 0.010*** (0.003) SI 0.748 (0.243) 0.886 (0.268) 0.533** (0.231) SIV −0.0004 (0.006) −0.003 (0.006) 0.006 (0.008) SRRV −0.030*** (0.007) −0.029*** (0.006) −0.033*** (0.007) SGV 0.020*** (0.004) 0.021*** (0.006) 0.024*** (0.004) SPV 0.005 (0.006) 0.004 (0.006) 0.020*** (0.007) R2 0.901 0.909 0.911 LR-test on fixed eff. 138.924*** (15) 118.727*** (15) 98.979*** (15) LR-test on interactions 156.083*** (25) 143.272*** (25) 160.111*** (25) LR-test on macrovar. 180.618*** (33) 163.402*** (33) 186.073*** (33) Breusch–Pagan 311.129*** (50) 253.017*** (48) 199.967*** (49) Durbin–Watson 2.066 2.152 2.169 Notes: ***=sign. at 1 per cent; **=at 5 per cent; *=at 10 per cent level. For variables, numbers in parentheses are S.E. For tests, numbers in parentheses are d.f.

These estimates indicate that without further disturbances it takes some 4 to 6 years for a country with no volatility to reach the equilibrium stock. We also calculated the estimated adjustment speed taking account of the volatility terms (at the means of the volatility variables). These estimates, at 0.156 for the whole sample, 0.144 when the USA and the UK are excluded and 0.095 when Japan is excluded,17 imply adjustment times of 10 or so years. This clearly indicates that volatility, as captured by the effects of all the volatility variables taken together, has led to slower adjustment towards the equilibrium robot stock. Because of the large number of interaction terms in the estimating equation individual parameter estimates are not particularly informative. Of more interest are the derivatives of In (St/St-1) with respect to each independent variable taking account of all the interaction terms. In Table 6.4 the estimates of these derivatives are reported with the derivatives being calculated at the means of the independent variables. When the independent variable is measured in log form the derivatives are also elasticities. In other cases the elasticities have been calculated at the sample means and are also reported in Table 6.4. We note these estimates are also very robust to changes in the sample with only one coefficient in the restricted samples (that on RR in the sample excluding the UK and the US) being different to the signs of the coefficients for the whole sample. These estimates indicate the following. 1. The elasticity of St/St-1 with respect to the price of robots (P) is negative, in line with expectations, and significant for all three samples (although for the full sample only at the 10 per cent level). The elasticity is however quite small in absolute value at 0.1 to 0.16. Table 6.4 Estimated derivatives and elasticities of ln(St/St-1) with respect to explanatory variables Variable

(1) Full sample

(2) UK and US excluded

(4) Japan excluded

LP RR

−0.140** (0.073) 0.001 (0.006) 0.007

−0.156** (0.082) −0.00002 (0.006) −0.0001

−0.097 (0.080) 0.004 (0.006) 0.019

TOIVANEN, STONEMAN AND DIEDEREN

Variable

(1) Full sample

(2) UK and US excluded

89

(4) Japan excluded

LIND 0.068 (0.048) 0.055 (0.046) 0.062 (0.044) DG 0.008** (0.004) 0.024 0.008** (0.004) 0.023 0.007** (0.004) 0.022 I 1.181*** (0.465) 0.046 1.276*** (0.500) 0.050 1.311*** (0.506) 0.052 LIV 0.003 (0.014) 0.002 (0.014) 0.001 (0.013) LRRV 0.012 (0.010) 0.021** (0.009) 0.003 (0.010) LGV −0.015* (0.009) −0.013 (0.009) −0.005 (0.010) LPV −0.027*** (0.010) −0.024** (0.010) −0.017 (0.011) Notes: Standard errors in parentheses, elasticities in italics. ***=sign. at 1 per cent; **=at 5 per cent; *=at 10 per cent level.

2. The derivative of ln (St/St-1) with respect to GDP growth (DG) is positive, in line with expectations, and significant for all three samples, indicating that in high growth economies, the diffusion of new technologies is faster. The elasticity is again small at 0.02. 3. The derivative of In(St/St-1) with respect to manufacturing output (IND), although positive as expected a priori, is insignificant for all three samples. The elasticity of St/St-1 with respect to the real interest rates (RR) is positive twice but negative when the USA and the UK are excluded from the sample, although in all cases insignificant and thus very imprecisely measured. Some real options models18 predict a positive relationship between the interest rate and the timing of investment; however, given the imprecise nature of the estimates, our conclusion is that the effect of real interest rates on the diffusion speed of robots is negligible. We should also note that, as stated above, we have been unable to separate out the impact of real wages on the equilibrium robot stock. 4. The derivative of ln (St/St-1) with respect to the rate of inflation (I), which we consider to be a volatility variable, is positive and very precisely measured for all three samples. The elasticity is calculated at 0.05. However we find that the elasticity of St/St-1 with respect to the volatility of inflation (IV), although positive for all three samples is never significant. Thus quite contrary to the policy doctrine which argues in favour of low and stable rates of inflation, we find that inflation speeds up rather than hampers the diffusion of new technologies and the impact of inflation volatility is negligible. 5. The elasticity of St/St-1 with respect to the volatility of robot prices (PV) is negative and if Japan is included in the sample, significant. It should be noted that given the construction of the robot price variable, the volatility of robot prices captures the volatility of exchange rates as well. 6. The other volatility variables perform less well. The elasticity of St/St-1 with respect to the volatility of the real interest rate (RRV) is positive, but only significant when the US and the UK are excluded from the sample. The elasticity with respect to the volatility of GDP growth (GV) is always negative but never significant at the 5 per cent level. These results lead us to conclude that the equilibrium robot stock, as expected a priori, is related to the price of robots (negatively) and GDP growth rates (positively). On the crucial issue of the impact of uncertainty (volatility) on the adoption of new technologies, we do find that overall, the presence of volatility has slowed the diffusion process (the calculated adjustment speed at the sample means is lower than it would be if volatility were zero). However, of the several volatility variables included, only the rate of inflation and the volatility of robot prices have a significant impact on the adoption of robot technology (after taking account of all interaction terms). The impact of the inflation rate is positive. The impact of robot price volatility is negative. This volatility as explained above also captures exchange rate volatility. Comparing the absolute values of the elasticities (ignoring those that are statistically insignificant) we find that all of them are moderate in size, but that the robot price elasticity is largest, followed by the inflation, price volatility and GDP growth elasticities. Looked at from an alternative viewpoint we observe significant impacts upon the adoption of robot technologies from country specific effects (which may include ARCH model uncertainty), the price of robots and its volatility, inflation but not its volatility, and GDP growth but not its volatility. We find little evidence that manufacturing output, the real interest rate or its volatility have impacted on the diffusion process. Conclusions This chapter has multiple objectives: on one hand it is an internationally comparative investigation of the relationship between the speed of diffusion of a particular new technology (robots) and macroeconomic volatility, while on the other it is an empirical contribution to the debate on the relationship between investment and uncertainty. To serve the former purpose, we adopted an approach widely used in empirical studies and estimated an epidemic model of technology diffusion. However the incorporation of terms reflecting uncertainty or volatility into the adjustment coefficient in the model provided the link to the investment and uncertainty literature. In fact the econometric specification of the model is closely related to the stock

90

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

adjustment models widely used in the aggregate investment literature. The resultant estimating equation also allows one to test whether uncertainty (volatility) affects the adjustment speed or the equilibrium stock in the model. The model was estimated on a panel data set covering 16 countries over the period from 1981 to 1993. Two sets of measures of uncertainty were modelled. The first based on ARCH estimations used the variance of forecasts, the latter used direct measures of variable volatility. The ARCH based measures showed insufficient variance over time to be separated from the country fixed effects in the model. Using the more direct measures the results show that volatility has a significant impact upon the adoption of new technology via the adjustment coefficient. The estimates indicate that variables conventionally used in diffusion estimation (e.g. technology price, GDP growth, lagged stock) impacted on the equilibrium stock in line with expectations. On the crucial issue of the impact of uncertainty (volatility) on the adoption of new technologies, we do find that overall, the presence of volatility has slowed the diffusion process (the calculated adjustment speed at the sample means is lower than it would be if volatility were zero). However of the several volatility variables included only the rate of inflation and the volatility of robot prices have a significant impact on the adoption of robot technology. The impact of the inflation rate is positive rather than being negative as the prevailing policy doctrine would suggest. The impact of robot price volatility is negative, however, this volatility also captures the volatility of exchange rates. The policy implications of these findings are not obvious. Clearly there is some virtuous circle in that a growing economy will invest more quickly in new technologies and this in turn should further stimulate growth. However no economy has yet solved the policy problem of how to generate faster long-term growth. The policy statement in the introduction considers that low and stable rates of inflation will generate such faster growth. Our results however throw considerable doubt on this proposition, suggesting instead, that controlling inflation may not directly yield the increases in investment with which the policy makers are hoping to drive the growth process (and may actually harm investment). In fact the only route on the basis of our findings by which stable and low rates of inflation may lead to increased investment would be through exchange rates. If low and stable rates of inflation lead to stable exchange rates this would directly stimulate investment. Alternatively, if low and stable rates of inflation lead to upward revaluations of the exchange rate this would make imported capital goods embodying new technology cheaper and again stimulate investment. However, if low and stable rates of inflation are sought through deflationary policies, one might argue, in a Keynesian manner, that such policies will impact upon the growth of GDP, and this in turn may lead to lower investment in new technology thereby nullifying the exchange rate effects. Although virtuous circles may therefore exist, it is by no means clear how policy should be directed in order to break into them. Appendix Construction of variables The dependent variable in the empirical analysis was defined and constructed as the natural log of the ratio of the current robot stock to the lagged robot stock. For Australia, the 1985 stock and for Switzerland, the 1989 stock were missing and the averages of the stocks of the previous and following year were used. The explanatory variables were calculated as follows: LS=the natural log of the robot stock lagged one period. LP=natural log of the real robot price index (1978=100). The price index is based on figures reported in the World Robot Statistics for years 1980–1993 on the unit value of robot production in Germany. Prices for 1979 and 1978 were calculated using the results of a regression of prices on a constant, years and squared years.19 We used the consumer price index as reported in the IFS statistics as a deflator (since it is available for all countries). To convert D-Mark prices into home currencies, we used the period average home currency/USD and D-Mark/USD series of the IFS statistics. LIND=natural log of the industrial output index, based on the manufacturing output figures in the IMF statistics (1978=100). DG=percentage annual change in GDP as calculated from IFS real GDP time series. I=percentage annual inflation rate calculated from the consumer price index in the IFS statistics. The consumer price index was chosen since it was available for all countries. RR=The real interest rate measured as the difference between the money market rate as reported in the IFS statistics and inflation (calculated as above). LRW=natural logarithm of an index of the real wage in manufacturing, based on ILO statistics (1978=100).

TOIVANEN, STONEMAN AND DIEDEREN

91

Country dummies have also been included to capture differences in manufacturing output and the price level in base years and hence enable us to use indexes for these variables. For Taiwan, data have been taken from statistics published by the Central Bank. These are designed to conform with the IFS statistics. Correlation patterns Table 6.A.1 details the patterns of correlation between the various explanatory variables and their volatilities (excluding interaction terms to save space) Table 6.A.1 Correlation matrix of some explanatory variables Variable

LP

RR

LIND

DG

I

LIV

LRRV

LGV

LPV

LP RR LIND DG I LIV LRRV LGV LPV

1.000

0.157 1.000

−0.136 0.048 1.000

−0.269 −0.004 0.250 1.000

0.401 −0.358 −0.252 −0.274 1.000

−0.339 −0.022 0.284 0.488 −0.458 1.000

0.163 −0.188 −0.126 0.0004 0.262 0.113 1.000

0.159 −0.014 −0.141 −0.244 0.084 −0.103 0.214 1.000

0.248 −0.046 −0.073 −0.114 0.318 −0.090 0.184 0.070 1.000

Restriction tests Here we present Wald test (1 d.f.) results on the restrictions given in Table 6.A.2, equations (4) and (5). Table 6.A.2 Wald restriction tests Tests of restriction: (4) i=volatility variable, j=level variable Variables

Whole sample

UK and USA excluded

Japan excluded

inflation, robot price inflation, real interest rate inflation, manufacturing output inflation, GDP growth vol. of robot price, robot price vol. of robot price, real interest rate vol. of robot price, manufacturing output vol. of robot price, GDP growth vol. of GDP growth, robot price vol. of GDP growth, real interest rate vol. of GDP growth, manufacturing output vol. of GDP growth, GDP growth vol. of real int. rate, robot price vol. of real int. rate, real interest rate vol. of real int. rate, manufacturing output vol. of real int. rate, GDP growth vol. of inflation, robot price vol. of inflation, real interest rate vol. of inflation, manufacturing output vol. of inflation, GDP growth

0.484 (1.071) −53.414 (762.97) 0.895 (0.891) −0.509 (1.798) −1.637 (10.555) −0.1977* (0.105) 22.617 (319.10) −0.291*** (0.098) 0.091 (0.152) −1.983 (2.473) 0.588** (0.277) 57.787 (787.02) −1.101 (1.516) 1.123 (2.660) 4.777 (8.967) 4.416 (3.370) 0.765 (10.208) 3.502** (1.436) −12.153 (23.752) 0.580 (0.609)

0.465 (1.034) −15.056 (39.838) 0.989 (0.974) −0.691 (1.839) −4.717 (15.357) −0.174** (0.095) 4.444 (10.823) −0.424*** (0.114) 0.122 (0.155) −1.937 (2.783) 0.599** (0.254) −2.623 (25.551) −0.512 (1.703) 1.531 (2.642) 4.4515 (9.948) 4.634 (3.502) 4.184 (28.096) 2.401*** (0.729) −14.400 (32.295) 0.519 (0.972)

0.798 (2.202) −1.904 (5.814) −1.646 (1.523) −3.880 (2.395) −6.151* (3.660) −0.298* (0.178) −1.917 (2.176) −0.321*** (0.103) −0.011 (0.141) −0.581* (0.337) 0.860* (0.508) 0.458 (10.362) −0.200 (1.308) 2.470 (2.508) 1.836 (1.608) 2.717 (1.764) 5.099 (16.111) 4.125** (1.857) −85.940 (1020.9) 2.021** (0.905)

92

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

(5) i=volatility variable Restriction

Whole sample

UK and USA excluded

Japan excluded

inflation 1.947 (5.1046) 2.529 (2.605) 2.282 (5.309) vol. of inflation −0.073 (0.407) 0.166 (0.464) −0.105 (0.381) vol. of real interest rates 0.061 (0.240) −0.035 (0.280) 0.338 (0.270) vol. of GDP growth 0.039 (0.259) 0.100 (0.302) −0.162 (0.287) vol. of robot price 0.081 (0.230) 0.002 (0.230) −0.413 (0.354) Notes: Numbers in columns 2–4 are Wald test with 1 d.f. and (prob. value); ***=sign. at 1 per cent, **=sign. at 5 per cent, *=sign. at 10 per cent level.

Results from ARCH estimations As the robot time series (1982–1993) is too short to allow the estimation of time series models country by country, we collected as long data series as possible (from IFS statistics) for the macrovariables for which we wanted to obtain volatility measures. These were the exchange rate, the real interest rate, inflation, and GDP growth. The longest time series obtained was 1949–1994, but for some variables and countries the obtained series were shorter. Even so, the shortest time series had over 30 observations. Our estimation strategy was to obtain as parsimonious representations as possible for each of the time series. The models thus obtained would then give the predicted value, and the predicted variance for each variable. As the time series were fairly short we opted for the specific-to-general approach and started with the assumption that a variable is best represented by an AR (1) process. We then tested for autocorrelation up to 4th order, and if any was found, these were corrected. The resulting AR (x) model has thus no autocorrelation up to 4th order. We then tested for the presence of ARCH effects using the standard procedure as outlined, e.g. in Greene (1993). We tested for the (joint and individual) significance of ARCH effects of 1st order, 1st and 2nd, etc. up to 4th order. If none was found, we accepted the null of homoscedastic error terms. If some was found, we estimated the predicted variance for the variable in question. As reported in the main text, we only found at most four time series (i.e. countries) with ARCH effects for any of our macrovariables. Table 6.A.3. gives a summary of the results. These results differ somewhat from those published in the ARCH literature. For example, we could not reject the null of homoscedasticity for US inflation. One potential explanation for such differences is that we have used yearly data that is in line with data on robot diffusion, but other researchers often use quarterly or monthly data. Using, e.g. quarterly data would have led to complications in transforming results from ARCH estimations to robot diffusion estimations where we had only yearly data available. Also, as such data was not available in a format consistent over countries, we decided to use (yearly) IFS data as that guaranteed consistency. Table 6.A.3 Results from ARCH estimations Variable Inflation

Real interest rate

GDP growth

Exchange rate

Country Number of lags in final model

ARCH effects at lags 1,…,4

Number of lags in final model

ARCH effects at lags l,…,4

Number of lags in final model

ARCH effects at lags 1,…,4

Number of lags in final model

ARCH effects at lags 1,…,4

1 2 3 4 5 6 7 8 9 10 11 12 13

— — — — — — 1 — — — — — —

3 1 1 1 1 1 1 1 2 3 1 1 3

— — — — — 1,2,3,4 1 1,2 — — — — —

1 1 1 4 1 2 1 1 1 1 1 1 1

— — — — — — — — — — 1 — —

1 3 1 2 1 1 1 1 1 1 3 3 1

— — — — — — — — 1 1 — 3 —

3 4 1 1 1 1 1 2 1 1 1 1 1

TOIVANEN, STONEMAN AND DIEDEREN

Variable Inflation

Real interest rate

GDP growth

Exchange rate

Country Number of lags in final model

ARCH effects at lags 1,…,4

Number of lags in final model

ARCH effects at lags l,…,4

Number of lags in final model

ARCH effects at lags 1,…,4

Number of lags in final model

ARCH effects at lags 1,…,4

14 15 16

— 1 —

1 1 1

— — 1

1 1 2

— 1 —

1 3 1

— — 1

1 1 1

93

Notes 1 We thank seminar audiences at the MITI Research Institute Tokyo and the ESRC Conference on Risk and Human Behaviour, University of York for comments on an earlier version of the chapter. This version has benefited from comments received in the WBS Technological Innovation seminar, at the RES Conference in Swansea and at the EEA Annual Meeting in Istanbul. This research is part of an ESRC funded project on Risk Management and the Adoption of New Technology. The usual caveat applies. 2 By which we mean the equilibrium stock at time t, not the stock at the end of the diffusion process as t tends to infinity. 3 For a more in-depth discussion on real options than the one that follows, see Dixit and Pindyck (1994) and the references therein. 4 See Lawrence Summers (1987) for empirical evidence on investment thresholds. 5 This is the case when the increased volatility is in the variable that determines the realisation of the component of the economic system the value of which cannot be known in advance. An example is the variance parameter of a Brownian motion that could characterise, e.g. output prices. 6 In the estimations, we use a period of 3 years: hence VOLi,t=STDi,t,…, t−2/ |AVGi,t,…, t−2|. We experimented with longer periods, but the measures are very similar. The reason for using a short period is to avoid interpreting long run trends as volatility. We experimented with using and the VOLi,t=STDi,t,…,t−2 results are in line with those reported but not as good. 7 We also estimated a logistic model, the most common specification. The essential difference is that in the logistic, the right hand side variables are levels, not logs. Its performance was, however, clearly inferior to the Gompertz and hence only the latter results are discussed. 8 See Stoneman and Karshenas (1995) for a justification, although Pindyck and Solimano (1993) discuss reasons why there could be a positive relationship between investment and the interest rate. We also experimented with including the wage in manufacturing as an explanatory variable. However, when the interaction terms with other variables were included as per the structure of the model it resulted in severe multicollinearity problems. When included without the interactions, the wage never obtained a significant coefficient. We therefore excluded it from the final estimations. 9 In the Pindyck and Solimano (1993) data, inflation, the volatility thereof and the volatility of real interest rates were correlated with their measure of volatility and with each other. In our data, with the exception of inflation and its volatility, they are not significantly correlated (see the Appendix, Table 6.A.1.). 10 We have experimented empirically with the introduction of these three variables in log forms in the empirical analysis by imposing sample restrictions and other methods. The results are not significantly different from those reported below. 11 Twelve countries were excluded for a variety of reasons, for example, for Russia and Hungary there was no reliable data on macrovariables (and even the robot data is suspicious) and the Benelux countries (The Netherlands, Belgium and Luxembourg) were summed together in the robot statistics. 12 Our discussion on robots and robot statistics relies heavily on World Industrial Robots 1994, Geneva, United Nations. 13 The real prices measured in each country’s home currency were then transformed into a price index to create a common measure across countries. The consumer price index was used since it was the only price index available for all countries over the observation period. The country fixed effects capture the differences in base year (1978) prices. We also ran estimations using the unadjusted German and Italian price data. With the former, the results are almost identical with the price index-based results presented here. With the latter, the results were broadly in line with those presented. 14 For more detail on how we constructed our explanatory variables, see the Appendix. For Taiwan, Financial Statistics (various issues) published by the Central Bank were used. These conform with the IMF statistics. ILO statistics were used for manufacturing wages (apart from Taiwan, for which Monthly Bulletin of Statistics of the Republic of China were used). 15 The correlations between variables are presented in the Appendix. Most are very small in absolute value: there are only 6 with absolute values over 0.3, and 3 with absolute values in excess of 0.4. These were corr (LI,LIV)=−0.458 corr (LP,I)=0.401 and corr (LIV,DG) =0.488. 16 The correlations between variables are presented in the Appendix. Most are very small in absolute value: there are only 6 with absolute values over 0.3, and 3 with absolute values in excess of 0.4. These were corr (LI,LIV)=–0.458 corr (LP,I)=0.401 and corr (LIV,DG) =0.488. 17 All the adjustment speed estimates are significant at the 1 per cent level. 18 See Pindyck and Solimano (1993) for a discussion on this. 19 The results of the regression are available from the authors upon request.

94

UNCERTAINTY AND INVESTMENT IN NEW TECHNOLOGY

References Bollerslev, T. (1986). Generalized Autoregressive Conditional Heteroskedasticity, Journal of Econometrics, 31, 307–27 Caballero, R.J. (1991). On the Sign of the Investment-Uncertainty Relationship, American Economic Review, 81, 279–88 Chow, G.C. (1967). Technological Change and the Demand for Computers, American Economic Review, 57, 1117–30 Davies, S. (1979). The Diffusion of Process Innovations, Cambridge: Cambridge University Press Davis, G. and Kanago, B. (1996). On measuring the effect of inflation uncertainty of real GNP growth, Oxford Economic Papers, 48, 163–75 Dixit, A.K. and Pindyck, R.S. (1994). Investment under Uncertainty, Princeton, NJ: Princeton University Press Dixon, R., Hybrid Corn Revisited (1980). Econometrica, 48, 1451–61 Driver, C. and Moreton D. (1991). The Influence of Uncertainty on UK Manufacturing Investment, Economic Joumal, 101, 1452–59 Eberly, J. (1993). Comment, NBER Macroeconomics Annual, 303–12 Engle, R.F. (1982). Autoregressive conditional heteroskedasticity with estimates of the variance of United Kingdom inflation, Econometrica, 50, 987–1007 Greene, W.J. (1993). Econometric Analysis, 2nd edn, New York: Macmillan Griliches, Z. (1957). Hybrid Corn: An Exploration in the Economics of Technological Change, Econometrica, 25, 501–22 Hall, R.E. (1993). Comment, NBER Macroeconomics Annual, 313–16 Ingersoll, J.E. and Ross, S.A. (1992). Waiting to Invest: Investment and Uncertainty, Journal of Bminess, 65, 1–29 Jensen, R. (1982). Adoption and Diffusion of an Innovation of Uncertain Profitability, Journal of Economic Theory, 27, 182–93 Mansfield, E. (1968). Industrial Research and Technological Innovation, Norton, New York Nickell, S.J. (1978). The Investment Decision of Firms, Cambridge: Cambridge University Press Pindyck, R.S. and Solimano A. (1993). Economic Instability and Aggregate Investment, NBER Macroeconomics Annual, 259–303 Reinganum, J.F. (1989). The Timing of Innovation: Research, Development, and Diffusion, in Schmalensee, Richard and Willig, Robert (eds) The Handbook of Industrial Organization, Amsterdam: North Holland Rothschild, M. and Stiglitz, J.E. (1970). Increasing Risk: I. A Definition, Journal of Economic Theory, 2, 225–43 Stoneman, P. (ed.) (1995). Handbook of the Economics of Innovation and Technological Change, Oxford: Blackwell Stoneman, P. and Karshenas, M. (1995). Technological Diffusion, in Stoneman, P. (ed.) Handbook of the Economics of Innovation and Technological Change, pp. 265–97, Oxford: Blackwell Summers, L.H. (1987). Investment Incentives and the Discounting of Depreciation Allowances, in Feldstein, Martin (ed.) The Effects of Taxation on Capital Accumulation, Chicago, IL: Chicago University Press

Statistical sources Financial Statistics, Taiwan District Republic of China, the Central Bank of China, various issues International Financial Statistics γearbook, IMF, various issues Monthly Bulletin of Statistics of the Republic of China, various issues Statistical γearbook, International Labour Organisation, various issues World Industrial Robots 1994, United Nations

Part II The consequences of investment

7 Overview Investment, feedback and spillover Ciaran Driver and Paul Temple

Introduction In the post-war period, the UK has devoted a smaller share of resources to investment than other European economies. However, as Figure 7.1 illustrates, the share of GDP going to physical investment has been declining in the EU since the 1970s. Do these things matter? Should policy makers be worried? This section considers the implications of such a development. Answers are by no means obvious. Although many evidently believe that investment can be insufficient (and the resulting capital stock too small) it is clear that investment is not always a good thing: we all have experience of consumer durables which we rarely have put to use. And in an industrial context, the gains from many IT investments have been slow to materialise. Finally, there has been talk of ‘over-capacity’ in relation to the crisis in Asia. This, however, highlights that in an uncertain world excessive capacity may not always be the result of a mistake. Has the recent Asian crisis to be put down to the poor quality of decision-making or to the bad luck associated with debt denominated in increasingly expensive dollars? Whatever the determinants of investment, or the quality of investment decision making, the proximate outcome of the investment process is manifested most-obviously in terms of the capital stock—its maintenance, augmentation, and its distribution across alternative uses. Some of the consequences of investment can be discussed in terms of the standard neo-classical model. In the simplest version, investment responds passively to the flow of savings through the operation of the price mechanism. In a steady state, without technical progress and with labour force growth driven exogenously, available savings draw forth just sufficient investment to equip additional workers with the existing level of capital per worker —a process of capital widening. Any increase in the proportion of output saved (and hence invested) results in a process of capital deepening, which carries on until, at a higher level of the capital-labour ratio, savings are once again just sufficient for purposes of capital widening. Capital deepening is described by the movement along a production function displaying diminishing returns. The ultimate result is an economy with a higher level of productivity but no long run impact on the growth rate of productivity. The only feedback mechanism operating is the negative one of diminishing returns. Matters become slightly more plausible when it is realised that the transition to a new steady state from any increase in the savings rate may take decades (Bombach 1985) and further, that diminishing returns only apply at the technological frontier. For countries not at the forefront of technology, the exploitation of ‘catch-up’ with the frontier can delay the operation of diminishing returns (Clark 1993). The idea of catch-up itself is consistent with the conception of technology as a public consumption good or perhaps, in recognition of the contingent nature of catch up as a club good—see for example Abramovitz’s (1986) original reference to ‘social capability’. Clearly however, the question of technical progress at the frontier also needs to be addressed. Here the neo-classical stories are unconvincing. Technical progress is seen as being driven exogenously (that is, independently of the decisions being made by firms and individuals). In a steady state moreover it is purely labour augmenting (Harrod neutral). The story is essentially unchanged; in a steady state there will be a constant ratio of capital per efficiency adjusted unit of labour. With labour measured conventionally, the capital–labour ratio rises with the rate of technical progress. Given the one-commodity world in which many of these neo-classical tales are told, the technological frontier can only be associated with one economy at any point in time. For the twentieth century, this economy can only be the US—the productivity leader across the board. Here we might expect to find positive evidence of the dominance of diminishing returns. Yet even here we cannot find much overt evidence of diminishing returns to investment shaping the growth process. What then are the countervailing forces? In the neo-classical model of a growing economy at the frontier, these are represented, as we have seen, by exogenous forces of technological change —a convenient, but unconvincing, scenario. More plausible must be a positive feedback mechanism capable of delaying or neutralising the impact of capital deepening. That such forces exist, and are powerful, was recognised at least as far back as Adam Smith, who based his key economic work on the powerful forces unleashed by capital accumulation. Two aspects of the accumulation process, both of which could serve as important positive feedback mechanisms were dealt with by Smith. Because of their relevance today we consider both in this overview.

INVESTMENT, FEEDBACK AND SPILLOVER

97

Figure 7.1 Gross domestic fixed capital formation as a percentage of GDP. Source: OECD National Accounts.

First, and most famously, was the impact of investment on specialisation and the division of labour. Second, was the fact that capital—or ‘stock’ as Smith called it—was the entity that created ‘productive’ employment. Smith, and the classical economists more generally, never doubted that capital and labour were complementary factors. Both these propositions are central to two of the chapters presented in this section. Let us consider both propositions in turn. Both these feedback mechanisms may, beyond delaying the onset of diminishing returns, raise the social rate of return above the private rate of return. Potentially they therefore have considerable policy implications. We now consider both in turn. Spillovers The existence of the kind of positive feedback mechanisms described by Smith is of course the fundamental proposition of the burgeoning endogenous growth literature—the idea that the long run (i.e. steady state) growth rate of economies can be influenced by the share of resources devoted to investment activities. There are two basic strands to these studies. The first considers the role of human capital accumulation and its relationship to fixed capital accumulation; the second analyses the purposive generation of new knowledge by innovation processes in general and R&D processes in particular. The impact of the first strand of thinking for economic thought has been to stimulate debate about the role of human capital formation for economic growth. The second has begun to consider the micro-economics of purposive technological change, a subject which even the most die-hard ‘old growth’ theorist would admit to being a subject of both policy relevance yet totally underdeveloped (see for example Solow’s recently published series of Arrow lectures [Solow 1997]). Both strands have however stressed the empirical importance of spillovers. Consider first the role of human capital in economic growth. A more general formulation of the basic two factor production function would have human capital as one of its arguments: where H represents the stock of human capital. Such a function would provide the basis for endogenous growth if it displays constant returns to the reproducible factors of production K and H., e.g. in per capita terms: such that α+β=1. Of course, human capital can be introduced in ways which are quite consistent with old growth theory, as the contribution by Mankiw et al. (1992) makes clear. However, in the particular case α+β=1, we have a version of the socalled AK model: along a steady state growth path, with both forms of capital (‘broad capital’) growing at an equal rate we may write: (1) 1−α where A=A(h/k) is constant. This can also be written in the form, allowing for a rate of depreciation for capital of δ:

98

CIARAN DRIVER AND PAUL TEMPLE

(2) In other words, there is a steady-state relationship between the growth of output per head and the share of output devoted to investment, i. The existence in one form or another of such a relationship can be thought of as a hallmark of the endogenous growth literature, since an analogous relationship to (2) is derived in the endogenous innovation literature as well. Here, however, the steady state rate of growth is increased by expanding the share of resources devoted to innovation. This is limited because of problems of appropriability, with at least some of the increase in knowledge emanating from an innovation becoming more generally available. This problem is discussed further below. There, however, the similarity between the two strands of endogenous growth literature seems to end, especially in relation to the policy implications. For example, the richer micro-economic foundations of the endogenous innovation literature make the impact of increasing global economic integration ambivalent. While a wider market increases the potential gain from successful innovation, more competition tends to reduce the probability of success. In the AK model, by contrast, the increases in static efficiency resulting from freer international trade would raise the ratio of output to broad capital (for a fuller discussion see Crafts 1996). In both these models however, the role played by physical investment does not differ fundamentally from that in the standard neo-classical model. A high rate of innovation (whether exogenous or endogenous) leads to a higher rate of output growth and to a higher investment share (Helpman 1993). However, the rate of innovation is influenced in these theories by rates of time preference, since the flow of R&D inputs occurs earlier than the flow of innovations. A number of tests of a relationship such as that in (2), (which suggests that permanent shocks to the investment ratio should be accompanied by permanent shifts to the steady state growth rate) can be found in the literature. Certain problems suggest themselves immediately. The first is that observations of growth rates and investment shares are not likely to have been steady-state observations. Even in the neoclassical model, outside the steady state, a boost to the investment share will provide a temporary lift to the growth of output before the economy settles down at the same growth rate but at a higher level of output. This, by itself, could explain the robust relationship between the investment share and growth rates found in the extensive analysis of international cross-sectional data of Levine and Renelt (1992). Second, the relationship between i and g could easily be turned around to reveal an accelerator type of relationship, so that the causal story may be highly complex, even given a strong statistical relationship. Finally, there are fundamental measurement problems, both with the left-hand and right-hand side of (2) that make empirical resolution of the issues problematic. For example, what (especially in the presence of international spillovers) is the growth rate that is germane to the investment rate being considered? Some of the problems involved in considering a relationship as in (2) can be illustrated by reference to some recent challenges to endogenous growth theory. Jones (1995) for example has focused on the increasing share of the OECD’s labour force that is being devoted to R&D. Is the fact that we have not observed any obviously commensurate rise in conventional total factor productivity growth a falsification of the endogenous innovation story? Not necessarily; measured R&D, whether in terms of personnel or expenditures is simply an imprecise proxy for innovative inputs. Moreover, since it has been widely used as a policy instrument to encourage innovation, there may be clear and increasing incentives for firms to describe activities as being R&D. In a more theoretical vein, it may be that a constant rate of innovation requires an increasing rate of investment in R&D, due for example to the increasing complexity of technology. Aghion and Howitt (1998) for example argue that, in a world where technology is described by the variety of products, R&D has a smaller and smaller proportional spillover effect. Our own interest in this burgeoning literature stems from the hypothesis that physical capital may be an important part of a broad capital stock subject to constant returns. This also has received substantial criticism. Jones (1995a) has for example argued that growth rates are stationary while gross investment rates are non-stationary. However, as Demetriades et al. (1998) argue in a recent paper, a first line of defence for endogenous growth theory might be that of a regime shift involving the parameter A, or indeed the rate of depreciation δ. Investigating the former idea using a panel for 20 countries for the period 1950 to 1990, Demetriades et al. find support for constant returns to reproducible capital. But they find none for the longrun constancy of A. As far as the latter possibility is concerned, there is some evidence that today’s capital stocks do lean more heavily toward machinery and equipment, and hence a higher rate of depreciation. Figures 7.2 (all industries) and 7.3 (manufacturing) show that there is a fairly clear-cut trend toward equipment in the four economies for which data were available, the one exception being US manufacturing. Another important question concerns the precise form that human capital takes within the production function. While alternative approaches and measures may be highly correlated from a statistical perspective, there are important policy implications resulting from the distinction: inter alia between the education of the workforce, the skills it acquires from purposive training, and the knowledge and skills acquired through a process of learning by doing. All these things are of course difficult to measure and to specify in an appropriate empirical form. Naturally enough many economists have tended to focus on measures of formal educational attainment, which are most capable of international comparison. However, results have been largely inconclusive, as a recent study by Oulton and Young (1996) makes clear. Unlike a number of other investigations,

INVESTMENT, FEEDBACK AND SPILLOVER

99

Figure 7.2 Machinery and equipment as a proportion of gross capital stock (all industries). Source: OECD sectoral database.

Figure 7.3 Machinery and equipment as a proportion of gross capital stock (manufacturing). Source: OECD Sectoral Database.

they find that the incorporation of measures of educational attainment is generally statistically insignificant, which they suggest may be the result of measurement error. At the other extreme of hypotheses which place formal education as a causal link in the process of economic growth (and not just as an accompaniment to growth) are those which emphasise the more tacit process of learning by doing. Here the controversial work of DeLong and Summers (1991, 1992) takes pride of place. Having examined mainly cross-sectional data they argued that there is a robust, causal link running from equipment investment (properly measured to distinguish between actual investment and the current price investment share) to variations in economic growth across economies. Their hypothesis is based on the observation that technological change is typically capital using and that there is a close relationship between technological knowledge and equipment investment in particular. Such knowledge is often ‘tacit’—based on hands on experience, and hard to transfer into formal education. There is some affinity in this idea with the ‘embodiment’ approach

100

CIARAN DRIVER AND PAUL TEMPLE

to the capital stock, in which productivity is influenced by the impact of investment on the average age of the capital stock— the greater the rate of investment, the greater the share of output in more technologically advanced equipment. However, in practice, it turns out that the average age in the capital stock is relatively insensitive to changes in the rate of investment (e.g. Dennison 1964). Moreover, the rate of technological advance is independent of the rate of investment. Here, this is not the case; moreover, the resulting productivity improvement cannot be appropriated by firms very easily and, on the DeLong and Summers view, is a source of a significant wedge between private and social returns to investment. While recognising that there must always be a question mark about the direction of causation between growth and investment, DeLong and Summers argue that, if high equipment investment were simply a consequence of high rates of growth, then we would expect rather higher equipment prices in rapidly growing economies. In fact they find no such relationship, suggesting that supply factors are the dominant force. Whatever the merits of this argument, the question of causality can also be addressed by more formal statistical means, and this is the basis of the contribution of Jerry Coakley and Andrew Wood to this section (Chapter 8). They note that the statistical results of DeLong and Summers are based on regressions of labour productivity on estimates of the share of output devoted to equipment investment and various possible exogenous influences on labour productivity. In theory, such an approach could be derived from a Cobb—Douglas technology in which labour augmenting technical progress is driven by such influences. Coakley and Wood instead examine the cointegration properties of the relationship between investment and equipment and investment in structures within a VAR framework. This allows them to test explicitly for the validity of the exogeneity assumption in the DeLong-Summers regressions and to examine more fully the role of structures investment. The exogeneity assumption is very important in empirical work, because without it, it is impossible to interpret the estimated output elasticity on investment as representing a social rate of return. Although they do find evidence of causality running from both forms of investment, they show that causality runs in both directions and they can find no evidence for the ‘privileged’ position of equipment investment. Note, however, that, as with all attempts to infer causality from the behaviour of time series, this analysis carries a health warning, for it depends upon the ability to associate causality with the sequencing of events in time. This may be especially problematic in looking at investment, which is almost by its nature forward looking. More importantly however, they argue that only detailed analyses of the social rate of return can provide answers to the riddles posed by the impact of investment. In effect, this implies careful analysis, at a micro level, of spillover processes, and arguably a sound knowledge of these is the only sensible foundation for policy making.1 The idea of technological spillovers has been particularly important in the second strand of endogenous growth theory— that which emphasises innovation processes as the key ingredient in growth rather than ‘broad’ capital formation. In many of the models based on innovation, what matters is the share of resources devoted to innovation, which is now viewed as an endogenous phenomenon, determined by the profit maximising activities of firms. This requires the injection of monopolistic elements so that firms can cover the fixed costs associated with the ‘research’ required to produce innovations. The seminal contribution is by Grossman and Helpman 1991; for a general discussion of both strands of endogenous growth analysis, see Crafts 1996. From a policy perspective, much of the theory emanating from this approach is highly contentious. The importance of fixed costs in research means that the rate of return to innovation depends upon a size effect, while the need to appropriate returns suggests a positive role for a degree of market power. For those with even a fleeting knowledge of European industrial and technology policy, with its emphasis on preventing R&D duplication, and its creation of ‘national champions’ this may appear strange indeed (see for example Ergas 1987). However, many of the policy issues raised by theories of endogenous innovation depend for their interpretation upon accurate information about the size and nature of technological spillovers. Here theoretical perspectives vary considerably. One variant on the endogenous innovation theme deals with the spillover of firm specific R&D efforts into a general pool of knowledge via which all firms can participate (e.g. expanding product variety models). By contrast in the so-called ‘quality ladder’ approach, the appropriability of R&D is strictly limited to firms within an industry (e.g. Grossman and Helpman 1991a). Empirical investigation first of all requires an ability to measure technological activity. Two methods have been important in practice: patent counts and R&D expenditures, although other measures (such as innovation counts and counts of technical standards) have also been used. Attempts to construct knowledge stocks (and hence a proxy for the services of knowledge capital) on the basis of R&D expenditures is largely based on the pioneering work of Griliches (1973) Armed with such a measure, production functions can be estimated at different levels of aggregation. For example, at the firm level, arguments in the function may include both in-house and external R&D, with a significant impact of the latter suggesting an externality. But where do we draw the boundaries of these external effects? Data tends to be available at firm, industry and economy wide levels. But are official industrial classifications appropriate? Do spillovers cross national boundaries within an industry more easily than they move across industries? In relation to questions such as these, some insights have certainly been gained, particularly perhaps when patent data have been used to identify the ‘technological proximity’ of ideas. Current research is ambivalent on many aspects of spillover and much new work remains to be done. However, research does point to the potential importance of spillovers and, inter alia, to the geographical and organisational ‘stickiness’ of knowledge flows. This

INVESTMENT, FEEDBACK AND SPILLOVER

101

fact may help to explain the strategic motives behind the extent of FDI flows, in the sense that knowledge flows more easily within a firm than it can across firms. Technological spillovers and their empirical identification thus provide a useful method of examining endogenous growth theory. In Chapter 9, Ray Barrell and Nigel Pain estimate a model in which endogenous technological change in the economy is related to both the level of FDI and the stock of patents. The argument that FDI might affect the growth rate of productivity is implied by their theory of FDI which gives a prominent role to the exploitation by firms of their specific tacit knowledge, which can none the less be applied in different locations. We find this argument quite compelling. One of us has been involved in a study of multinationals in Ireland (with a domestic or international base) where firms were asked to identify the reason as for investment abroad. The replies are described in Table 7.1 (see Driver and Whelan 1998). While firm-specific knowledge was not top of the agenda in the Irish context—cost factors being most important—there was a significant vote for that motivation in carying out FDI. A possible implication of this motivational theory—which Barrell and Pain go on to test in their chapter—is that FDI becomes a source of technological spillover as domestic firms seek to emulate the firm-specific characteristics of their foreign rival. The exact source of FDI productivity is controversial (for the car industry, see Griffiths, 1999). However, Barrell and Pain note that UK inward FDI appears to have a stimulating effect on domestic productivity in the manufacturing sector but not in the non-manufacturing sector. This could indicate either increased scope for transferable skills in the manufacturing sector or, as the authors suggest, it may reflect the initial relative disadvantage of UK manufacturing operations. It is, perhaps, unclear whether the manufacturing effect is truly a technological one or a more simple competitive effect. To put it more directly we cannot be sure that FDI stimulates technological progress or whether it stimulates productivity by way of X-efficiency Table 7.1 Reasons for foreign investment Reason: % within each column responding:

Response (%)

Very important (important) Proximity to markets Firm has unique competitive advantage Production and labour costs abroad Geographical diversification to minimise risk Home market is saturated Other (trade barriers, grants, tax, etc.) Source: Driver and Whelan (1998).

40 (59) 25 (71) 60 (35) 19 (23) 11 (12) 42 (35)

or an intensification of labour. The short lag of four quarters for the effect to operate suggests that it is unlikely to be entirely due to technological adjustment. Evidence in Hay and Liu (1997) suggests that firms’ relative inefficiency (derived from efficiency frontiers) is related in the long run to capital investment. On the other hand the same research found a contemporaneous increase in efficiency in response to rival firms increase in efficiency. It seems unlikely that all of this can be attributed to capital investment, though some may be and it might be a fruitful research topic to explore further. If increased domestic investment and efficiency is indeed the general response rather than defensive cost-cutting and the type of submissive response discussed in Scherer (1991), then FDI could be seen in a more positive light. The attempt to find a role for domestic patents in the equation for technological progress was less successful in that it seems to find significance only when constraints are imposed. This is perhaps not all that surprising. Patents are imperfect indicators of technological progress and this is particularly true in the engineering sectors where technological progress can be rapid. The finding that patents enter the demand for capital equation alone is interpreted by the authors as suggesting a role for process innovation. However patents are sometimes argued to be even less important in protecting process innovations than product innovations as it is easier for firms to rely on first mover advantages, learning, tacit knowledge and secrecy (Levin et al. 1987). Since patents and R&D are closely correlated at economy level, the effect may be capturing lagged R&D and the incentive to innovate. In that case the true lag is more diffuse and almost certainly longer than the four-quarter observed lag on patents and this would cohere with what we know about the sluggishness of the diffusion process. Investment and employment Adam Smith’s second theme concerned the relationship between the capital stock and employment: along with most classical economists he believed that a larger capital stock would generate higher employment. Many modern economists appear to deny this, at least for the long run, noticing that while capital intensity has risen secularly, unemployment has not been so trended, although it has fluctuated quite dramatically between one period and the next. This belief in the long-run

102

CIARAN DRIVER AND PAUL TEMPLE

independence of the rate of unemployment from capital accumulation is particularly associated with the work of Layard et al. (1991); one important consequence of this is a tendency to emphasise the role and the institutions of the labour market in accounting for unemployment. It is however worth noting that the two views are not entirely incompatible since the rate of unemployment depends upon the decision to enter the labour force in the first place, on migration decisions, and on the ways in which work is organised. A century and more ago it seems certain for example that there was much more underemployment than there was today, not just in agriculture, but also because more work was organised on a casual basis. According to Matthews et al. (1982), for example, a general trend decline in the informal sector of Britain’s labour market in the nineteenth century did not manifest itself in any fall in unemployment simply because most of the workers in this sector were outside the scope of the unemployment figures. Given this kind of measurement difficulty, it might be more meaningful to consider the relationship between investment and employment. A recent study of the OECD for the period 1960 to 1992 by Rowthorn (1995) showed a significant impact of capital stock growth on employment, especially in the manufacturing sector of the economy. In contrast to the position of those who believe in the long run independence of unemployment from investment, a number of writers have begun to stress the possibility that unemployment may be a symptom of an inadequate capital stock. A key early contribution was that of Malinvaud (1980); more recent contributions have come from Bean (1989), Bean and Gavosto (1990), and Rowthorn (1995, 1996). On the face of it, the independence assertion may seem somewhat strange. Evidently an addition to the economy’s stock of physical capital requires additional workers to operate it. However, there is a further impact operating through any increase in wages which create an incentive for ex post substitution between factors on the existing capital stock. Thus the empirical value of elasticity of substitution between capital and labour partly determines the longer-run relationship between investment and employment, together with the impact of increased investment on wages. Both aspects of the impact of investment can be investigated within the so-called NAIRU (non-accelerating inflation rate of unemployment) framework. Developed specifically to tackle the macro-economic problem of the relationship between unemployment and inflation, this approach has become the standard vehicle for the attempt by labour economists to understand the persistently high levels of unemployment experienced across much of Europe over the 1980s and 1990s. For a useful discussion of this approach to the problem see Bean 1994. Typically, the NAIRU framework allows for bargaining power on the part of the labour force, and market power on the part of firms. These models are based on a labour demand schedule, and a so-called ‘wage setting’ or supply schedule: both are set in real product wage—employment space. Equilibrium (un)employment is at a level where the employer’s mark up on wages is consistent with the real wage anticipated by the workforce. Other levels of employment generate either a rising or falling rate of inflation, and are inconsistent with the realisation of expectations. These levels of employment are not expected to persist. It is typical of these models that the equilibrium level of employment is independent of the level of effective demand.2 Since most of these models assume that the representative firm is always on its labour demand curve,3 its derivation takes precedence. Given a production function Y=F(AL, K) where Y is output, L and K are employment and the capital stock, and A an index of labour augmenting technical progress, let the technology display both constant returns to scale and a constant elasticity of substitution. Then with employment as the decision variable (and A=1), the first order optimising condition for the firm can be written as an equality between the marginal product of employment and (ε/[ε−1])W/P, where ε=the elasticity of product demand, W is the money wage to the employer, and P is the price of value added. This can be written in logarithms (written in lower case and writing the mark-up of price over marginal cost as M=(ε/[ε−1]) as: (3) where α=σ/s and σ=the elasticity in substitution between capital and labour and s is the share of profits in total revenue. The actual outcome for employment from an exogenous increase in the capital stock will however depend upon the specifics of the wage bargaining process captured by the wage setting curve. Two aspects stand out. The slope of the wage setting curve, and the impact (if any) of the productivity gains generated by the increased capital intensity upon the wage setting curve. Layard et al. (1991) and others have considerably developed our understanding of the micro-economics of this first aspect of the wage bargaining process. Many discussions focus on a Nash bargaining solution where wages are maximised by the product of the employer’s and union’s net gains weighted by each party’s bargaining strength. James Nixon and Gioυanni Urga in Chapter 10 adhere to a similar paradigm. In general formulations the net gain for a unionised workforce will depend upon probabilities of obtaining employment elsewhere and upon benefits obtainable if a worker were to enter a spell of unemployment. As Nixon and Urga show, the tradeoff for employers and the workforce in bargaining a higher wage depends naturally enough upon the elasticity of employment and profits with respect to the real wage. Note that equation (3) and the production function are sufficient to determine these elasticities. It can be seen that their value depends upon the elasticity of substitution between capital and labour (σ), the elasticity of demand in the product market (ε), and the share of profits in total revenue (s). In the special Cobb-Douglas case, with σ=1, the share of both profits and wages is constant with respect to the

INVESTMENT, FEEDBACK AND SPILLOVER

103

Figure 7.4 The impact of capital deepening on employment.

real wage, and (with given elasticities of both product demand and substitution), the wage setting schedule is completely inelastic. However, much (if not most) empirical work suggests (see for example the review of estimates in Rowthorn 1996), that the Cobb—Douglas assumption, while frequently being used as a benchmark, is a considerable overestimate of ex post substitution possibilities.4 If indeed σ<1, the wage setting curve will be positively sloped (less than completely inelastic). Thus a process of capital deepening would not only raise the real wage but also the wage share. In the general context of model where there is conflict over income shares, such a rise in wage share could act as a buffer against such forces as changes in bargaining strength, which would otherwise require an increase in the equilibrium unemployment rate. But the process does not necessarily end here, since the productivity gains implied by the bargain may react back on the wage setting curve, a process discussed by Sargent (1995); this represents a potential barrier to the employment generating possibilities of an increase in the capital stock. He considers a once and for all rise in the capital-labour ratio (as opposed to a permanent rise in its growth rate), contrasting between various more or less ‘optimistic’ scenarios. These depend upon how the rise in productivity reacts upon the wage setting curve in subsequent periods5 and are illustrated in Figure 7.4. The initial impact of capital deepening is to raise the price setting schedule from P(K0) to P(K1) and hence (ceteris paribus) to shift the equilibrium level of employment from e0 to e1 with higher real wages and higher employment. In the most optimistic case, the wage setting curve is unaffected by the rise in real wages generated by the increase in the capital-labour ratio. Here the outcome is for real wages to rise on a ‘one-off basis, as implied by a movement along the given wage setting curve. But what happens in subsequent periods? A more pessimistic conclusion follows if the rise in productivity raises the wage setting curve in the next period to the extent that they were actually raised in the first period. In a new equilibrium, these aspirations cannot be realised so the rise in wages is somewhat less than that achieved in period 1. Nevertheless, wages rise to w2 in this latter scenario, so some of the reduction in unemployment is reversed. In a super-pessimistic scenario, the real wage aspirations of employees in this subsequent period are raised to the full extent that they believe that real wages could have been increased in the initial period without any increase in employment, i.e. to w3. Here, all the reduction in unemployment is reversed and this corresponds to a ‘pure insider’ model where the impact of capital accumulation on employment is reversed through its impact on the wage setting curve. Sargent presents no evidence as to which of the above scenarios might obtain, but does put an interesting gloss on the modern emphasis on the need for flexibility in the labour market. First he recalls that, in the UK at any rate, a high rate of growth of investment, rather than budget deficits, was the proximate cause of full employment in the 1950s and 1960s (on this see Matthews 1968). Moreover, this acceleration in investment did not trigger any immediate rise in inflation. Paradoxically, this combination was possible precisely because of the real wage inflexibility engendered by collective bargaining. With the acceleration of investment, the ‘warranted’ real wage as implied by the technology in (1), was constantly moving ahead of the actual real wage creating a negative real wage gap. In addition, more rapid improvements in productivity in some sectors rather than others resulted, not in higher profits or wages, but by falling relative prices, so that gains from productivity were distributed quite generally through lower prices. The possibility of such a benign impact of an investment boom on inflation has if anything been lessened by structural changes to the labour market in the UK in recent years. The absence of a national norm or target for real or money wage growth, or one compatible with a full employment policy, has interacted with the increasing prevalence of company level pay negotiations and the widespread belief that wage increases should be paid for by productivity increases. If anything, this could be imparting an inflationary bias to the current UK labour market, since those firms and sectors experiencing low productivity growth will find themselves having to grant ‘inflationary’ wage increases (or face recruitment difficulties) if they cannot match wage increases in the more progressive industries.

104

CIARAN DRIVER AND PAUL TEMPLE

In their chapter Nixon and Urga take up the point that equilibrium unemployment will, outside of the special case of an elasticity of substitution equal to one, be dependent upon the capital stock (and hence relative factor prices). Increases in capital intensity may therefore allow for a reduction in the equilibrium unemployment rate, but the actuality will depend upon the nature of the wage bargain. They present important new evidence for the UK on the degree of factor substitutability, and upon the nature of the wage bargaining process. To generate estimates of the substitutability between labour and capital for the UK, they utilise a translog cost function defined over four factors of production which imposes no a priori restriction on any of the elasticities of substitution. Their results suggest an elasticity between capital and labour of 0.4, which is broadly in line with a survey of such estimates contained in Rowthorn (1996). Their model of wage determination is derived from Manning (1993) and based on an ‘insider’ framework in which those employed are powerful enough to capture productivity gains, whatever their source, including changes in factor intensity. For simulation purposes, the estimated model is closed by a relationship between aggregate demand, a measure of fiscal policy, and the real money supply. The final and most interesting simulation they report shows that an increase in the real cost of capital raises equilibrium unemployment, despite the favourable effect coming from the substitution of labour for capital. It is perhaps surprising that this simulation suggests that there is a favourable short-term employment response from this source, since we might generally expect changes in factor proportions to take time to work through, at least when compared with the impact on aggregate demand. This impact is ultimately swamped by the reduction in aggregate demand. Nevertheless, they take this result as an indictment of the European policy regime which has emphasised the use of real interest rates to control inflationary pressures; this result probably needs to be benchmarked against a straightforward deflationary fiscal policy. It is perhaps worth commenting on the difference between Nixon and Urga and the related work of Rowthorn (1995, 1996). The difference broadly corresponds with the optimistic versus pessimistic outcomes discussed above. The empirical estimates of Nixon and Urga appear to favour a rather pessimistic outcome with labour’s target wage increasing pari passu with any such gain. For an empirical matter of such importance it is clear that much more work needs to be done, and the Nixon and Urga position may in the end prove rather extreme. Certainly if it were true, capital deepening would be self-defeating for firms. In such a case business would tend to emphasise capital widening. However no such trend has been observed—if anything the converse.6 Some implications This overview has focused on two consequences of investment which would raise social rates of return above the private rate of return and therefore form at least a basis for policy intervention. Some of the relevant policy considerations are considered in Part III of this book. The first type of process driving a wedge between private and social returns was the presence of spillovers, the existence of which forms one basis for some of the theorising that surrounds endogenous growth theory. This burgeoning (if not entirely new) branch of literature has already had an important impact on the way economists, and to some extent governments, think about economic progress. The second type of process involves the relationship between capital accumulation and employment, and the role that investment might play in a full-employment policy. Here what is remarkable is the continuing emphasis in both theory and policy on the role and reform of the labour market and its institutions. For both types of process what is clearly still missing are the solid empirical foundations without which policy interventions are unlikely to be effective. The contributions in this section go some way towards redressing the balance. Notes 1 Attempts to pin down the nature of external effects by observing the influence of aggregate activity on industry performance have not been entirely successful: see the stream of literature stemming from Caballero and Lyons (1992). 2 An alternative model which does feature the level of aggregate demand is that of Arestis and Sawyer (1998) 3 This assumption is often referred to as that of the ‘right to manage’. It is frequently assumed that bargaining between employers and workers takes place over the real wage but not employment itself. Management then treats the real wage as the basis of an employment decision taken by itself alone. 4 We may note in passing that, as Rowthorn himself asserts (see Rowthorn 1995), the elasticity of substitution is likely to be lower in manufacturing than in services. This must present the possibility of sample selection bias in Rowthorn’s 1996 survey of estimates, since a much higher proportion of empirical work relates to manufacturing. 5 In Sargent’s discussion he assumes that there is a ‘background’ rate or growth of real wages set by the growth of labour augmenting technical change (the A parameter above). In this discussion we suppress this for simplicity.

INVESTMENT, FEEDBACK AND SPILLOVER

105

6 Moreover, it may not be reasonable to treat the capital stock as exogenous. While an increase in the wage share may help to reduce inflationary pressure, this may in its turn react back on the inducement to invest. However, there may be some leeway arising from the historically very high levels of real rates of interest now being experienced.

References M.Abramovitz (1986) ‘Catching up, forging ahead, and falling behind’, Journal of Economic History, 46, 385–406 P.Aghion, and P.Howitt (1998) Endogenous Growth Theory, Cambridge: MIT Press P.Arestis and M.Sawyer (1998) ‘The macroeconomics of New Labour’, Jerome Levy Foundation working paper C.R.Bean (1989) ‘Capital shortages and persistent unemployment’, Economic Policy, 8, 12–58 C.R.Bean (1994) ‘European unemployment: a survey’, Journal of Economic Literature, 32, June, 573–619 C.R.Bean, and A.Gavosto (1990) ‘Outsiders, capacity shortages, and unemployment in the United Kingdom’, in J.H.Dreze and C.R.Bean (eds) Europe’s Unemployment Problem, Cambridge: MIT Press G.Bombach (1985) ‘Post-war economic growth revisited’, Professor Dr.de Vries Lectures in Economics: theory; institutions, policy, vol. 6, Amsterdam: North Holland R.Caballero and R.Lyons (1992) ‘External effects in US procyclical productivity’, Journal of Monetary Economics, 29, 209–26. P.Clark (1993) ‘Tax incentives and equipment investment’, Brookings Papers on Economic Activity, 1, 317–48 N.F.R.Crafts (1996) ‘Post-neo-classical endogenous growth theory: what are its policy implications?’, Oxford Review of Economic Policy, 12(2), 30–47 J.B.DeLong, and L.H.Summers (1991), ‘Equipment investment and economic growth’, Quarterly Journal of Economics, 106(2), 445–502 J.B.DeLong and L.H.Summers (1992) ‘Equipment investment and economic growth: how strong is the nexus?’, Brookings Papers on Economic Activity, 2, 157–211 P.O.Demetriades, P.Arestis and C.Kelly (1998) ‘New evidence on the endogenous growth debate’ University of East London, Department of Economics working paper 12, March E.F.Dennison (1964) ‘The unimportance of the embodied question’, American Economic Review, 54 C.Driver and B.Whelan (1998) ‘Investment and business risk in Irish manufacturing’, in H.Kanterelis (ed.) Business and Economics for the 21st Century, New York: Business Economics and Society International. H.Ergas (1987) ‘Does technology policy matter?’, in B.R.Guile, H.Brooks (eds) Technology and Global Industry, Washington: National Academy Press R.Griffith (1999) ‘Productivity and foreign ownership in the UK car industry’, (mimeo), Institute for Fiscal Studies. Z.Griliches (1973) ‘Research expenditures and growth accounting’, in B.R.Williams (ed.) Science and Technology and Economic Growth, London: Macmillan G.M.Grossman and E.Helpman (1991) Innovation and Growth in the Global Economy, Cambridge: MIT Press G.M.Grossman and E.Helpman (1991a) ‘Quality ladders in the theory of growth’, Review of Economic Studies, 58, 43–61 D.A.Hay and G.S.Liu (1997) ‘The efficiency of firms: what difference does competition make?’, Economic Journal, 107, 597–617 E.Helpman (1993) ‘Endogenous macroeconomic growth theory’, European Economic Review, 36, 237–67 C.I.Jones (1995) ‘R&D based models of economic growth’, Journal of Political Economy, 103, 759–84 C.I.Jones (1995a) ‘Time series tests of endogenous growth models’, Quarterly Journal of Economics, 110(2), 495–525 R.Layard, R.Jackman and S.Nickell (1991) Unemployment: Macroeconomic performance and the labour market, Oxford: Oxford University Press R.C.Levin, A.K.Klevorick, R.R.Nelson and S.G.Winter (1987) ‘Appropriating the returns from industrial R&D, Brookings Papers on Economic Activity, 783–820 R.Levine and D.Renelt (1992) ‘A sensitivity analysis of cross-country growth regressions’, American Economic Review, 82(4), 942–63 E.Malinvaud (1980) Unemployment, Cambridge: Cambridge University Press N.G.Mankiw, D.Romer, and D.Weil (1992) ‘A contribution to the empirics of economic growth’ Quarterly Journal of Economics, 107(2), 407–37 A.Manning (1993) ‘Wage bargaining and the Phillips Curve: the identification and specification of aggregate wage equations’, Economic Journal, 103, 98–118 R.C.O.Matthews (1968) ‘Why has Britain had full employment since the war?’, Economic Journal, 78, 555–69 R.C.O.Matthews, C.H.Feinstein and J.C.Odling-Smee (1982) British Economic Growth 1856–1973, Oxford: Oxford University Press N.Oulton and G.Young (1996) ‘How high is the social rate of return to investment?’, Oxford Review of Economic Policy, 12(2), 48–69. R.Rowthorn (1995) ‘Capital formation and unemployment’, Oxford Review of Economic Policy, 11(1), 26–39 R.Rowthorn (1996) ‘Unemployment, wage bargaining, and capital-labour substitution’, Cambridge University, mimeo J.R.Sargent (1995) ‘Roads to full employment’, National Institute Economic and Social Review, February, 74–89 F.M.Scherer (1991) ‘International R&D races: theory and evidence’ in L.G. Matteson and B.Stymne (eds) Corporate and Industry Strategies for Europe, Barking: Elsevier Science Publishers BV R.M.Solow (1997) Learning from ‘Learning by Doing’, Stanford: Stanford University Press

8 Components of investment and growth Jerry Coakley and Andrew Wood

Introduction Growth theory has returned to the political and research agenda in recent years. Interest in the topic has received a stimulus from the extensive empirical literature on the links between investment and growth and developments in endogenous growth theory. In particular attention has focused on the relative contributions of different forms of investment—human capital, research and development, structures and equipment—to economic growth. Simplifying somewhat, endogenous growth theory highlights that increasing returns in the form of spillover effects from these components of investment results in their respective social returns exceeding their private returns.1 Two strands of empirical research have emerged from the new growth literature. One attempts to identify precisely which factors are relevant for economic growth whilst the other seeks to quantify the social rates of return attributed to different types of investment. This chapter focuses on the large empirical literature on the determinants of growth. This has produced a number of contradictory findings. For instance, Levine and Renelt (1992) have argued that the aggregate physical investment ratio is one of the more robust cross-sectional determinants of economic growth. By contrast DeLong and Summers (1991, 1992, 1994) dispute this role for aggregate investment and controversially argue that equipment investment is the only form of physical investment that can improve the growth record of an economy and that investment in private and public non-residential structures make no contribution. This chapter addresses this issue by undertaking cointegration analysis of output per capita and the components of investment using a long span of data for six OECD countries. It is argued below that the equipment— growth nexus is not a unique association and is not based on a simple, uni-directional, causal relationship. The empirical methodology of DeLong and Summers is questioned and an alternative methodology based on time series techniques is suggested. Empirical results are then reported, followed by a conclusion. The relative importance of investment in equipment and structures Given that the new approaches to economic growth have identified investment as a long-run determinant of growth it is natural to wish to identify which forms of investment have the greatest beneficial impact. Theoretical models have been based on the view that investment in human capital, physical capital and R&D can all result in positive externalities. It is the task of the empirical researcher to corroborate and attempt to quantify these externalities. An influential contribution to this debate has come from DeLong and Summers (1991, 1992, 1994) who argue that the impact of fixed investment on economic growth has been under-estimated due to an aggregation problem in which diverse forms of fixed investment are lumped together. By disaggregating investment into equipment and non-equipment they claim to explain statistically many of the differential growth experiences of industrialised and developing economies. More controversially, their results show that only equipment investment is related to economic growth and that buildings investment in non-residential structures has a zero or negligible association with growth. They establish a positive association between equipment investment and growth over different time periods and across different groups of countries controlling for a selection of other variables. This led the authors to describe the relationship as a ‘robust stylised fact’ (DeLong and Summers 1992, p. 174), a description that is increasingly accepted in the literature (Abel 1992), although Oulton and Young (1996) dispute this view for a sample of leading industrial countries. DeLong and Summers claim that the equipment–growth nexus is special because it is based upon a causal relationship in which the direction of causality runs entirely from equipment investment to growth. This allows them to interpret the parameters of their estimated regressions as social rates of return. For example, countries with market-orientated policies are estimated to have a social rate of return to equipment investment of the order of 30 per cent, far in excess of a reasonable approximation of the private rate of return. This striking result leads them to argue that there is ‘a strong case…for at least a modest bias in favour of equipment investment’ (DeLong and Summers 1992, p. 159). Rather than using estimates of capital stock, DeLong and Summers use gross investment as the explanatory variable in their empirical work. They rationalise this by arguing that increases in capital formation are likely to embody the latest

JERRY COAKLEY AND ANDREW WOOD

107

technologies and incremental adaptation of new ideas. In particular, new fixed equipment investment is seen to embody the complex learning process in which new technologies are incorporated into production techniques (Mowery and Rosenberg 1989). The link between new technology and higher economic productive potential is brought about by a process of trial and error. Therefore new investment plays a critical role since it includes not only the adoption of new technologies, but also the adaptation and modification—‘improvement engineering’—which is invariably required to successfully incorporate new technology in production (Rosovsky 1972). This process of innovation through learning-by-doing results in spillover effects which benefit future users of the new technology and consequently drives a wedge between the private and social rates of return of new investment.2 Although many of the arguments put forward by DeLong and Summers are persuasive, the complete dismissal of the role of other forms of investment appears difficult to justify beyond the narrow confines of their empirical results. At its most extreme, their conclusion is that not only does new building investment play no direct role in the growth process but it may in many cases have a negative effect. Intuitively it is unclear why building investment cannot yield positive external economies. For example, to what extent can equipment investment provide large rates of return without accompanying building investment, be it investment in more appropriate, modern buildings to contain the new equipment or in new infrastructure? A variation on this theme is suggested by Aschauer (1989, 1993) who argues that the post-1970 productivity slowdown in the United States can in part be explained by the downturn in the stock of public capital.3 Aschauer argues that public investment not only has a direct effect on growth, but also an indirect one by inducing an increase in the rate of return to private investment and thereby stimulating an increase in private investment. These types of feedback mechanisms are ruled out by DeLong and Summers on the basis of low partial correlations between equipment investment and public investment (DeLong and Summers 1992, p. 175). While the findings of Aschauer has been subject to a great deal of subsequent research,4 those of DeLong and Summers have until recently received a considerably more favourable response. This may be explained by the apparent robustness of their empirical evidence and the intuitive appeal that machinery and equipment are, in some sense, more important than mere bricks and mortar. And yet, as Aschauer demonstrates, external economies are not exclusive to equipment investment, and indeed the productive contribution of equipment may depend to some extent on the availability of an efficient infrastructure. It follows that the claim that equipment investment is special rests largely on the empirical results of DeLong and Summers. The methodology that these results are based on should therefore be considered carefully and compared to alternative methodologies, specifications and data sets. DeLong and Summers estimate cross-section regressions with the average annual growth rate of output per worker as the dependent variable and average equipment investment and non-equipment investment shares of GDP and the growth rate of the labour force as the independent variables, controlling for the productivity gap with the leading countries. In DeLong and Summers (1991) the data was averaged over ten, fifteen, and twentyfive year periods (1975–85; 1960–75; and 1960–85). In a later paper (DeLong and Summers 1992) they repeated the exercise for the 1950s and 1985–90, in addition to estimating a ‘very long run panel’ which consisted of eight periods of approximately fifteen years (1870–85; 1885–1900; 1900–13; 1913– 29; 1929–38; 1938–50; 1950–65; and 1965–80) for eight countries.5 All the results essentially give the same message; there is a strong association between equipment investment and growth but not between other forms of investment and growth. This approach raises a number of issues. It is striking that their regressions are based on cross-section data, for what is essentially a time series issue—explaining economic growth.6 Also of concern is that the data are averaged over a maximum of twenty-five years. While this may be regarded as the ‘long term’ for most economic series in so far as two to three cycles can work through completely allowing a secular trend (if relevant) to be observed, the long run behaviour of investment in built structures is very different. While GDP and investment in equipment are typically characterised by fairly short cycles of between five and nine years, the underlying path of investment in built structures often contains long swings of twenty-five years or more (Figures 8.1 and 8.2 contain trends and actual data for the UK). For example, most industrialised countries experienced very rapid increases in building investment following the Second World War which continued to rise until it peaked for most countries during the late 1960s since when real investment in buildings has tended to fall while the share of buildings investment has fallen or has levelled off (Ball et al. 1996). Past experience suggests that such stagnation of building investment cannot be expected to continue indefinitely. The existence of relatively long building cycles may go some way towards explaining DeLong and Summers’ argument that building investment has little or no effect on economic growth. The relatively short periods over which they average their data are clearly more conducive to demonstrating a relationship between output growth and equipment investment because they have similar, relatively short cyclical characteristics. The longest time period used by DeLong and Summers (1960–85) covers the post-war peak and subsequent long decline of building investment and is therefore unlikely to result in a strong positive association between building investment and output growth. The final area of concern is that of causality. DeLong and Summers are acutely aware that the strength of their result, and therefore the weight of their prescriptions, rely crucially on the claim that the direction of causality runs from equipment investment to growth.7 This implies that higher investment in equipment leads to higher growth rates and not that higher growth

108

COMPONENTS OF INVESTMENT AND GROWTH

Figure 8.1 UK GDP and investment in buildings.

Figure 8.2 UK GDP and investment in equipment.

rates lead to higher equipment investment rates. It is very difficult to use econometric techniques to provide convincing and unambiguous evidence of the direction of causality between two variables. While tests for Granger-causality and exogeneity can be applied to time series data no comparable methods can be applied to cross section data. DeLong and Summers go to some trouble to demonstrate that their equipment-growth nexus is causal. Their strongest evidence is a negative relationship between the relative price of equipment investment and average growth rates and equipment shares. They argue that if higher growth is leading to higher equipment investment, one would expect this to be related to a rise in the relative price of equipment. However, their data are confined to average growth rates and equipment shares on the one hand, and the relative price of equipment at a single point in time on the other.8 They therefore demonstrate that on average countries with higher equipment investment shares also have low equipment prices, but it is very difficult to draw any conclusions about the dynamic relationship between the rate of equipment investment and the relative price of equipment from this evidence. Data and methodology Due to the considerations outlined in the previous section, time series techniques are adopted to test for long run relationships between different subsets of fixed investment and output per worker. The data used must cover a long time period so as to take account of the long cycles which may dominate medium term movements in building investment. This restricts our analysis to a comparison of those countries for which reasonably reliable long run data sets have been collected—the US, Canada, Japan, UK, Finland and Germany.9 For these countries we have data for output per worker and gross fixed capital formation separated into two sub-categories: equipment and non-residential buildings. We also have the corresponding estimates of the capital stock for the US, UK, Germany and Japan. The time period covered varies from country to country and between assets (see Table 8.1). The results of DeLong and Summers are based on an apparently ad hoc statistical equation relating output per worker to investment shares and a sample of other possible explanatory variables. (3.1)

JERRY COAKLEY AND ANDREW WOOD

109

This specification can be derived from a conventional Cobb-Douglas production function with the somewhat restrictive implicit assumptions of constant returns to scale and constant capital-output ratios. The central theme of this paper is twofold: to test the uniqueness of the equipment–growth nexus, and to test the endo– exogenous division implied by (3.1). To do this a VAR methodology is adopted based on a re-formulation of (3.1) which exploits the common order of integration of the levels of Y/N, E and B.10 First, the following bivariate VARs are estimated: (3.2) Table 8.1 Data coverage GDP per worker: Y/N

Equipment: E

US 1869–1992 1890–1992 Canada 1890–1992 1926–1992 Japan* 1885–1990 1885–1989 UK 1856–1992 1856–1992 Germany 1860–1992 1920–1991 Finland 1860–1992 1860–1992 Note: * Japan has missing data for 1941–1951.

Buildings: B

Equipment, capital stock: KE

Buildings, capital stock: KB

1869–1992 1926–1992 1885–1989 1856–1992 1860–1992 1860–1992

1890–1989 n.a. 1890–1989 1856–1991 1935–1990 n.a.

1890–1989 n.a. 1890–1989 1856–1991 1935–1990 n.a.

(3.3) (3.4) Tests for cointegration will allow us to determine if there is a long-run, linear relationship between output per worker and investment. The DeLong and Summers ‘stylised fact’ suggests that cointegrating relationships exist between Y/N and E (3.2) but not between Y/N and B (3.3). The loadings of the cointegrating vectors (αi) can be analysed in order to test the endo—exogenous split of (3.1). If the causality does run from equipment investment to output per worker, we would expect α1 in equation (3.2) to be significantly different from zero but α2 to be insignificant. Alternatively, if new building investment has a positive effect on equipment investment we would expect α1 in equation (3.4) to be significant. The results from these equations are then complemented by first combining the three variables and carrying out exclusion tests on both the cointegrating space and the loading matrix. Then the investment data are replaced by estimates of the capital stock and four variable VARs consisting of Y, KE, KB and N are estimated. Results Bivariate VARs Table 8.2 presents the results for pairwise regressions between Y/N and E and between Y/N and B, over the whole sample and over the post-war period. The lag of the VAR was chosen to be just sufficient to ensure the residuals have no serial correlation. The results are varied, but it is apparent that the association between equipment and output is only marginally more robust than that between building investment and output. While cointegration between Y/N and E is rejected in three out the ten cases, cointegration between Y/N and B is rejected in only four cases. The relationship between equipment investment and output is clearly not unique since Table 8.2 reveals very long run, linear relationships between output per worker and different components of investment. These results cannot be explained by reference to a stable (mean stationary) long-run investment share because over the periods covered investment and its components parts do not take a constant share of GDP per worker (hereafter GDP). This suggests that the identified cointegrating vectors may signify some form of causal relationship. Table 8.2 also shows evidence of a strong relationship between investment in equipment and investment in buildings. This is especially evident

110

COMPONENTS OF INVESTMENT AND GROWTH

Table 8.2 Pairwise regressions: trace and eigenvalue test statistics Whole sample Max. eigenvalue 1 Y/N and E UK 29.88 US 15.48 Japan 9.15 Germany 12.85 Finland 13.44 Canada 21.89 2 Y/N and B UK 21.44 US 12.70 Japan 5.47 Germany 10.88 Finland 21.38 Canada 8.81 3 E and B UK 15.20 US 13.60 Japan 7.08 Germany 6.17 Finland 22.52 Canada 1.29 Note: Bold indicates significance at the 90 per cent level.

Trace

Post-war Max. eigenv value

Trace

29.93 16.87 9.39 13.57 14.59 22.28

10.17 9.39 20.73 21.82 19.95 16.18

12.40 11.72 23.16 27.55 22.71 21.33

21.46 13.36 6.10 12.87 22.20 8.90

3.90 7.64 17.37 16.60 11.18 14.30

5.85 11.42 22.75 27.73 14.37 20.06

16.45 15.23 7.72 6.84 23.42 1.33

3.81 14.10 14.86 30.08 26.61 21.13

6.46 14.95 14.87 30.56 30.12 21.53

during the post-war period when the association is rejected only for one country, the UK (the post-war buildings boom and slump is particularly long and pronounced in the UK). While this does not conform exactly with Aschauer’s thesis since the buildings category includes both public infrastructure and private structures, it does suggest more subtle feedback mechanisms between equipment investment and non-equipment investment than is accepted by DeLong and Summers. Tests of significance of the loading parameters are presented in Table 8.3. The claim that the equipment-growth association represents a relationship running from higher equipment investment to higher growth and not an accelerator relationship running in the opposite direction is supported by post-war Germany. However an accelerator-type relationship is much more evident overall. Surprisingly, the loading parameters of equation (4) indicate that investment in buildings is more likely to adjust to restore the equilibrium between E and B. Only post-war Japan, Finland and Germany offer any evidence in favour of a positive effect whereby building investment may encourage higher equipment investment.11 Table 8.3 Pairwise regressions: tests for significance of loading parameters (t-ratios) Whole sample α1 1 Y/N and E (equation 2) UK US Japan Germany Finland Canada 2 Y/N and B (equation 3) UK US Japan Germany

Post-war

α2

α1

α2

2.92 1.14 n.a. 0.77 2.38 1.92

3.13 2.19 n.a. 3.06 2.31 4.94

n.a. n.a. 5.23 5.06 0.30 1.82

n.a. n.a. 2.14 0.63 4.86 1.61

3.88 1.66 n.a. 0.99

3.18 2.95 n.a. 3.01

n.a. n.a. 4.41 4.48

n.a. n.a. 3.20 1.53

JERRY COAKLEY AND ANDREW WOOD

Whole sample α1

111

Post-war

α2

α1

α2

Finland 3.04 2.42 0.48 3.44 Canada n.a. n.a. 2.17 4.00 3 E and B (equation 4) UK 0.56 3.62 n.a. n.a. US 1.03 2.92 0.13 3.36 Japan n.a. n.a. 2.96 4.13 Germany n.a. n.a. 2.48 6.56 Finland 4.17 2.54 3.15 2.56 Canada n.a. n.a. 0.12 4.31 Notes: n.a.: no cointegrating relationship so not available. Bold indicates the t-ratio is significant at the 95 per cent level.

Three variable VAR analysis The next stage combines Y/N and the two investment variables together in a VAR. Table 8.4 shows that at least one cointegrating relationship was found for each country covering the longest possible time span. Table 8.5 shows the long run parameters normalised on Y/N. The choice of Y/N as the normalisation variable does not imply that it is the endogenous variable; it is merely done for convenience. The often used procedure of choosing different variables on which to normalise in order to make sense of the parameters as structural equations is not used here because that would detract from the central objective of the chapter: to identify long-run relationships, and to describe the adjustment mechanism which ensures the stability of those systems. Inspection of the loading parameters in Table 8.5 suggests that it would be wrong to treat the relationship as a single equation with Y/N as the endogenous variable. Indeed, the results for the US and pre-war Japan suggest that E is the only endogenous variable while B and Y/N are both Table 8.4 Trace and maximum eigenvalue test statistics H0

UK

US

Japan pre-war

Japan post-war Germany

Finland

Canada

1 Maximum eigenvalue test statistics r=0 49.37 16.75 16.22 23.73 21.87 30.03 24.24 r=1 20.25 10.50 4.80 7.59 10.07 14.16 2.11 r=2 0.01 0.71 0.25 0.11 3.66 0.23 0.13 2 Trace test statistics r=0 69.63 27.96 21.28 31.42 35.59 44.42 26.48 r=1 20.26 11.21 5.06 7.70 13.73 14.39 2.24 r=2 0.01 0.71 0.25 0.11 3.66 0.23 0.13 Notes: The lag of the VAR is chosen in each case to be sufficient to remove serial correlation from the residuals of the dynamic equations. Bold test statistics indicate significant at the 90 per cent critical value. Table 8.5 α and β matrices from cointegrating VARs UK β2t αlt α2i US αi Japan pre-war αi Japan post-war αi

βli 1.00 −0.04 (1.65) 0.01 (3.51) βi −0.03 (0.68) βi −0.07 (0.95) βi −0.15 (5.18)

Y/N

B

E

1.00 −3.33 0.06 (0.49) −0.05 (3.25) 1.00 0.15 (1.34) 1.00 0.28 (1.14) 1.00 −0.34 (3.77)

−0.23 3.16 0.51 (5.33) 0.02 (1.35) 0.21 0.40 (2.44) 0.25 1.10 (4.00) −0.04 −0.23 (2.30)

0.80

0.30 0.15 0.51

112

COMPONENTS OF INVESTMENT AND GROWTH

Y/N Germany βli 1.00 αli 0.05 (1.57) 0.33 (4.69) Finland βli 1.00 β2i 1.00 −0.84 αli −0.02 (2.32) 0.10 (3.68) α2i −0.01 (2.20) −0.01 (0.54) Canada βi 1.00 αi 0.06 (1.85) 0.26 (3.50) Notes: t-ratios in parentheses, bold indicates significance at 95 per cent level.

B

E

−0.80 0.31 (4.32) 2.02 1.36 −0.14 (2.28) 0.11 (2.84) 0.21 0.58 (5.15)

1.45 −0.91

0.53

weakly exogenous. Similarly, Canadian and German Y/N is weakly exogenous. These results suggest that not only would it be mistaken to condition on both E and B due to the implied loss of information (Engle, Hendry and Richard (1983); Ericsson (1994)) but that it is also wrong to interpret long-run parameters estimated in single equation models as rates of return. Formal test statistics for the exclusion of the αi confirm the endogeneity of E within the identified systems (Table 8.6). These results show that investment in equipment does not have a unique association with output per worker, and that it is often a mistake to treat either investment in equipment or investment in buildings as weakly exogenous with respect to explaining productivity. However, as indicated in the introduction it is not obvious that the focus should be on the flow of new investment goods. For this reason the procedure is repeated, substituting capital stock estimates for the investment data. Panel cointegration tests were performed for the countries with continuous observations from 1890–1992 (US, UK, Finland).12Y/N cointegrated with both E and B with ADF(1) statistics of –4.70 and –4.37 respectively. The corresponding tests statistics for the post-war period (1952–92) are –2.69 and –1.13 for the US, UK and Finland, and –3.17 and –1.42 for all six countries. These results confirm that both buildings and equipment cointegrate with per capita GDP. The role of the capital stock Again the emphasis is on estimating the VAR over the longest time period possible. Long-run measures of the capital stock are scarce so the sample of countries is more restricted than the previous section. A four variable VAR was estimated, with employment, N, being added as a separate variable. This resulted in at least one cointegrating vector for each country except the US. However, a cointegrating relationship was found for the US when Y, KE and KB were each divided by N.13 Table 8.6 Restrictions on the α matrix (x2(r)) ΔY/N

ΔB

ΔE

UK 14.22 (0.00) 10.38 (0.01) 27.27 (0.00) US 0.35 (0.55) 0.87 (0.35) 3.78 (0.05) Japan pre-war 0.88 (0.35) 0.95 (0.33) 11.20 (0.00) Japan post-war 15.24 (0.00) 8.88 (0.00) 4.15 (0.04) Germany 1.39 (0.24) 10.81 (0.00) 10.58 (0.00) Finland 9.70 (0.00) 13.02 (0.00) 12.49 (0.00) Canada 3.29 (0.07) 10.32 (0.00) 22.13 (0.00) Note: The probability of falsely rejecting the null that the respective αi=0 are in parentheses. Accept the null if p<0.05. Table 8.7 Trace and maximum eigenvalue test statistics H0

UK

1 Maximum eigenvalue test statistics r=0 49.39 r=1 36.31 r=2 7.15 r=3 0.08 2 Trace test statistics r=0 92.93 r=1 43.54

US*

Japan pre-war

Japan post-war

Germany

20.72 7.70 0.15

37.38 14.27 9.74 0.46

44.95 19.38 16.61 0.31

31.55 17.21 13.30 1.72

28.57 7.85

61.85 24.46

81.25 36.30

63.78 32.22

JERRY COAKLEY AND ANDREW WOOD

H0

UK

US*

Japan pre-war

Japan post-war

113

Germany

r=2 7.23 0.15 10.20 16.92 15.02 r=3 0.08 0.46 0.31 1.72 Notes: The lag of the VAR is chosen in each case to be sufficient to remove serial correlation from the residuals of the dynamic equations. Bold test statistics indicate significant at the 90 per cent critical value. * Variables are as a ratio of total employment. Table 8.8 α and β matrices from cointegrating VARs Y

KB

KE

UK β1i 1.00 −0.32 0.92 β2i 1.00 1.82 1.48 3.57 α1i −0.03 (1.13) −0.03 (0.30) 0.50 (5.49) −0.00 (0.34) α2i −0.03 (4.78) 0.13 (4.37) −0.03 (1.00) 0.00 (0.40) US* βi 1.00 0.95 −0.84 αi −0.02 (0.63) −0.09 (3.69) −0.06 (2.24) Japan pre-war β1i 1.00 −5.47 0.58 β2i 1.00 0.37 0.09 2.07 α1i −0.02 (0.74) −0.00 (0.13) 0.00 (0.38) 0.02 (6.80) α2i −0.69 (3.67) 0.02 (0.54) 0.13 (1.45) 0.05 (1.63) Japan post-war β1i 1.00 0.21 0.32 β2i 1.00 0.58 0.05 1.19 β3i 1.00 0.75 0.75 −0.14 α1i −0.32 (2.65) 0.02 (0.87) 0.39 (4.75) −0.07 (1.77) α2i −0.43 (2.03) 0.04 (1.15) −0.39 (2.75) 0.15 (2.38) α3i 0.20 (3.54) 0.04 (4.29) 0.09 (2.35) 0.01 (0.72) Germany β1i 1.00 0.37 0.24 β2i 1.00 1.34 −0.67 4.80 β3i 1.00 0.79 0.02 0.12 α1i −0.20 (5.06) 0.07 (1.69) 0.03 (0.18) −0.02 (0.78) α2i 0.10 (1.86) 0.00 (0.17) −0.56 (2.40) 0.09 (2.70) α3i 0.02 (0.18) −0.06 (3.67) −0.29 (0.63) −0.00 (0.03) Note: * Variables are divided by N; t-ratios are in parentheses, those in bold are significant at the 5 per cent level.

N −0.21

25.19

1.64

−0.91

Table 8.8 shows the estimated long run parameters and the loading parameters. Table 8.9 shows the test statistics for restricting either KB or KE from the long run relationship. Somewhat surprisingly, and in contrast to the implications of DeLong and Summers’ findings, both KB and KE are necessary to obtain the cointegrating relationships identified in Table 8.7. This again underlines the fact that buildings as well as equipment are important in explaining growth. The loading parameters again show that it is not just Y that responds to shocks. Indeed, the one cointegrating vector found for the US does not enter the dynamic equation for ΔY, indicating that it does not represent a growth equation. The complexity of the causal relationships represented by the cointegrating vectors is confirmed by tests for restrictions on the rows of the α matrix. These are reported in Table 8.10 which indicates that for the UK, the US and post-war Japan neither KB or KE can be treated as weakly exogenous. Conclusion Using time series techniques, cointegrating relationships were found between output per worker and investment in equipment and investment in nonresidential buildings for six OECD economies. Underlying the estimated cointegrating systems is a complex web of feedback mechanisms indicating Table 8.9 Tests for restrictions on the β matrix UK US

KB

KE

33.17 (0.00) 11.19 (0.00)

40.12 (0.00) 6.70 (0.01)

114

COMPONENTS OF INVESTMENT AND GROWTH

KB

KE

Japan pre-war 12.30 (0.00) 7.04 (0.03) Japan post-war 15.44 (0.00) 23.14 (0.00) Germany 12.39 (0.01) 16.39 (0.00) Note: The probability of falsely rejecting the null that the respective βi=0 are in parentheses. Accept the null if p<0.05. Table 8.10 Tests for restrictions on the α matrix ∆Y

∆KB

∆KE

∆N

UK 21.17 (0.00) 17.82 (0.00) 27.76 (0.00) 0.23 (0.89) US 0.38 (0.54) 3.17 (0.07) 8.34 (0.00) Japan pre-war 4.81 (0.09) 0.11 (0.94) 1.90 (0.39) 24.25 (0.00) Japan post-war 18.14 (0.00) 16.08 (0.00) 24.89 (0.00) 8.26 (0.04) Germany 20.54 (0.00) 12.11 (0.01) 5.23 (0.16) 6.60 (0.09) Note: The probability of falsely rejecting the null that the respective αi=0 are in parentheses. Accept the null if p<0.05.

that it is a mistake to treat any one of the variables as endogenous. The evidence shows that causality runs in both directions between output per worker and the two forms of investment and between the different forms of investment. Finally, qualitatively similar results were found when the investment data is replaced by capital stock data. The long-run time series results reported in this chapter demonstrate that the association between output per worker and equipment investment identified by DeLong and Summers’ cross-section analysis is not a unique association. Our results cast doubt on the privileged role of equipment investment for growth proposed by DeLong and Summers. The existence of long-run relationships between components of investment and output raises questions about the relevance of the simple neo-classical growth model. The implication of our results for endogenous growth theory requires analysis of social rates of return and these warrant further investigation. Appendix Canada: Investment in buildings and equipment 1926–1992 is from Statistics Canada (1993). GDP and employment are from Urquhart and Buckley (1965) and Statistics Canada (1993). Finland: Investment in buildings and equipment, GDP and employment 1860–1985 are from Hjerppe (1989). Statistics Finland is used to update to 1992. Germany: Data prior to 1960 are for a Germany standardised as being the West German Lande, excluding Saarland and West Berlin, and spliced on to later data which include these regions. Building investment data from 1880 is from Maddison (1993), prior to 1880 the data is from Kirner (1968). Equipment investment is from Maddison (1993). GDP is from Maddison (1991) and is updated using OECD’s main economic indicators. Japan: Data for the 1885–1940 period are from Ohkawa and Shinohara (1979). Data for 1952–1992 are from the Economic Planning Agency. UK: Feinstein (1972) is the source for pre-1950 data. Investment data for the war years were obtained from the estimates of Maddison (1993). The CSO’s economic trends was used to update the series post 1950. US: Building and equipment investment are from US Department of Commerce, Bureau of Economic Analysis. GDP is from Romer (1989). Capital stock: All capital stock data is from Maddison (1993). Notes 1 For a discussion of endogenous growth theory see Romer (1996, chapter 3). 2 A similar version of the relationship between investment and economic growth is that the rate of investment determines the rate of technical advance (Scott 1989, 1992). 3 The data used by Aschauer is for non-military capital stock, 93 per cent of which is structures. 4 See Gramlich (1994) for a survey of this literature. 5 The above papers all used gross investment categories for their independent variables. DeLong (1992) estimated the same model only using net investment in place of gross investment and found the same qualitative results. 6 There are parallels here with another well established stylised fact: the savings-investment association identified by Feldstein and Horioka (1980). Much of the evidence in favour of this association comes from cross-section analyses. More recently the association

JERRY COAKLEY AND ANDREW WOOD

7 8

9 10

11

12 13

115

has been tested using time series techniques with conflicting results; the results of Leachman (1991) and Coakley et al. (1996) are contrary to the finding, while those of De Haan and Siermann (1994) are favourable. A similar point is made by Blomstrom et al. (1993). Figure 3 of DeLong and Summers (1992) shows a negative relation between the average equipment investment share (1960–85) and the relative price of equipment in 1980. Figure 7 of DeLong and Summers (1991) shows a negative relation between average output per worker (1960–85) and the average relative price of equipment in 1975 and 1980. See appendix for data sources. Cointegration analysis forces us to analyse the links between the components of investment and the level of output per worker rather than the growth of output. However, since growth in the level of output is a necessary condition for economic growth, conclusions from our analysis can plausibly be extended to growth. To obtain a more direct comparison with Aschauer’s results we repeated the test using UK data, replacing building investment with general government investment. The result was reversed, with investment in equipment being ‘caused’ by government investment. Not only does this offer some support for Aschauer’s findings, it also highlights the importance of an appropriate level of disaggregation. The panel cointegration tests used have been developed by Im et al. (1995). This specification resembles a Cobb-Douglas production function with constant returns to scale. However, the results reported below indicate it is wrong to interpret the cointegrating relationship in this way.

References Abel, A.B. (1992) Comments and discussion, Brookings Papers on Economic Activity, 2, 157–211. Alexander, W.R.J. (1994) The investment-output ratio in growth regressions, Applied Economics Letters, 1, 74–6. Aschauer, D.A. (1989) Is public expenditure productive?, Journal of Monetary Economics, 23, 200–5. ——(1993) Genuine economic returns to infrastructure investment, Journal of Policy Studies, 21, 380–90. Ball, M., T.Morrison and A.Wood (1996) Structures investment and economic growth: A long-run international comparison’, Urban Studies, November, 1687–1706. Blomstrom, M., R.E.Lipsey and M.Zejan (1996) Is fixed investment the key to economic growth?, Quarterly Journal of Economics, 111(1), 269–76. Coakley, J., F.Kulasi and R.Smith (1996) The savings-investment association, Economic Journal, 106(436), 620–7. DeLong, J.B. (1992) Productivity growth and machinery investment: a long run look, 1870–1980, Economic History Review, 52, 307–24. DeLong, J.B. and L.Summers (1991) Equipment investment and economic growth, Quarterly Journal of Economics, 106, 445–502. ——(1992) Equipment investment and economic growth: how strong is the nexus?, Brookings Papers on Economic Activity, 2, 157–99. ——(1994) How robust is the growth-machinery nexus?, in M.L.Baldassarri, L.Paganetto and E.S.Phelps (eds), International Differences in Growth Rates, New York: St Martin’s Press. Denison, E.F. (1991) A new view of economic growth: a review article, Oxford Economic Papers, 43, 224–36. Engle, R.F., D.F.Hendry, and J.-F.Richard (1983) Exogeneity, Econometrica, 51(2), 277–304. Ericsson, N.R. (1994) Testing for exogeneity: an introduction, in N.R.Ericsson and J.S.Irons (eds), Testing Exogeneity: Advanced Texts in Econometrics, Oxford: Oxford University Press. Feinstein, C.H. (1972) Statistical Tables of National Income, Expenditure and Output of the UK 1855–1965, Cambridge: Cambridge University Press. Feldstein, M. and C.Horioka (1980) Domestic savings and international capital flows, Economic Journal, 90(2), 314–29. Gramlich, E.M. (1994) Infrastructure investment: a review essay, Journal of Economic Literature, 32, 1176–96. Hjerppe, R. (1989) The Finnish Economy 1860–1985: Growth and Structural Change, Helsinki: Bank of Finland. Im, K.-S., M.H.Pesaran and Y.Sin (1995) Testing for units roots in dynamic heterogeneous panels, DAE working paper 9526, University of Cambridge. Kirner, W. (1968) zeitreihen fur das anlagevermogen der wirschaftsbereiche in der Bundesrepublik Deutschland, Berlin: Duncker and Humblot. Leachman, L. (1991) Savings, investment and international capital flows, Open Economies Review, 2, 137–63. Levine, R. and D.Renelt (1992) A sensitivity analysis of cross-country growth regressions, American Economic Review, 82(4), 942–63. Maddison, A. (1991) Dynamic Forces in Capitalist Development: A Long-run Comparative View, Oxford: Oxford University Press. ——(1993) ‘Standardised estimates of fixed capital stock: a six country comparison.’ Innovazione e Materie Prime, April. Mowery, D.C. and N.Rosenberg (1989) Technology and the Pursuit of Economic Growth, Cambridge: Cambridge University Press. Ohkawa, K. and Shinohara, M. (1979) Patterns of Japanese Economic Development: A quantitative appraisal, London: Yale University Press. Oulton, N. and G.Young (1996) How high is the social rate of return to investment?, Oxford Review of Economic Policy, 12(2), 48–69. Romer, C. (1989). ‘The pre-war business cycle reconsidered: new estimates of gross national product, 1869–1908’, Journal of Political Economy, 97, 1–37. Romer, D. (1996) Advanced Macroeconomics, New York: McGraw-Hill. Rosovsky, H. (1972) What are the lessons of Japanese economic history? , in A.J.Youngson (ed.), Economic Development in the Long Run, New York: St Martin’s Press.

116

COMPONENTS OF INVESTMENT AND GROWTH

Scott, M.F. (1989) A New View of Economic Growth, Oxford: Clarendon Press. Scott, M.F. (1992) A new theory of endogenous economic growth, Oxford Review of Economic Policy, 8(4), 29–42. Statistics Canada (1993) Canadian Economic Observer. Statistics Finland (1993) Statistical γearbook of Finland. Urquhart, M.C. and K.A.H.Buckley (1965) Historical Statistics of Canada, Toronto. US Department of Commerce (1993) Fixed Reproducible Tangible Wealth, 1926–1989, Washington, DC.

9 Foreign direct investment, innovation and economic growth within Europe Ray Barrell and Nigel Pain1

Introduction Increasing attention has been paid in recent years to the impact of cross-border investments by multinational firms. This reflects both the rapid growth in the importance of foreign-owned firms in many economies, particularly within Europe, along with recent improvements in the quality and availability of data on foreign direct investment. At the heart of the debate is a focus on the costs and benefits of foreign investment, such as whether inward investments affect employment and economic growth and whether outward investment is simply ‘job exporting’, with firms moving to low-cost, labour-abundant locations. An understanding of the motives and consequences of firms’ decisions to invest abroad is of particular importance for the UK, whose aggregate stocks of outward and inward foreign direct investment reached 30 per cent and 21 per cent of GDP respectively at the end of 1995. In this chapter we investigate the factors affecting the growth of multinational firms and foreign direct investment, the national characteristics that affect the pattern of inward investment within European countries and the implications of foreign-owned firms for the economic performance of host economies. How large is foreign investment? Multinational companies have become more influential in the world economy since the 1970s, helped by the removal of many national barriers to capital movements. One illustration is provided by the rapid growth of foreign direct investment (FDI). The aggregate stock of FDI in the world economy is estimated to have risen from 4.5 per cent of world output in 1975 to 9.5 per cent in 1995. In addition the value of sales by the foreign affiliates of domestic companies is now estimated to exceed the value of world exports by around one-quarter (UNCTAD, 1996). The broad pattern of foreign direct investment stocks is shown in Table 9.1. It is clear that the vast majority of investments have been made by firms from the developed economies of the OEGD and are located within Table 9.1 Global foreign direct investment stocks Outward FDI World % of GDP OECD % of GDP % of total EU-15 % of GDP % of total Inward FDI World % of GDP OECD % of GDP % of total EU-15 % of GDP

1980

1985

1990

1995

$ billion 4.9 $ billion 6.8 97.7 $ billion 7.4 41.5

513.7 5.9 501.9 6.1 95.9 213.2 7.1 41.8

685.5 8.1 657.4 10.6 95.4 286.5 13.8 46.1

1684.1 9.7 1606.2 13.2 91.7 777.2 17.4 44.3

2730.1

$ billion 4.6 $ billion 4.8 74.0 $ billion 6.4

481.9 6.3 356.4 4.9 71.6 185.0 5.6

734.9 8.3 526.3 9.0 79.3 226.5 12.7

1716.9 9.4 1361.4 10.1 72.3 712.2 14.8

2657.9

2503.2

1208.8

1922.0

1028.1

118

RAY BARRELL AND NIGEL PAIN

1980

1985

1990

% of total 38.4 30.8 41.5 Source: Authors’ calculations from UNCTAD (1996), and OECD National Accounts.

1995

38.7

other developed countries. This geographical distribution of existing investments suggests that multinational investment cannot simply be characterised as the movement of production to low wage economies. The global level of FDI has risen particularly sharply since the middle of the 1980s, especially within Europe in the aftermath of the common deregulation of national capital and product markets. 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. During this time there have been rising levels of investment in the EU both by non-EU nationals and EU firms themselves. The proportion of investments held within the OECD economies has fallen notably since 1990, suggesting that there has been some relocation towards developing economies, especially China, but is still no higher than it was in the first half of the 1980s. The pattern of investment flows within the OECD is summarised in Table 9.2. The largest foreign investors are the United States, the UK, France, Germany and Japan. Whilst the first three of these economies also receive a high level of inward investment as well, less investment takes place in Japan. The situation of Germany is unclear; recorded FDI inflows are low, but this may in part reflect differences in their national definition of the inflow of direct investment (Jost, 1997). The change in the stock of inward investment in Germany between the end of 1990 and the end of 1995 is $92 billion, considerably greater than the recorded capital inflow over this period. Table 9.2 FDI flows and GDP in selected OECD economies ($ billion, period totals) 1976-80

1981-85

1986-90

1991-95

1995 GDP

USA Outflows 82.2 42.8 130.3 292.9 6954.8 Inflows 37.6 92.9 266.8 204.9 Japan Outflows 10.1 25.5 160.4 103.4 5114.0 Inflows 0.6 1.7 1.6 5.2 UK Outflows 39.1 46.1 140.5 127.1 1101.8 Inflows 27.8 21.6 108.7 85.8 France Outflows 9.6 13.5 83.4 116.0 1537.6 Inflows 11.2 10.8 40.4 95.0 Germany Outflows 18.7 21.1 71.9 110.3 2412.5 Inflowsa 5.9 3.8 13.5 18.3 Inflowsb 13.8 22.1 71.5 92.3 Belgium Outflows 2.9 1.0 21.1 33.7 269.2 Inflows 6.3 5.6 23.4 49.1 Denmark Outflows 0.9 1.0 5.5 12.6 173.3 Inflows 0.9 0.4 3.0 13.4 Greecec Inflows 2.4 2.3 3.8 5.3 114.3 Italy Outflows 2.2 8.4 20.4 31.4 1087.2 Inflows 2.7 5.3 19.4 16.3 d Ireland Outflows n.a. n.a. 3.3 11.6 64.3 Inflows n.a. n.a. 6.9 16.8 The Netherlands Outflows 21.0 19.2 50.2 62.9 395.5 Inflows 6.3 7.6 31.9 35.1 Spain Outflows 0.9 1.5 7.4 16.7 559.6 Inflows 5.2 8.9 37.5 49.5 Sweden Outflows 3.0 7.0 41.2 25.3 230.6 Inflows 0.5 1.5 7.0 30.5 Switzerlande Outflows n.a. 6.2 25.7 40.2 306.1 Inflows 4.6 2.6 12.6 11.3 Sources: Direct investment statistics from IMF International Financial Statistics γearbook 1996 and International Financial Statistics, June 1997, and Deutsche Bundesbank Monthly Report, August 1997. GDP from OECD Main Economic Indicators, June 1997.

FDI AND EUROPEAN ECONOMIC GROWTH

119

1976-80 1981-85 1986-90 1991-95 1995 GDP Notes: aFigures from FDI flow data in balance of payments. bFigures based on change in stock of inward investment. cOutflow data not available. dData based on Eurostat figures in European Union Direct Investment γearbook, 1996. eNational data only available from 1983. Data prior to then from US, German and UK statistics.

The flow of investment from EU firms into other EU member states and North America from 1982 to 1994 is described in greater detail in Pain and Lansbury (1997). During the period from 1982 to 1987 investment in North America was greater than investment within the EU. However intra-EU investment exceeded investment in North America from 1988, after the start of the Single Market Programme, reaching a peak of 1.25 per cent of GDP in 1990. Over the five years to 1994, intra-EU FDI flows were equivalent to 4.5 per cent of gross domestic fixed capital formation in the EU on average. Such data appear to suggest that moves towards greater European integration and the eradication of barriers to cross-border capital mobility have generated a significant change in investment patterns. Inward direct investment in countries such as Belgium, Denmark, France and Ireland during 1991–95 was more than twice the level in the previous five-year period. Borders and institutions at the supra-national level seem to have a significant impact on the pattern of investments. Membership of the dominant regional trading bloc appears to matter, particularly for attracting investments from Japanese and American companies (Barrell and Pain, 1998a and b). The level of inward investment into Spain and Portugal rose significantly after their accession into the EU in 1986. Sweden appears to have had a similar experience in more recent years.2However the contrasting experience of Greece since entry into the EU in 1981 suggests that membership of the regional ‘club’ is necessary but not sufficient, implying that national institutions and social capabilities may also affect location decisions (Barry and Bradley, 1997). The scale of direct investment inflows should also be judged against the size of the economies in question. Although the US was the largest recipient of inward FDI between 1985 and 1995, it is also the largest economy and inward investment is relatively low as a proportion of GDP. The UK has the highest ratio of investment/GDP of all the G7 economies. However a number of smaller open economies notably Ireland, Belgium, The Netherlands, Spain and more recently Sweden and Portugal have proportionately higher levels of flows of inward investment. The UK authorities’ claim that the primary factors behind the high level of inward investment into the UK have been the labour market reforms introduced in the 1980s and the pro-business climate adopted by the government during this period, with low levels of corporate taxation and few direct regulatory burdens. Comparisons between the UK and Italy and, to a lesser extent, Germany, might appear to give some weight to this argument. Comparisons with more heavily regulated economies such as France, Belgium and Sweden which have succeeded in attracting a significant level of inward investment suggest that many other factors are likely to be important. Despite its supposed locational advantages, on balance the UK is a net outward investor, both within Europe and elsewhere and within manufacturing and non-manufacturing sectors (Barrell and Pain, 1997a). Indeed UK companies have been the largest single investors in France since 1990. This suggests that foreign investment is about much more than simply moving production to either countries with flexible labour markets and low labour costs per unit of output or countries with large domestic markets, even though such factors do undoubtedly influence investment decisions. The UK is more distinctive in the methods by which foreign investment takes place. Such investment can be by means of cross-border mergers and acquisitions (M&As), or by investment in greenfield sites. Table 9.3 reports the value of crossborder sales and purchases for selected economies. Table 9.3 Cross-border mergers and acquisitions 1991-1995 Country United States Sales Japan UK Sales Germany Sales France Sales The Netherlands Sales

Purchases 188.79 Purchases Sales Purchases 90.53 Purchases 34.57 Purchases 43.75 Purchases 24.13

Total ($ billion)

As % of FDI

180.78 92.1 55.09 5.72 104.15 105.5 55.77 37.5 64.68 46.1 41.19 68.7

61.7 53.3 110.0 81.9 50.6 55.8 65.5

120

RAY BARRELL AND NIGEL PAIN

Country

Total ($ billion)

Spain Purchases 8.39 Sales 20.50 41.4 European Union Purchases 347.17 Sales 273.09 66.9 Source: UNCTAD (1996), Annex tables 7 and 8. FDI data from Table 9.2 above.

As % of FDI 50.2 64.2

Although the growth in the world-wide aggregate value of M&As has mirrored the growth of FDI flows in recent years, it is important to bear in mind that there is no necessary correspondence between the two measures. For example, if a merger is financed by capital raised in the host country it will not be included in the FDI data. In other cases a merger might not be reflected in the FDI figures as it consists of a takeover by one foreign firm of an existing subsidiary of another foreign firm. In this case the level of FDI would not change. It is clear that foreign investments into and out of the UK, the US and The Netherlands are more likely to involve mergers and acquisitions than investments in countries such as Germany, France and Spain, where greater barriers exist in national equity markets. Agarwal (1998) stresses the extent to which the particular structure of corporate equity holdings acts as an impediment to new inward investments through corporate M&As in Germany. The reinvestment of profits by existing investors and the transfer of additional funds from parent companies to their affiliates appear to have been more important sources of inward investment in Germany in recent years (Jost, 1997). The figures in Table 9.3 reflect the fact that new inward investment is increasingly likely to occur by means of mergers and acquisitions rather than through investment in new greenfield sites. One implication of this is that the figures for foreign direct investment may not provide a complete indication of the importance of multinational firms within domestic economies. This is particularly true of foreign investments in the manufacturing sectors of many European economies, where significant involvement has often been built up over a decade or more. Table 9.4 gives some figures for the share of output and employment attributable to foreign-owned manufacturing firms in a number of host economies. In the large European economies around one-fifth of employees work for foreign-owned firms. These firms account for around one-quarter of total output. In some small, open economies such as Ireland and Hungary (and also Belgium), foreign firms account for around 60 per cent of all activity. In addition it needs to be remembered that employment abroad by the affiliates of multinationals from countries such as Germany, France and the UK is even higher than employment by foreign-firms in these countries, suggesting that many firms now have a considerable degree of scope to shift production around from one location to another. Within Europe the practices of many multi-plant car producers provide one example of this pattern (Hubert and Pain, 1998). It is also clear from Table 9.4 that in all countries the average labour productivity of foreign-owned firms is higher than that of domestic firms. In part this may reflect differences in the industry mix, but as we shall see later, there is also evidence that the presence of foreign firms in a particular industry can have a beneficial effect on indigenous firms within that industry. One interesting point from the relative productivity figures is the high level seen in Germany; it seems likely that this partially reflects a degree of ‘technology sourcing’, with foreign manufacturing firms attracted to Germany to benefit from the advanced technologies and skills employed in the chemicals and engineering sectors. Why do firms become multinationals? It is clear that traditional general equilibrium models based on the Heckscher-Ohlin framework (Mundell, 1957) cannot provide a full explanation for the observed pattern of cross-border corporate investments. Such models explain capital flows purely in terms of relative factor endowments. Table 9.4 The importance of foreign-owned firms, manufacturing sectors Country Francea Germanyb Hungaryb Irelandc The Netherlandsa Swedena United Kingdomd

1994 1994 1996 1993 1994 1994 1994

Foreign firms output share (%)

Foreign firms employment share (%)

Foreign/domestic labour productivity

30.8 25.3 61.4 68.4 27.0 16.3 25.2

26.3 16.1 36.1 44.7 23.2 13.6 18.6

1.25 1.76 2.81 2.68 1.22 1.24 1.47

FDI AND EUROPEAN ECONOMIC GROWTH

Country

Foreign firms output share (%)

Foreign firms employment share (%)

121

Foreign/domestic labour productivity

United Statesa 1991 14.0 11.1 1.30 Sources: Holland and Pain (1998, table 12); Hunya (1998); Howenstine and Shannon (1996) and authors’ calculations from Activities of Foreign Affiliates in OECD Countries and OECD National Accounts. Notes: aValue added. bSales. cNet output. dGross value added at factor cost.

Capital abundant economies either export capital-intensive goods, or invest directly abroad if there are strong barriers to trade. If this was the only factor causing capital flows, then we should see capital rich countries, generally those in the OECD with high incomes, exporting capital to poor (in terms of both capital and income) countries. However, investments are predominantly concentrated in the OECD countries rather than in the more capital scarce developing economies. Within the OECD there are, of course, labour-abundant low wage economies, and some, such as Spain, Portugal and Mexico, have specific locational advantages arising from their membership of a regional trade area. However, other key destinations for inward investment in recent years such as the United States, the UK, France, Belgium and Sweden are relatively high cost, at least in terms of wages per hour, and relatively capital rich. It is clear from our research into the investment decisions of US, German, Japanese and British companies that relative unit labour costs, which take account of wage and productivity differences, remain an important element in the investment decision.3 However they cannot account fully for the observed pattern of foreign investment between developed economies in recent years, nor for why individual companies may elect to have multiplant operations rather than produce in a single location and serve foreign markets by trade. This suggests that models which stress relative factor endowments and relative factor costs need to be augmented in order to explain FDI flows from capital rich countries to similar countries. In particular, account needs to be taken of the benefits from the agglomeration of complementary activities when constructing models to explain the location of activities (Krugman and Venables, 1995; Barrell and Pain, 1998). In a world with imperfect competition the size of nations, and the quality and variety of goods they produce, is to be explained, not taken as given. Two key factors in determining the increasing spread of multinational firms within the developed economies appear to be knowledge-based, firm-specific assets4 and the evolving structure of barriers to trade between and within large regional markets such as the EU and NAFTA. With the changing nature of technology and production, firm-specific knowledge-based assets appear to be coming more common. Such assets may act as a joint input across plants, giving economies of scale at the level of the firm rather than at the level of the individual plant (Markusen, 1995). The need to safeguard proprietary knowledge also encourages firms to undertake direct investment rather than license existing foreign firms to undertake production. In some circumstances a single multi-plant firm may therefore have a cost advantage over two separate single plant firms and locate production nearer to final markets, setting up cloned plants using common knowledge-based assets. Such possibilities appear to have increased with the move away from ‘Fordist’ modes of production, based on mass assembly of identical products and exploitation of economies of scale at plant-level, to more customised, smaller-scale production for niche markets. Modern research-based production is increasingly related to blueprints for ‘ways of doing’ rather than blueprints for basic machinery. In a large economy, such as the United States, a firm with a specific product can have a multi-plant operation to serve regional markets without investing abroad. In a large economy such as Europe, which has many countries, a firm whose optimal plant structure requires dispersed production to adapt products for local markets would become a multinational. The gradual removal of non-tariff barriers to trade appears to have stimulated cross-border investment in Europe over the past decade. The expected concentration of tradable activities within a single location in a barrier-free Single Market does not appear to have happened. Estimates from the European Commission (1996, chapter 6) suggest that concentration ratios, measured as the market share of the four largest firms, rose at the EU level between 1987 and 1993, but not at the national level. The results in Pain (1997) and Pain and Lansbury (1997) imply that the Single Market Programme had raised intra-European FDI by German and British companies by $27 billion by 1992 (at constant 1990 prices), or around 0.5 per cent of EU GDP. Empirical work in Pain (1997), Hubert and Pain (1998) and Barrell and Pain (1998) also shows that registered patents and R&D expenditures help to explain the industrial pattern and level of foreign investment by British, German and US multinational firms, with ‘innovating’ companies being more likely to invest abroad. This suggests that multinational companies may be an important channel for the transmission of innovations and ideas (Romer, 1993). Moreover, as information technology is increasingly utilised, process innovations and human resource management (Ferner, 1997) are becoming more important in the improvement of production. In such an environment the inherent advantages of foreignowned firms can often be more easily transferred through entry by mergers and acquisitions. With economies of scale at plant level having become relatively less important than in the heyday of mass production, greenfield investments have become a

122

RAY BARRELL AND NIGEL PAIN

less common mode of entry. The existing market structure in the host economy may also affect the decision over the mode of entry (Görg, 1998). What determines investment patterns in Europe? Empirical and survey evidence also indicates that investment decisions are affected by the characteristics of potential host economies. Labour costs are not as important an element in many investment decisions as they once were, particularly for companies making high technology products. Cultural factors such as language matter as do the skills and abilities of the host country workforce, the quality of infrastructure and the existence of supportive institutional arrangements for knowledge transfer between a well-developed science base and private industry. Such factors can act as centripetal forces, helping to bring an agglomeration of related activities to a particular location (Barrell and Pain, 1998). As we argued above, membership of the European Union also appears to improve investment prospects. From a policy perspective it is of interest to ask whether we can identify factors which affect the type of activities placed in different locations within Europe. One approach is to use econometric evidence to estimate the importance of particular factors in the growth of FDI. In Table 9.5, drawn from the results of a panel data analysis of the foreign investments made by German manufacturing firms in Hubert and Pain (1998), we show that the UK has clearly benefited from changes in the performance of the labour market in recent years, both in terms of an improvement in labour cost competitiveness and through the decline in industrial conflict. However, whilst it is clear that the growth in R&D undertaken by German firms has had a large effect on the level of inward investment received in all locations, the effects have been consistently smaller for the UK and Ireland than for the other locations. The UK appears to have performed relatively poorly in attracting investments from those sectors in Germany where innovations have risen most rapidly. It is also worth noting that many other countries also appear to have gained investment as a result of favourable movements in their relative labour cost position and a decline in the number of strikes, even though labour market institutions in these countries are very different to those in the UK. Perhaps the most interesting results are obtained for host country labour costs relative to those in Germany. There are some interesting distinctions to be drawn between the various locations. Trend unit labour costs in countries such as The Netherlands, France and Austria whose currencies Table 9.5 Accounting for inward-investment growth in Europe (percentage change in stock of inward investment from Germany due to each factor) Host location

Period

Relative labour costsa

Host strikes

Belgium 1984–1989 4.7 0.4 1989–1994 3.0 3.8 18.1 The Netherlands 1984–1989 5.0 −7.9 1989–1994 4.1 4.9 15.9 UK/Ireland 1984–1989 7.4 3.2 1989–1994 8.0 8.2 14.8 France 1984–1989 4.6 3.0 1989–1994 3.9 1.5 16.4 Italy 1984–1989 3.1 2.6 1989–1994 8.4 3.0 17.7 Spain/Portugal 1984–1989 2.1 2.4 1989–1994 1.0 –0.2 21.5 Source: Hubert and Pain (1998). Note: a Unit labour costs in host relative to those in Germany, expressed in a common currency.

German R&D 22.9 18.1 18.5 21.0 19.9 23.7

have maintained a close link with the D-mark are estimated to have risen less rapidly between 1984 and 1994 than those in Germany. In contrast, unit costs in countries such as the UK, Spain, Italy and Sweden actually rose more rapidly than German costs over this period. However this was more than offset by the depreciation of their nominal bilateral exchange rates against the D-mark, particularly in the early 1990s, resulting in a depreciation of their real exchange rates. Thus whilst movements in relative labour costs have helped to raise the level of foreign investment by German companies since the mid-1980s, this has as much, if not more, to do with movements in market exchange rates than with high basic wage costs in Germany or labour market reforms in the host economies. To a large extent the higher level of wage costs in many continental European economies is simply the counterpart to a higher average level of labour productivity. Both factors matter in the location decision.

FDI AND EUROPEAN ECONOMIC GROWTH

123

It is possible to derive a similar story from the detailed statistics available in some source countries on the operations of their foreign affiliates. In particular we can draw on the long-established surveys undertaken into the operations of United States multinationals. Table 9.6 reports figures on the gross product of majority-owned foreign affiliates,5 a value-added measure, along with estimates of their R&D intensity and labour productivity (per head). The key role of United States companies in Ireland, particularly in the manufacturing sector, is clear from these figures. The gross product of United States affiliates was equivalent to nearly 12 per cent of Irish GDP in 1994 and also amounted to over 5 per cent of GDP in Table 9.6 Selected data for US non-bank foreign affiliates in Europe (1995 unless otherwise stated) All sectors

Manufacturing

Gross product ($bn) Employment (000s) Total ($bn)

R&D expenditures (1989–95 average) Gross product ($bn) Value added per heada (1993 prices)

As % of gross product

EU15 273.9 2701 7.66 3.46 139.8 UK 70.6 813 1.85 3.12 28.5 Germany 61.5 549 2.53 5.02 40.1 France 34.7 378 1.02 3.41 19.0 Italy 20.9 184 0.36 1.89 9.5 The Netherlands 17.5 129 0.44 3.12 8.5 Belgium 14.1 101 0.35 3.75 7.7 Ireland 10.0 56 0.45 7.83 7.4 Spain 9.6 136 0.21 2.60 6.6 Sweden 5.1 49 0.16 4.78 2.7 Sources: Calculated from annual articles on ‘Operations of US Foreign Affiliates’, Survey of Current Business, selected issues. Note: aAverage based on four years from 1992–95.

100.0 83.3 111.6 89.0 104.4 129.2 129.8 176.1 83.2 71.5

both the UK and Belgium. Within the manufacturing sector the differences between investment in Britain, Germany and France are less marked, reflecting the high level of investment in oil-related industries in the UK. The activities located within the UK have been relatively employment intensive, accounting for a larger share of employment than output in European affiliates. Labour productivity in manufacturing, measured in terms of gross product per employee over the four year period 1992–95, is over 16 per cent lower in UK affiliates than the European average. Even so, the productivity of US affiliates in the UK appears to be somewhat higher than the productivity of UK firms overall. For example, in manufacturing the US affiliates only account for some 10.5 per cent of total employees in employment in the UK, but produce around 13.5 per cent of value added.6 The research intensity of affiliates in the UK, measured as R&D expenditures relative to sales, is also below the European average, with the most research intensive affiliates being located within Germany, Sweden, Belgium and Ireland. However it is noteworthy that on average the United States affiliates have a greater propensity to undertake R&D expenditures in all countries than domestic firms (as measured by the share of business enterprise R&D in GDP), pointing to a further channel through which multinational firms may benefit host economies. The findings from these detailed statistics mirror those from econometric studies. Both suggest that whilst the United Kingdom has clearly been able to attract relatively labour intensive investments, it has fared relatively poorly in attracting more capital intensive investments, and it has not been particularly attractive for sectors where R&D has been more significant. This is consistent with many cross-country studies of the skill mix of the workforce, a factor discussed at length in Prais (1995), as well as the extent to which increased educational investments have enabled countries such as Ireland to increasingly attract high-tech investments. The United Kingdom has performed comparatively well in attracting investments in the non-manufacturing industries. However, as we discuss below, it is in this sector that the evidence for the existence of beneficial supply-side effects from foreign investment is weakest. The impact of FDI on trade and technical progress Given that more technologically progressive sectors undertake more FDI, it is reasonable to expect that they would alter the production possibilities of the host country by bringing new technologies and ideas as well as additional capital. Indeed, as Tables 9.4 and 9.6 reveal, foreign-owned firms typically have a higher level of labour productivity than domestic ones and undertake more R&D. It would be wrong to suggest that the benefits of FDI must be limited in many developed economies just because a relatively high proportion of investments consist of mergers and acquisitions rather than investments in ‘greenfield’ sites. Take-overs and the associated reorganisation of existing capacity and introduction of new ideas may raise

124

RAY BARRELL AND NIGEL PAIN

the rate of technical progress and hence the long-run rate of economic growth, even if they do not add to final demand directly. Survey evidence suggests that the mode of entry of foreign firms makes relatively little difference to the impact of such firms on domestic suppliers in the UK (PACEC, 1995). The effects of FDI need not be confined to the host company. Microeconomic evidence on licensing and direct investment suggests that inflows of new technology and working practices from the affiliates of multinational firms create a significant potential for spillovers to local firms in the host country (Blomström and Kokko, 1996). Such effects can be direct, if foreignowned firms are more productive or transfer firm-specific knowledge to local suppliers, or indirect, with local firms subsequently able to employ workers whose knowledge and skills have been upgraded as a result of employment within foreign-owned firms. However whilst there is widespread empirical evidence that inward direct investment has raised the growth rate of many developing countries (Balasubramanyam et al. 1996), much less is known about the impact of foreign investments on economic performance in Europe. The high level of inward investment in manufacturing activities has clearly been particularly important in the economic development of many smaller European economies such as Ireland (Barry and Bradley, 1997; Ruane and Görg, 1998) as well as the transition economies in Eastern Europe (Holland and Pain, 1998). Within the United Kingdom, there are frequent claims that inward investment has helped to transform the supply-side of the economy (Eltis, 1996). It is certainly clear that foreignowned companies account for a rising proportion of employment and exports in the tradable sectors of the economy. However this does not provide evidence that the incomes of factors of production owned by residents of the particular economy have been raised above the levels they would otherwise be at in the absence of any such investments. FDI and export performance One channel by which foreign firms change the economic prospects of host economies is through trade. In a world of imperfect competition, new inward investments can help to change the variety and quality of goods and services produced within nation states, factors which are widely thought to be important determinants of trade performance and growth prospects (Grossman and Helpman, 1991). It is important to focus on both inward and outward direct investment in order to allow for the fact that most countries import and export capital simultaneously. The key empirical issue is whether export performance, that is export sales relative to foreign market size, is affected by FDI given the other relative characteristics of domestically produced goods. Whilst it is possible to conceive of situations in which both inward and outward investment may either create or displace exports, on balance the cross-country evidence points to a small negative impact of outward investment on home country export performance, offset by a corresponding positive impact from inward investment on host country export performance. This pattern holds for most large countries, including the United States, the UK, France and Germany, but not for Japan where net outward investment is found to improve export performance (Barrell and Pain, 1997b; Pain and Wakelin, 1998a, 1998b). There is also some evidence that the negative relationship between net outward investment and export performance has strengthened over time, particularly within Europe (Pain and Wakelin, 1998b). In the 1960s and 1970s investment was often undertaken to bypass barriers to market entry, and hence had little effect on exports of finished products. More recently trade and capital market barriers have been lowered by the single market programme and hence investment is more likely to represent a deliberate decision to serve foreign markets from foreign production, at least in the tradable goods sector. At face value these results might suggest that the transmission of ideas through FDI clearly benefits host economies but is not without cost to investing countries. However it would not be appropriate to conclude that outward investment ‘exports’ jobs simply because we find a negative relationship between outward investment and net trade performance. Account also needs to be taken of the factors responsible for the higher level of outward investment. If this has arisen as a result of the creation of a higher level of firm-specific assets, and hence a faster rate of domestic innovation, then domestic economic growth and non-price competitiveness may have been enhanced. Equally, demand may be raised in host economies. Alternatively if outward investment is driven by cost differentials, then domestic production is likely to be reduced. In both cases a full general equilibrium cross-country analysis is required in order to take account of the endogenous linkages between trade, investment and technological change. Inward investment, technical progress and labour productivity In order to provide an illustration of the potential impact of inward investment on the major developed economies we initially look at time series evidence on the determinants of factor demands and technical progress in West Germany. We then consider some cross-sectional evidence for a number of manufacturing industries in the UK, France and the US. Our initial analysis assumes an underlying CES production function of the form: (1)

FDI AND EUROPEAN ECONOMIC GROWTH

125

Here v denotes returns to scale, γ and s are production function scale parameters, and the elasticity of substitution (σ) is given by 1/(1+ρ). If σ=1 (ρ=0), then production is Cobb-Douglas. K and L denote the net capital stock and labour input measured in terms of employee hours. The production function allows for the possibility of both labour and capital augmenting disembodied technical progress. It is possible to obtain estimates of σ, λ and κ using the factor demand equations implied by the marginal productivity conditions that the marginal product of each input should equal its (mark-up adjusted) real price. Thus: (2a) (2b) where w, c and p respectively denote labour costs per head, the nominal user cost of capital and the price of value added (at factor cost) and β denotes the mark-up. If we impose long-run constant returns to scale (v=1) we obtain log-linear factor demand equations of the form: (3a) (3b) If we assumed either that capital augmenting technical change is zero (K=0), or that technical change is ‘neutral’ (κ=λ=0, with γ replaced by eγt), then it would be possible to calibrate the production function using the labour demand equation alone. However we feel it is necessary to test such restrictions and hence we jointly estimate the labour and capital demand equations with appropriate long-run cross-equation restrictions being imposed. Our main interest lies in the extent to which technology transfers and other spillovers from foreign-owned firms affect the pace of technical change and hence economic growth. Endogenous technical change is typically investigated either by introducing specific variables in the production function itself or by endogenising technical progress. Keller (1989) provides a comprehensive overview of the empirical literature. Here we assume that technical progress is dependent on the aggregate level of foreign-owned assets within the economy, the cumulated stock of patents registered by domestic residents, together with an exogenous element proxied by a linear time trend: (4) The appropriate functional form for endogenising technical progress is undoubtedly worthy of further investigation. The specification used here implies that technical progress will grow at a constant rate if the real stock of direct investment and the stock of patents grow at a constant rate. We arbitrarily assume that the FDI and patents variables enter with a four quarter lag. Our model is closely linked to that of Coe and Helpman (1995) who link the growth of total factor productivity to the domestic and ‘world’ stocks of R&D expenditure. The research into the determinants of inward FDI indicates that this is itself driven by the growth of firm-specific assets outside the host country. We allow for the existence of adjustment costs by estimating a data-based dynamic model for employment and investment in which the factor demand expressions implied by the marginal productivity conditions ((3a) and (3b)) are embedded as the long-run steady-state solution. Estimation is on a consistent, quarterly, time series data set for West Germany available up to 1994. Results are reported in Table 9.7, with the derived estimates of the production function parameters reported at the foot of the table, along with their respective t-statistics. In all cases current values of output and real factor prices were treated as endogenous, with parameter estimates being obtained using three-stage least squares (3SLS) with appropriate cross-equation restrictions imposed. Terms in the growth of EU output (denoted EUY) and lagged real wage inflation were used as additional instruments. Initial estimates are shown in equation 1. A significant effect is obtained from inward foreign direct investment in the labour demand equation. Labour-augmenting technical progress also appears to contain a significant exogenous element (given by the Time variable) worth around 1.7 per cent per annum. The elasticity of substitution is estimated to be 0.33 per cent, and the restrictions required to return to a Cobb—Douglas production function are rejected by the data. All three indicators of technical progress are correctly signed in the capital demand equation, but none is individually significant, and the hypothesis of exogenous capital-augmenting technical progress cannot be rejected using a quasi-likelihood ratio test [QLR (2)=3.51]. However the joint hypothesis of exogenous technical progress in both factor demand equations (λPAT, λYDI, κPAT and κFDI all zero) is decisively rejected [QLR(4)=27.53]. Two alternative formulations are shown in equations 2 and 3. In the first equation we impose the cross-equation restrictions required to give neutral technical progress in the long-run [QLR(3)=3.60]. The patents variable remains insignificant in both equations, but the time trend and FDI stock are now significant in the capital stock equation as well as in the labour demand one. The long-run parameter implies that a 1 per cent rise in the stock of inward FDI will raise technical progress by 0.28 per cent (standard error 0.06 per cent). In equation 3 we drop the time trend from the capital stock equation and the patents stock from the labour demand equation, while retaining the restriction of neutral effects from FDI [QLR(3)=1.08]. Both the patents and FDI stocks are now significant in the capital stock equation. The significance of the former possibly reflects the extent to

126

RAY BARRELL AND NIGEL PAIN

which process innovations may primarily affect the productivity of capital equipment. Inward foreign direct investment engenders neutral technical progress, Table 9.7 3SLS estimates of a West German production function with endogenous technical progress Equation 1

Equation 2

Equation 3

Labour demand Constant 1.0767 (3.1) 0.7149 (2.4) 1.1964 (4.0) ln (L/Q)–1 –0.3679 (5.8) –0.3557 (5.7) –0.3723 (6.4) Time/(102) –0.1044 (5.2) –0.0884 (3.6) –0.1087 (5.8) ln (w/p)–1 –0.1225 (4.1) –0.1496 (5.3) –0.1146 (4.0) Δ ln (Q)* 0.4134 (5.9) 0.4035 (5.7) 0.4210 (6.2) Δ ln (w/p)* –0.4187 (4.4) –0.4796 (5.2) –0.4002 (4.3) D751 –0.0287 (4.6) –0.0275 (4.4) –0.0287 (4.6) ln (PAT)–4 0.0038 (0.3) 0.0071 (0.6) ln (FDI)–4 –0.0599 (4.8) –0.0580 (4.7) –0.0591 (5.0) Capital stock Constant 0.0386 (2.8) 0.0179 (2.6) 0.0440 (3.1) ln (K/Q)–1 –0.0065 (2.8) –0.0031 (2.6) –0.0068 (2.8) Time/(102) –0.0005 (1.0) –0.0008 (2.7) ln (c/p)–2 –0.0022 (3.1) –0.0013 (2.3) –0.0021 (3.2) Δ ln (K)–1 0.6828 (7.8) 0.7396 (8.8) 0.6905 (7.8) Δ ln (K)–2 0.1171 (1.5) 0.0993 (1.4) 0.1233 (1.7) Δ ln (Q)* 0.0330 (6.5) 0.0340 (6.9) 0.0334 (6.9) Δ ln (c/p)* –0.0026 (1.5) –0.0018 (1.1) –0.0026 (1.5) ln (PAT)–4 –0.0005 (0.5) 0.0001 (0.5) –0.0009 (2.0) ln (FDI)–4 –0.0009 (0.9) –0.0005 (2.4) –0.0011 (2.6) Implied parameters σ 0.3331 (5.2) 0.4204 (7.2) 0.3079 (5.2) λTIME 0.0043 (20.0) 0.0043 (17.0) 0.0042 (25.0) KTIME 0.0011 (0.9) 0.0043 (17.0) — λPAT –0.0155 (0.3) –0.0343 (0.5) — KPAT 0.1238 (0.5) –0.0343 (0.5) 0.1874 (1,7) λFDI 0.2445 (4.8) 0.2811 (4.6) 0.2294 (7.6) KFDI 0.2032 (1.0) 0.2811 (4.6) 0.2294 (7.6) R2: Labour 0.772 0.779 0.767 Capital 0.963 0.962 0.963 Standard error: Labour 0.0056 0.0055 0.0057 Capital 0.0004 0.0004 0.004 Notes: An *indicates a variable that is instrumented. Additional instruments were Δln (w/p)–1, Δln (EUY, ln (EUY)–1. T-statistics reported in parentheses. Dependent variables Δln (L), Δln (K). Sample 1971Q4–1993Q4.

with a 1 per cent rise in the stock estimated to raise technical progress by 0.22 per cent. Patents affect capital-augmenting technical progress, with a 1 per cent rise in the stock raising technical progress by 0.18 per cent. Labour-augmenting technical progress remains exogenous at an estimated 1.7 per cent per annum. Two tests were undertaken in order to check the statistical adequacy of the factor demand equations. The presence of serial correlation of order φ in a simultaneous model implies that: (5) where Y and X are vectors of dependent and explanatory variables, U–φ denotes the matrix of lagged residuals from the system estimates and A, B and R are matrices of parameters. E is a covariance matrix with zero off-diagonal elements, see Godfrey (1988) for further details. The lagged residuals are included in the instrument set. A quasi-likelihood ratio test of R=0 is a

FDI AND EUROPEAN ECONOMIC GROWTH

127

valid test for the absence of serial correlation. There did not appear to be any significant first-order autocorrelation in the estimated equations. For example the test statistic for equation 3 was [QLR(4)=7.48]. We have also undertaken a limited analysis for parameter stability, using a systems analogue of the Salkever (1976) dummy variable test. Although our estimated equations are derived on West German data alone, it is obviously of interest to test whether there were any significant changes in the aftermath of German unification. In particular it might be expected that unification has affected the structure of the labour market in the former West Germany. Separate (1,0) dummies for each quarter from 1991Q1 to 1993Q4 were added to the labour demand and capital stock equations and the instrument set. The additional dummies were jointly insignificant in equation 3 [QLR(24)=18.99], suggesting that there was little evidence of parameter instability. These results provide one way of seeking to investigate the importance of technology transfer and process and product innovations on technical progress. There is evidence that inward foreign direct investment has brought significant benefits to Germany, where it now accounts for some 17 per cent of total employment within manufacturing and 7.25 per cent of aggregate whole economy employment, although it is not possible to distinguish between different types of technical progress. Overall, the results illustrate that innovations from outside Germany have helped to raise productivity within Germany. At the same time, innovations from within Germany have been transferred elsewhere through outward foreign direct investment. Related work for the UK is reported by Barrell and Pain (1997b and 1998) who focus solely on labour-augmenting technical progress. Capital-augmenting technical progress is assumed to be zero. Separate results are presented for the private sector, manufacturing sector and private sector services. There is clear evidence that FDI has a significant effect on technical progress in the private sector. On closer inspection, this appears to primarily arise from the inward investments that have occurred in the manufacturing sector. In this sector a 1 per cent rise in the FDI stock is estimated to raise technical progress by 0.26 per cent. By contrast there does not appear to be any discernible significant effect from non-manufacturing FDI in the private service sector, even though some two-thirds of inward investment in the United Kingdom has taken place in these sectors. Exogenous technical progress is put at 3 per cent per annum for both sectors. One possible explanation for these findings is that the benefits offered by inward investment are more apparent and more quickly felt in those sectors where domestic producers are at a comparative disadvantage and relatively less productive. Barrell and Pain (1997b) estimate the impact of FDI on economic growth in the UK manufacturing sector, where there was a marked acceleration of labour productivity growth from the middle of the 1980s. Labour productivity (measured in terms of employee hours) rose by 3.85 per cent per annum in the decade after 1985, compared to growth of 2.85 per cent per annum over the period from 1972 to 1985. During this period the stock of inward investment in the manufacturing sector is estimated to have risen by 78 per cent in real terms. Barrell and Pain estimate that the growth of inward investment in the decade from 1985 raised manufacturing output by 12.50 per cent, or by approximately 1.2 per cent per annum, implying that up to 30 per cent of the growth in UK manufacturing productivity since 1985 might be attributable to the impact of inward investment. These results provide some evidence that international investment may potentially affect the pace of economic growth in Europe. However it is clear that it cannot be the only factor behind the estimated growth of technical progress, or total factor productivity. It remains to be seen whether similar findings can be obtained for other countries or for a more detailed industrial model. Cross-section evidence It remains possible that aggregate results and statistics of the kind reported in Tables 4 and 7 primarily reflect a ‘batting average’ effect, rather than genuine spillovers into domestically-owned companies. Such an effect might arise from the rapid growth of inward investment along with a disproportionate representation of foreign firms in higher-productivity industries (Solomon and Ingham, 1977; Davies and Lyons, 1991). One way of testing for within-industry spillovers from foreign firms to domestic firms is to examine the direct relationship between the productivity of domestic firms and the share of output produced by foreign firms using cross-section data. If the presence of inward investors exerts a positive influence on the productivity of local firms, then it might be expected that domestic firms in those industries with a significant foreign presence will have a higher level of productivity (Farinha and Mata, 1996). We use two digit manufacturing industry-level data for the United Kingdom, France and the United States to estimate an equation of the form: (6) where DQj denotes the value added by domestic firms in industry j, FQj denotes the value added by foreign firms and Lj the number of employees of domestic firms. This relationship provides a means of summarising one aspect of the within-industry impact of foreign firms on the performance of domestic firms.7The particular functional form employed implies that a rise of 1 percentage point in the share of industry output produced by foreign firms will raise the level of labour productivity of

128

RAY BARRELL AND NIGEL PAIN

domestic firms in industry j by γ per cent. An instrumental variables estimator is required since it is possible that foreign investors will seek to enter those domestic industries with the highest level of labour productivity. We use the foreign sector share in the previous year as an instrument. The results are summarised in Table 9.8. As in the time series results discussed above there is some evidence of significant positive spillovers from the presence of foreign firms in all three economies, suggesting that inward investment can lead to a change in the behaviour of domestic companies. Similar results are reported for Portugal in Farinha and Mata (1996). However much more remains to be learnt about the channels through which such effects operate in future research. Conclusions Changing patterns of demand for products inevitably alter the pattern of locational advantage. The process of the growth and decline of countries and regions is not, however, autonomous. Comparative advantage has to be seen as path dependent. There is a growing body of research into the interrelationships between trade, investment and economic growth in Europe. Industrial policies, trade policies and other developments within regional economic groups, such as the Single Market Programme, affect patterns of advantage because they change the environment within which firms operate. Firm-specific knowledge seems to be the major factor influencing the decision to undertake FDI and hence such investments can act as an important channel for the diffusion of ideas and new innovations even between Table 9.8 Within-industry spillovers from foreign firms in manufacturing United Kingdom

United States

France

μ 3.289 (30.5) 3.707 (28.2) 5.233 (32.3) γ 0.0082 (2.2) 0.0384 (4.3) 0.0216 (3.0) Year 1994 1990 1994 No. of industries 19 17 14 2 R 0.151 0.399 0.468 Notes: T-statistics reported in parentheses. All data for two-digit manufacturing industries. UK and United States results from Holland and Pain (1998, table 15). French data for employment and output taken from Dupont and François (1997, p. 46).

developed economies. The often made distinction between investment by means of mergers and acquisitions and investment in greenfield sites is becoming increasingly outmoded. New ideas can arrive through both forms of investment. Such investments can enhance the growth process in the host economy and raise welfare in the home economy by providing an additional flow of income to an investment in knowledge. The effects on the host country are likely to be more beneficial if the FDI involves process innovations and hence helps enhance the skills and knowledge of the indigenous workforce. Thus FDI should not necessarily be seen as a zero sum game, with symmetric effects on winners and losers. However detailed econometric evidence on the macroeconomic and industrial impact of inward direct investment in Europe is limited. It is to be hoped that our evidence which suggests the presence of significant spillovers into technical progress leads to more research into these issues. Much more remains to be learned. New technologies may arrive via international trade as well as by foreign investments. Both point to a need for a greater understanding of the role played by host country institutions and workforce attributes in attracting and implementing new ideas and technologies. Such factors inevitably influence the processes affecting growth, the policies available to increase growth and the overall costs and benefits to the economy as a whole. Notes 1 We are grateful to Ciaran Driver, Bob Rowthorn, Paul Temple, Garry Young and other conference participants for helpful comments on this paper, and to the ESRC for financial support (grants number L116251012 and R022250091). 2 A similar picture is apparent in North America. The cumulated inflow of foreign direct investment in Mexico rose from $10.6 billion over 1986–1990 to $31.5 billion over 1991–1995, reflecting in part a higher level of investment as a result of the North American Free Trade Agreement. 3 Representative results for the US, Japan, the UK and Germany can be found in Barrell and Pain (1996), Barrell and Pain (1999), Pain (1997) and Hubert and Pain (1998). 4 Knowledge-based assets is used as a general term to encompass factors such as firm-specific process or product innovations as well as intangible assets such as managerial and marketing skills and reputation. 5 A majority-owned foreign affiliate is one in which US investors hold more than 50 per cent of the equity stake and hence have a clear controlling interest.

FDI AND EUROPEAN ECONOMIC GROWTH

129

6 See Barrell and Pain (1997a) for further details. 7 An extension would be to allow for spillovers across industries as well as within industries.

References Agarwal, J. (1998), ‘European integration and German FDI: implications for domestic investment and central European economies’, in Barrell and Pain (eds), (1999). Balasubramanyam, V.N., Salisu, M. and Sapsford, D. (1996), ‘Foreign direct investment and growth in EP and IS Countries’, Economic Journal, 106, 92–105. Barrell, R. and Pain, N. (1996), ‘An econometric analysis of U.S. foreign direct investment’, Review of Economics and Statistics, 78, 200–207. Barrell, R. and Pain, N. (1997a), ‘The growth of FDI in Europe’, National Institute Economic Review, 160, 63–75. Barrell, R. and Pain, N. (1997b), ‘Foreign direct investment, technological change and economic growth within Europe’, Economic Journal, 107, 1770–1786. Barrell, R. and Pain, N. (1998), ‘Real exchange rates, agglomerations and irreversibilities: macroeconomic policy and FDI in EMU’, Oxford Review of Economic Policy, 14, 3, 152–167. Barrell, R. and Pain, N. (1999), ‘Trade restraints and Japanese direct investment flows’, European Economic Review, 43, 1, 29–45. Barrell, R. and Pain, N. (eds) (1999) Investment, Innovation and the Diffusion of Technology in Europe. Cambridge: Cambridge University Press. Barry, F. and Bradley, J. (1997), ‘FDI and trade: the Irish host country experience’, Economic Journal, 107, 1798–1811. Baunerhjelm, P. and Ekholm, K. (1998), ‘Foreign activities by Swedish multinational corporations’, in Barrell and Pain (eds), (1999). Blomström, M. and Kokko, A. (1996), ‘Multinational corporations and spillovers.’ CEPR Discussion Paper 1365. Coe, D.T. and Helpman, E. (1995), ‘International R&D spillovers’, European Economic Review, 39, 859–887. Davies, S.W. and Lyons, B.R. (1991), ‘Characterising relative performance: the productivity advantage of foreign owned firms in the UK’, Oxford Economic Papers, 43, 584–595. Dupont, M. and François, J. (1997), L’implantation étrangère dans I’industrie au 1 Janvier 1994, Service des statistiques industrielles, Ministère de l’Industrie de la Poste et des Télécommunications, Paris. Eltis, W. (1996), ‘How low profitability and weak innovativeness undermines UK industrial growth’, Economic Journal, 106, 184–195. European Commission (1996), The Single Market and Tomorrow’s Europe. London: Kogan Page. Farinha, L. and Mata, J. (1996), ‘The impact of foreign direct investment in the Portuguese economy.’ Banco de Portugal working paper 16– 96. Ferner, A. (1997), ‘Country-of-origin effects and human resource management in multinational companies’, Human Resource Management Journal, 7, 19–37. Godfrey, L.G. (1988), Misspecification Tests in Econometrics: The Lagrange Multiplier Principle and Other Approaches. Cambridge: Cambridge University Press. Görg, H. (1998), ‘Analysing foreign market entry: the choice between greenfield investment and acquisitions’, Trinity College, Dublin, technical paper 98/1. Grossman, G.M. and Helpman, E. (1991), Innovation and Growth in the Global Economy. Cambridge: MIT Press. Holland, D. and Pain, N. (1998), ‘The diffusion of innovations in Central and Eastern Europe: a study of the determinants and impact of foreign direct investment’, NIESR discussion paper 137. Howenstine, N.G. and Shannon, D.P. (1996), ‘Differences in foreign-owned US manufacturing establishments by country of owner’, Survey of Current Business, 76, 3, pp. 43–60. Hubert, F. and Pain, N. (1998), ‘Innovation and the regional and industrial pattern of foreign direct investment’, in Barrell and Pain (eds), (1999). Hunya, G. (1998), ‘FDI penetration in Central European manufacturing industries: an introduction and some findings’, presented at ACE Project Workshop, Vienna, May 1998. Jost, T. (1997), ‘Direct investment and Germany as a business location’, Economic Research Group of the Deutsche Bundesbank discussion paper 2/97. Keller, A. (1989), ‘Econometrics of technical change: techniques and problems’, in P.Hackl (ed.), Statistical Analysis and Forecasting of Economic Structural Change. Hamburg: Springer-Verlag. Krugman, P. and Venables, A.J. (1995), ‘Globalisation and the inequality of nations’, Quarterly Journal of Economics, 110, 857–880. Markusen, J.R. (1995), ‘The boundaries of multinational enterprises and the theory of international trade’, Journal of Economic Perspectives, 9, 169–189. Mundell, R.A. (1957), ‘International trade and factor mobility’, American Economic Review, 47, 321–335 OECD (1995), Foreign Direct Investment, Trade and Employment. Paris: OECD. PACEC (1995), Assessment of the wider effects of foreign direct investment in manufacturing in the UK. Report by PA Cambridge Economic Consultants for Department of Trade and Industry. Pain, N. (1997), ‘Continental drift: European integration and the location of UK foreign direct investment’, The Manchester School Supplement, 65, 94–117.

130

RAY BARRELL AND NIGEL PAIN

Pain, N. and Lansbury, M. (1997), ‘Regional economic integration and foreign direct investment: the case of German investment in Europe’, National Institute Economic Review, 160, 87–99. Pain, N. and Wakelin, K. (1998a), ‘Export performance and the role of foreign direct investment’, The Manchester School Supplement, 66, 62–88. Pain, N. and Wakelin, K. (1999), ‘Foreign direct investment and export performance in Europe’, in Read, R., Thompson, S. and Milner, C. (eds.), New Horizons in International Trade and Industry. London: Macmillan Press, forthcoming. Prais, S.J. (1995), Productivity, Education and Training. Cambridge: Cambridge University Press. Romer, P. (1993), ‘Idea gaps and object gaps in economic development.’ Journal of Monetary Economics, 32, 543–573. Ruane, F. and Görg, H. (1998), ‘Irish FDI policy and investment from the EU’, in Barrell and Pain (1998c). Salkever, D.S. (1976), ‘The use of dummy variables to compute predictions, prediction errors and confidence intervals’, Journal of Econometrics, 4, 393–397. Solomon, R.F. and Ingham, K.P.D. (1977), ‘Discriminating between MNC subsidiaries and indigenous companies: a comparative analysis of the British mechanical engineering industry’, Oxford Bulletin of Economics and Statistics, 39, 127–138. UNCTAD (1996), World Investment Report 1996. Geneva: United Nations.

10 Investment, growth and unemployment Modelling the supply side of the UK economy James Nixon and Gioυanni Urga1

Introduction Unemployment has been one of Europe’s most persistent problems since the late 1970s, rising to over 11 per cent for the EC on average. Much of the emphasis has been on the labour market; for example, examining the role played by trade unions, the unemployment benefit regime or the tax incentives for effort. Now in the late 1990s many of these labour market-supply side explanations are beginning to look unconvincing in the face of continued high unemployment. In the UK for example, unions are much weaker, membership has fallen significantly, but unemployment remains high. Equally, pinning an argument on the benefit regime would suggest that equilibrium unemployment would have fallen in the 1980s. In this chapter we argue that the explanation for Europe’s persistent unemployment problem may in fact lie with inadequate aggregate demand and in particular with the depressed level of investment spending in Europe over the last ten years. Will Hutton (1995) for example writes: ‘No longer can the economy, with its ageing and inadequate stock of factories, machines and industrial infrastructure provide work for all those who want it’ (p. 1). Traditionally, the natural rate hypothesis has ruled out aggregate demand as a possible cause of persistent unemployment for what largely amount to statical reasons: demand has clearly risen for many centuries while unemployment, though sometimes rising to very high levels, has remained essentially untrended over similar periods of time. This contention is picked up later, where we argue that a distinction can be drawn between rising levels of income, associated with trend increases in productivity and changes in the demand for labour that arise from shifts in the level of capital shock. We argue that, at any point in time, the latter may represent a locus of capital–labour ratios consistent with a range of non-inflationary unemployment rates. The implication is that increases in the demand for labour, warranted by increases in the capital stock may give rise to changes in employment, that do not necessarily lead to increased wage demands by driving unemployment below the natural rate. Investment, growth and unemployment Economists have some very strong (prior) theoretical convictions about long run labour market behaviour. These can be expressed in shorthand as the idea that unemployment cannot be trended over time. In other words, as economists we believe that the Luddites were wrong; advancing technological change that enables firms to produce the same output but with fewer people, does not lead to lower and lower levels of employment for the economy as a whole. This proposition is clearly supported by the data over very long periods of time. In Figure 10.1 for example, unemployment is untrended over a very long time horizon, while on the other hand our economy’s productive potential (Figure 10.2) has persistently risen. Put another way, in mature industrial economies, this growth in output has been typically achieved with broadly stable populations. Per capita income has therefore grown enormously. This could not have been achieved without technical progress, otherwise diminishing returns would have made it impossible to maintain per capita growth for so long just by accumulating more capital per worker. Our models of the economy therefore have to allow technology to improve over time. More than this however, the neoclassical economists of the 1950s and 1960s noted that as well as being positive, per capita growth rates do not tend to diminish over time (Kaldor, 1963). If we take as a stylised empirical fact the observation that per capita growth rates approach a constant in the long run, then this implies that the economy tends towards a steady state dynamic equilibrium.2 Taken together, the requirement to have a constant growth rate in the long run and to have neutrality of unemployment to growth requires that the technical progress is biased in a very specific way. In particular it requires technical progress to be labour augmenting or Harrod neutral, for the economy as a whole, over very long periods of time.3 A similar argument applies to the idea that there may be a relationship between unemployment and the capital stock. Bean (1989) probably puts it the most succinctly:

132

JAMES NIXON AND GIOΥANNI URGA

Figure 10.1 UK unemployment rate.

Figure 10.2 UK gross domestic product and trend: constant prices (log scale).

capital intensity and productivity have been rising steadily since the Industrial Revolution. Yet the unemployment rate shows no discernible trend; consisting of a few minor wiggles at business cycle frequencies, plus the occasional major change of level. It does not require any formal statistical analysis to confirm that the unemployment rate and capital intensity cannot be related in the long run. The import of these observations about technical progress and the capital stock means that in the empirical literature suitable cross equations restrictions are usually imposed between the wage, price and employment equations such that neither technical progress or capital accumulation can affect employment determination. This is done to ensure that unemployment is independent of productivity and therefore untrended in the long run. This was recognised in the seminal model of unemployment by Layard, Nickell and Jackman in 1991. Their analysis of the supply side can be summarized in the following wage-price system: (1) (2) where (k−1) is the capital-labour ratio (their proxy for productivity) and u is the unemployment rate. Solving (1) and (2) simultaneously and setting ∆2p=0, and β3=γ3, we can derive an expression of the unemployment rate consistent with stable inflation. Thus the non-accelerating inflation rate of unemployment, or NAIRU is given by: (3) Such models place all the onus of adjustment on the labour market, requiring wages to adjust to clear the market. Any unemployment that continues to exist in equilibrium is therefore due to structural factors that influence the relative bargaining strengths of employer and worker, such as trade unions or high levels of unemployment benefit. However, as Manning (1992) points out, the imposition of the restriction β3=γ3 amounts to what he describes as ‘super-neutrality’ of the equilibrium

INVESTMENT AND UNEMPLOYMENT

133

unemployment rate with respect to productivity, such that ‘unemployment is independent of the entire time path of capital accumulation and productivity growth’ (p. 5). This restriction is therefore ‘much stronger than the original neutrality proposition, which was that in an economy experiencing steady-state growth, the unemployment rate should be constant’. The problem arises because in the steady state the capital-labour ratio will grow in line with technological progress, and hence k−1 can be used as a proxy for productivity. However, it is clear if one considers the underlying production function that technological progress is not the only source of output growth. Increases in output can also arise from increases in intensity for factor inputs for a given level of technology. These changes in factor intensity will be reflected in changes in the factor shares of output. The problem with a Cobb–Douglas specification is that factor shares are constant. A more general functional form will allow factor shares to vary. We should then be able to discriminate between increases in output that arise from technological progress and increases that arise from increased factor intensity. In this chapter we argue that by modelling the unrestricted production set we are able to separate the restrictions we require for a steady state (i.e. Harrod neutral progress) from those restrictions that imply ‘super-neutrality’. This requires modelling the economy in a more general framework than the Cobb–Douglas case. In particular, we allow the elasticity of substitution between factors to be freely determined using a four factor translog cost function. It is then an open question as to just what wages respond to. For unemployment to still be independent of capital accumulation, and therefore of investment, would require a very particular sympathetic (to use Bean’s terminology) adjustment of wages to factors that influenced the level of the capital stock, such as the cost of capital. We argue here however, that the evidence appears to be that the aggregate level wages respond to productivity regard less of the source of that productivity. This would suggest that wages are unlikely to respond in the required sympathetic way. But again, with a more general functional form we can separately identify the restrictions on the wage equation that are needed for the required sympathetic response of wages for unemployment to be independent of the determinants of the capital stock. Below we argue that the restrictions for super-neutrality are highly unlikely to be fulfilled. In this case the long run equilibrium, or natural rate will be a function of the level of capital stock (or more specifically the determinants of the relative factor inputs such as the relative factor prices).4 The remainder of this chapter examines these arguments empirically. The next section introduces a consistent set of factor demands based around a flexible production technology and considers what restriction would be required on wage behaviour to retain the classical independence of unemployment from demand side factors. The remaining sections estimate the model on UK data and present model simulations to underscore the arguments being made here. An aggregate production structure for the UK economy The long-run structural model We think of the supply side in terms of a representative, imperfectly competitive firm, operating in a small open economy with five aggregate commodities; goods (Y), capital (K), labour (L), fuels (F) and non-fuels (M).5 Fuels and non-fuels are essentially assumed to be raw materials whose price is set exogenously. We assume there is a market for labour and capital which determines their respective prices, although in so far as the cost of capital is influenced by interest rates this too is exogenous, set by an inflation targeting authority. Additionally, we assume there is disembodied exogenous technical progress (t). The imperfectly competitive firm decides its required input volume, taking factor prices as given, to produce an expected level of output, given the current state of technology. It then sets price on the basis of a markup over marginal costs, which in turn determines the real value of factor incomes. This then determines actual demand, through the demand side of the economy. For our empirical specification, we assume that the cost function can be approximated by a second-order translog cost function. The translog is a flexible functional form which can be interpreted as a second-order approximation to any arbitrary cost function (see Denny and Fuss, 1977). It has enough parameters to allow us to estimate empirically an unrestricted set of elasticities of substitution, between the different factors of production. We therefore are not constrained to restrict all of the elasticities of substitution to be unity a priori, as with the Cobb–Douglas production function. We therefore write the general equilibrium translog cost function in the form:

134

JAMES NIXON AND GIOΥANNI URGA

(4)

where pi is the ith input price, C is the equilibrium total cost, γ is output, t is a time trend. By Sheppard’s lemma, differentiating the long-run cost function with respect to each of the factor input prices generates the firm’s long-run cost minimising factor demands. If we differentiate ln C with respect to ln Pi we obtain the following system of input share equations: (5)

(6) and Xi is the quantity demanded of input I. A number of specific restrictions can then be tested for and imposed on this general model: 1 Since the shares must sum to one, the following parameter restrictions must hold. This amounts to imposing linear homogeneity in factor prices.

(7)

2 Equally we require symmetry for the translog to be viewed as a quadratic approximation to an arbitrary cost function6 (Denny and Fuss, 1977). Thus the cross-partial derivatives must be equal. This requires: (8) These restrictions will be imposed throughout the paper. 3 Additionally, the cost function will be linearly homogeneous in output if:

INVESTMENT AND UNEMPLOYMENT

135

(9) 4 The homogeneous translog cost function will be homothetic if: (10) 5 Finally, labour augmenting technical progress requires the following restrictions between the coefficients on the price of labour and the coefficients on the rate of growth of technology, υ:

(11)

A consistent model of wage determination We now move to derive a model wage determination that is consistent with the determination of productivity in the rest of our supply side. What follows is a very typical statement of the ‘right to manage’ bargaining framework, very much in the spirit of Manning (1993). This is deliberate because we wish to concentrate on the innovation of extending the model to the capital stock. Although this model of wage determination is based around a union bargaining model, we believe the model has wider scope than simply the allocation of rents between an employer and a trade union. Pencavel (1985) for example, argued that a bargaining scenario is intrinsic to any situation where firm specific skills are important. Thus the firm’s profits can be thought of as a ‘cake’ to be divided up, where the share that workers receive is a function of their bargaining power and the strength of their negotiating position. This amounts to an argument that in the long run wages must be related to some measure of underlying productivity or profitability. We assume wages (w) are therefore set to maximise a weighted average of the union’s utility and firm’s profits. The Nash bargain is then: (12) Where U is union’s utility, Π is firm’s profits and Χ represents the relative power of the union in the bargain. A typical form for the union’s utility function to take is (13) Where Vt is the value of employment and is some measure of the value of alternatives elsewhere in the economy. represents the union preferences for employment relative to wages. With this form of utility function the union therefore cares about both wage and employment levels. Substitution equation (13) into equation (12) and taking logs, we have (14) Differentiating with respect to log wages gives: (15) which after some rearrangement can be written as: (16) where εNW and εПW are the elasticity of employment and profits with respect to the wage. Thus the first term, the elasticity of employment with respect to the wage, is included because the union are explicitly concerned about employment levels. It is not immediately clear why a union should be concerned about employment in the steady state. We also have in mind the situation where individual skilled workers are able to bargain over their wage but clearly have little influence over employment. We therefore set this term to zero. Following Manning (1993), defining the value functions in a standard way, leads to a relationship for real wages of the following form: (17)

136

JAMES NIXON AND GIOΥANNI URGA

Figure 10.3 The determination of equilibrium unemployment.

Finally, by making plausible assumptions about the relationship between unemployment and the transition probabilities q and s, i.e. that job quitters have the probability of staying unemployed if γs, we can express down Manning’s (1993) structural wage equation: (18) where R is the replacement ratio R=B/W. To borrow the nomenclature in Manning (1995), our original wage setting schedule, as given by (17), can be drawn as WW in Figure 10.3. However, on the assumption that benefits are indexed to the level of wages, i.e. replacing B/W by the replacement ratio, we can eliminate wage from our wage equation to obtain the structural unemployment equation (18). Thus the WW curve is completely inelastic as represented by W′W′. This gives us the equilibrium unemployment rate as determined by the wage setting relationship alone, that is as a function of union bargaining power, the replacement ratio, the job competitiveness of the unemployed and μ. As Manning points out, εΠW will be a constant for the Cobb–Douglas case, leading to an exogenous μt depending only on union power. Thus the assumption of Cobb–Douglas in addition to the cross equation restrictions between the wage equation and the price-employment schedule lead to the imposition of super-neutrality of unemployment with respect to any of the determinants of productivity, such as the level of capital stock. But this is not the case for a more general functional form. We now therefore derive the structural determinants of μt consistent with our supply side. This in turn will lead us to four alternative but inter-related possible specifications for wage equations, relating wages to productivity, costs, profits, and individual components of the cost function, respectively. We first derive an expression for the elasticity of wages with respect to profits so that we can substitute for μt.7 The profit function for the imperfectly competitive firm is therefore: (19) where C(Pi,Y) is the unrestricted cost function. Then from the envelope condition we have: (20) The elasticity of profits with respect to the wage is therefore: (21) Which if we divide through by total costs gives: (22) where Ψ, the difference between revenue and costs as a percentage of costs, is the profit rate. Given the homogeneity of the cost function in the long run, this is equivalent to the markup on marginal costs.8 Log linearising the structural wage equation (18) and substituting for μt, we derive a wage equation from the factor share:

INVESTMENT AND UNEMPLOYMENT

137

(23) A little inspection indicates the wage equation can be expressed in one of three equivalent ways: 1 One that uses the cost to employment ratio as above in (23). 2 If we appeal to duality theory we can replace this term with productivity, i.e. Y/N. 3 A third is to use the envelope condition (20) to derive an expression for wages in terms of nominal profits per employee (where in this instance the markup term Ψ is included in profits.) Thus we see the broad equivalence between wage equations based on profits (see for example the argument made by Carruth and Oswald, 1989), and wage equations based around productivity. Each of these formulations suggests that wages respond purely to changes in productivity, arising from whatever source. This property arises by virtue of the ‘insider’ model we have adopted for wage determination. The implication is that a change in productivity that arises from increased factor intensity will still be captured by wage bargainers, other things, particularly bargaining power, being constant at the expense of employment opportunities for outsiders. In the next section we present what we believe is the compelling econometric evidence that this is exactly how wages are determined. For it to be otherwise would require very particular restriction on the wage equation that essentially distinguishes between technological progress and changes in factor prices (that produces changes in factor intensities). This is discussed in an appendix and future work will test such restrictions explicitly but a priori, we see no theoretical justification for such a complex non-linear relationship. In the absence of such sympathetic adjustments in wages the unemployment schedule W'W' will be a function of factor shares. Thus an increase in the cost of capital will shift the W'W' schedule to the right decreasing equilibrium unemployment, if labour and capital are substitutes. Unemployment is therefore a positive function of the labour share, other things being equal. In reality we feel that the sympathetic adjustment in wages is more likely to come through some of the other parameters, such as union bargaining power, that we have so far held constant. In so far as such restrictions would require institutional changes in the real world, they are likely to be very slow to come about. This suggests that high levels of equilibrium unemployment, stemming from deficient employment demand may persist for very long periods of time. To try to gauge the relevance of these remarks, we now turn to implementing our model empirically. In the interests of brevity we leave discussion of the firm’s pricing behaviour to our related papers (see Henry, Nixon and Williams, 1997; and Nixon and Williams, 1997). But essentially the models we discuss there, are imperfectly competitive pricing ones, where price is set as a markup of marginal cost; where cost in this case is the total cost, consistent with our cost function and factor demands. Estimation results The dynamic factor demands This section considers the empirical evidence for Harrod neutrality and for the relationship between wages and productivity. We estimate the dynamic cost function and factor shares jointly by non-linear least squares by a systems generalisation of the two-stage Engle–Granger technique. Thus we jointly estimate the full set of equations in level form (omitting the dynamics) and test for cointegration in each equation separately, using conventional Dickey-Fuller tests. We apply our restrictions to the model in three stages: homogeneity, homotheticity and then Harrod neutral technical progress, testing for continued cointegration at each stage. We then estimate the full dynamic model. Ideally, we would have liked to test the validity of our restrictions via conventional log likelihood ratio tests but this proved not to be possible for reasons discussed below. Table 10.1 reports cointegration tests on each equation in the system, as we successively impose the three groups of restrictions. In order to achieve cointegration we extend our model in two directions. In particular, to apply linear homogeneity with respect to output requires the addition of capacity utilisation as a variable in our system. This finding mirrors the arguments made in the real business cycle literature about the need to measure capital services accurately (see for example Burnside, Eichenbaum and Rebelo, 1995). Clearly, it is utilised factors that go into the production function so this result is hardly surprising. In principle it would be possible to adjust the data for factor volumes employed (most easily labour could be multiplied by hours, for example). However, hours data is only available for manufacturing and in any case this series has been recently discontinued. We therefore include capacity utilisation in our system as an extra regressor. This implies the addition of six extra terms to the cost function; cu, pi. cu, for I=1 to 4, and cu. t, with the appropriate crossequation restrictions between them.

138

JAMES NIXON AND GIOΥANNI URGA

The second extension stems from the observation that our measure of fuel input shows a marked drop at the start of the 1980s. This seems to reflect a fundamental asymmetry of response, possibly associated with irreversibility of investment or permanent technical change. Thus the long fall in real fuel prices over the 1980s has not resulted in a return to the same Table 10.1a Cointegration tests on unrestricted levels system TCOST SL SK SF

ADF(n)

(n)

SSR

SE

LogL

4.61 [0.028] 6.44 [0.000] 4.93 [0.011] 5.01 [0.008]

(0) (3) (3) (0)

0.01064 0.00218 0.00345 0.00065

0.00912 0.00413 0.00519 0.00225

2110

Table 10.1b Cointegration tests when linear homogeneity with respect to output is imposed TCOST SL SK SF

ADF(n)

(n)

SSR

SE

LogL

4.71 [0.021] 5.81 [0.000] 4.73 [0.019] 5.57 [0.001]

(0) (1) (1) (0)

0.01608 0.00231 0.00467 0.00064

0.0112 0.00425 0.00604 0.00224

2065

Table 10.1c Cointegration tests when homotheticity is imposed TCOST SL SK SF

ADF(n)

(n)

SSR

SE

LogL

5.52 [0.001] 6.43 [0.000] 5.12 [0.005] 4.35 [0.057]

(0) (3) (3) (0)

0.0329 0.00214 0.00338 0.00071

0.0160 0.00409 0.00514 0.00236

2033

Table 10.1d Cointegration tests when Harrod neutral technical progress is imposed ADF(n)

(n)

SSR

SE

LogL

TCOST 4.41 [0.049] (0) 0.05090 0.0214 1940 SL 4.96 [0.009] (0) 0.00327 0.00505 SK 5.03 [0.007] (3) 0.00340 0.00515 SF 5.17 [0.004] (0) 0.00071 0.00236 Notes: ADF tests of order n are reported for residuals on each equation, where n is the minium lag required to remove serial correlation from the ADF regression. Cointegration probability values are for 6 regressors and are for guidance only. Round brackets are standard errors; square brackets are probability values. Table 10.2 Levels and dynamic estimates of main coefficients Sample: 65q1−96q4

Level equation

Dynamic model

FIML:

Estimate

t-statistic

Estimate

t-statistic

A0 A1 A2 A3 A11 A12 A13 A22 A23 A33 V M B11

−3.161 0.5528 0.5417 0.0640 0.1690 −0.0587 −0.0531 0.1118 0.0092 0.0311 −0.0048

250.0 92.19 73.74 −34.5 32.21 −32.71 −36.27 59.79 −12.10 41.98 −62.90

−3.157 0.5433 0.5521 0.0613 0.1630 −0.0586 −0.0567 0.1142 0.0089 0.0314 −0.0045 0.9502 −0.2479

284.4 91.63 69.10 28.44 38.03 30.74 54.03 69.10 15.08 53.77 53.62 48.48 3.23

INVESTMENT AND UNEMPLOYMENT

Sample: 65q1−96q4

Level equation

FIML:

Estimate

139

Dynamic model t-statistic

Estimate

t-statistic

B12 0.2118 2.43 B13 −0.2433 2.50 B21 0.1461 3.61 B22 −0.4364 9.72 B23 0.1703 3.37 B31 0.3679 9.57 B32 0.2196 5.06 B33 −0.8131 16.28 Notes: Where Ai j are the product coefficients and Bi j are the adjustment coefficients where 1=labour, 2=capital, 3=fuels and 4=non-fuels.

level of fuel use for a given level of output. Rather the price hike of the 1970s appears to have produced a permanent increase in fuel efficiency. To capture this effect, we additionally include a cumulated real fuel price as well as the share of manufacturing in GDP, as two further dummies in our system. This again implies the addition of a further six terms and two more restrictions for each dummy. The results from the ADF tests on the residuals from each equation on the system, indicate that we can restrict our model to be consistent with economic theory and still maintain cointegration. Thus imposing linear homogeneity, homotheticity and Harrod neutrality improve the cointegration properties of the system without increasing the standard error of the regressions markedly. Turning to the dynamic model (discussed in the Appendix), Table 10.2 reports estimates for the coefficients of interest. We are able to estimate a model where all the optimality conditions are satisfied (i.e. that all the own price elasticities are negative and where the long run elasticities of substitution are greater than the short run elasticities). The latter condition is vital to the stability of the model and ensures that the Le Chatelier principle holds (see Urga, 1996, p. 8). In terms of Table 10.3a Long-run Allen elasticities of substitution Labour Capital Fuel Non-fuel

Labour

Capital

Fuel

Non-fuel

−0.133

0.461 −0.832

−0.479 1.902 −6.876

0.366 −2.511 3.5146 −0.332

Labour

Capital

Fuel

Non-fuel

−0.151

0.487 −1.04

−0.405 1.913 −7.347

0.397 −2.33 3.391 −0.706

Labour

Capital

Fuel

Non-fuel

−0.089 0.215 −0.183 0.379

0.075 −0.136 0.278 −0.413

−0.027 0.112 −0.396 0.173

0.041 −0.284 0.397 −0.038

Labour

Capital

Fuel

Non-fuel

−0.101 0.0237 −0.141 0.393

0.0794 −0.170 0.272 −0.385

−0.023 0.110 −0.423 0.167

0.449 −0.263 0.383 −0.079

Table 10.3b Short-run Allen elasticities of substitution Labour Capital Fuel Non-fuel

Table 10.3c Long-run price elasticities Labour Capital Fuel Non-fuel

Table 10.3d Short-run price elasticities Labour Capital Fuel Non-fuel

Note: Elasticities are evaluated at sample means.

140

JAMES NIXON AND GIOΥANNI URGA

individual coefficients, we find that the necessary concavity conditions are global and that the estimated Allen elasticities are consistent with previous studies. The elasticity of substitution between capital and labour is 0.42 which is broadly in line with a wide average of studies (see Rowthorn, 1996, for a survey). Finally, our estimate of technical progress is for growth of around 1.8 per cent per year. Pre-empting some of the comments we make below, where we simulate the model; the main coefficient of dynamic adjustment (m), is estimated to be 0.95. While significantly different from one, the closeness of this coefficient to unity means that there is only a small variation between our estimates of the short- and long-run elasticities (see Table 10.3). The system also suffers from serial correlation, a result which proved robust across a number of permutations of the specification, which meant we are unable to test our restrictions in the dynamic model. The problem seems to stem from the observation that while, in principle, the system should allow for different speeds of adjustments in individual factors; in practice, the m coefficient appears to impose a significant element of common response. In mitigation we appeal to the underlying economic theory, the strong evidence of cointegration when the restrictions are applied, and the simulation properties of the final system. Thus while we accept there is evidence of misspecification, which is addressed by Hall and Nixon, 1998, we would argue that in practice this does not detract significantly from the long run properties of the aggregate system we are attempting to justify. A cointegrating vector for wages Table 10.4: Cointegrating regression for wages with productivity LWW=

OLS 65q1 96q4

OLS 66q1 96q4

OLS 66q1 96q4

Constant (0.227) LPC (0.0177) LPROD (0.0634) LU (0.01022) LMTUR

−4.963 (0.256) 1.0875 (0.019) 0.9459 (0.077) −0.06155 (0.013)

−5.208 (0.57605) 1.0743 (0.0131) 1.0369 (0.0540) −0.0426 (0.0087) −0.13952 (0.0308)

−1.0767

(0.047) LPFIT

Johansen 66q1 96q4

Johansen 67q1 96q4

Johansen 67q1 96q4

1.0690 (imposed) 0.9484

1 (imposed) 1.1223

1 (imposed) 0.99593

1

−0.01461

0.017178

−0.0597

−0.03130

0.59555

0.37534

−0.10117 −0.01461

1.0018

0.0020

(0.0033) 2.972 [4.84] 2.906 [4.85]

DF 2.649 4.539 [4.19] [4.53] ADF(4) 3.346 3.848 [4.19] [4.53] HO r=0 39.2 26.7 37.8 HA r=1 [20.9] [27.0] [33.4] HO r<1 10.4 22.4 27.3 HA r=2 [14.1] [20.9] [27.0] Notes: a Dependent Variable is the log of wedge adjusted wage (LWW). b A unit coefficient on prices was tested for and accepted on the Johansen estimates. This is therefore imposed in the results reported here. c LPC is log consumer prices, LPROD is log productivity, LU is log unemployment rate, LMTUR is proportion of unemployed, unemployed for more than 26 weeks and LPFIT is the profit rate as defined in equation (34). d VAR length in Johansen=4 (minium required to avoid mis-specification). e Round brackets are standard errors; square brackets are probability values. Table 10.5 Dynamic wage equations based on labour productivity ΔLW= ΔLW−1

65q2 96q4 0.45537 (0.0861)

65q2 96q4 0.50940 (0.0740)

65q2 96q4 0.49836 (0.0724)

65q2 96q4 0.43489 (0.0723)

INVESTMENT AND UNEMPLOYMENT

ΔLW−4 ΔLPC ΔLPC−1 ECM term LWW−1 Constant LPC−1 LPROD−1 LPFIT−1 LU−1 LMTUR−1

0.17104 (0.0811) 0.07030 (0.1015) 0.19337 (0.1027) −0.0822 (0.0339) 1 −2.0981 (0.3480) 1.0496 (0.0664) 0.86611 (0.2890) 0.01013 (0.0217) −0.09408 (0.0497) 0.42858 (0.2995)

0.21978 (0.0811)

0.21153 (0.0703)

0.22217 (0.0676)

0.23572 (0.0931) −0.9141 (0.03268) 1 −2.1994 (0.24047) 1.0472 (0.0558) 0.95971 (0.1902)

0.25111 (0.0905) −0.07820 (0.02740) 1 −2.1787 (0.27684) 1

0.27037 (0.0931) −0.07195 (0.02371) 1 −2.0584 (0.22902) 1

1.0824 (0.1324)

1

−0.09291 (0.0432) 0.31608 (0.2119)

−0.08289 (0.0462) 0.36362 (0.24903) 0.71222 0.008574

−0.0771 (0.0474) 0.43157 (0.2266) 0.38710 0.008253

141

Chi-Sq (1) [0.399] [0.534] Regression 0.008623 0.008574 SE Serial 6.2091 [0.184] 7.4836 [0.112] 7.4836 [0.112] 5.9837 [0.200] correla’n Functional 7.1664 [0.007] 8.7037 [0.003] 8.7037 [0.003] 3.2287 [0.072] form Normality 0.83380 [0.659] 2.2247 [0.329] 2.2247 [0.329] 1.2878 [0.525] Heterosced. 4.2671 [0.039] 2.7670 [0.096] 2.7670 [0.096] 0.82332 [0.364] Note: a LW is the log of the nominal wage and LWW is the wedge adjusted wage. b Round brackets are standard errors; square brackets are probability values.

The theoretical models discussed above posit a relation between real wages, productivity and a number of variables that might affect the relative strength of each party to the wage bargain. Cointegration tests for wages from both first stage EngleGranger and a Johansen VAR are reported in Table 10.4. There is clear evidence of at least one cointegrating vector between real wages and productivity, and very possibly two. We find it useful to include the proportion of long term unemployment in the total unemployment count, as an extra regressor. A number of authors attribute this to hysteresis stemming from the depreciation of human capital associated with long term unemployment (Blanchard and Summers, 1986). The inclusion of long term unemployment is not strictly necessary to achieve cointegration but the parameter estimates from the Johansen regression that does, are much more appealing theoretically. Table 10.5 reports estimates of dynamic regressions for wages on the basis of the cointegrating vector reported above. Again the long run parameters point to a clear relationship between wages, prices and productivity with unit coefficients. The unemployment term and the error correction coefficients are small but still significant. Finally, the equation readily supports dynamic homogeneity; this last finding can be interpreted as suggesting that wages are set on the basis of expectations of inflation, rather than the price level, which seems reasonable. System simulation properties Steady state properties of the system This section reports the simulation properties of the entire supply side system, the cost function, factor demands, and wages. We note that in the long run the model will possess the following steady state properties:

142

JAMES NIXON AND GIOΥANNI URGA

Figure 10.4 Factor volumes: percentage deviation from base.

• Our estimate of v gives an exogenous rate of technical progress of 1.8 per cent per annum. • If real wages grow at this amount, demand (hence output) also grows at this amount. • Harrod neutral technical progress ensures that employment is constant while other factors of production are utilised at rate v in order to produce more output. The capital output ratio is therefore a constant, as are the factor shares. • Costs therefore grow at rate v plus some underlying rate of nominal (factor) price increase. Average costs and hence prices therefore grow at a constant rate. Inflation in the steady state is therefore a constant. This could be thought of as the rate of growth of the money supply or as the authorities’ inflation target.9 Taken together our system of equations define an equilibrium (or natural) unemployment rate, which will be a function of relative factor prices, as we have argued above. In so far as we have included long term unemployment in our wage equation this will be a function of the economy’s previous time path. Equally, the rate of unemployment at which inflation will begin to accelerate will in general be different from this structural equilibrium unemployment rate and will depend on the rate of adjustment towards that equilibrium.10 An exogenous factor price rise without any wage prices dynamics This section reports partial model simulations that illustrate the substitution response between factors. The four figures (Figure 10.4) show deviations of each factor from base for 10 per cent increases in each factor price respectively. We clearly see that fuel and employment tend to be complements in our system, which substitute for capital and non-fuels. Dynamically, the factor volumes make an initial jump and then adjust to their new optimum values. The factor demand system does settle down to a steady state equilibrium, where it takes around years five for complete adjustment. This is somewhat disappointing, in that even though the dynamic factors shares should in principle allow for different speeds of adjustment, the adjustment parameter, m, appears to impose a significant amount of common response. The speed of adjustment also appears quite rapid, especially in the context of the capital stock. This goes back to our earlier comments about the degree of common adjustment imposed by the dynamic (m) coefficient. Full simulations of an exogenous real factor price rise To close the model and derive the full system equilibrium we combine the factor demands and wages with a consistent price equation.11 Finally, we close the model with a reduced form IS-LM equation:

INVESTMENT AND UNEMPLOYMENT

143

Figure 10.5 Employment: percentage deviation from base after a 10 per cent increase in the money stock.

Figure 10.6 Unemployment rate: percentage deviation from base after a 1 per cent increase in the real cost of capital.

(24) where x is fiscal policy, m is the nominal money stock and p is the gross output price deflator. Figure 10.5 reports the effect of a 10 per cent increase in the money supply to illustrate the neutrality properties of the system. The model eventually settles down to a new equilibrium where the price level is 10 per cent higher but output and employment are unaltered. However, it takes more than the whole simulation period (60 years) for the equilibrium to be restored. Experimentally, it is easy to show that this result stems from the small coefficient on unemployment in the wage equation. When this coefficient is doubled the response time is nearly halved (see the second line on Figure 10.6) but still remains very long. This is an example of a real inertia; i.e. that the equilibrating mechanism of unemployment on real wages is very weak. In this reduced form demand side simulation, relative factor accumulation is the only equilibrating mechanism. The relatively slow adjustment to equilibrium therefore not only reflects the time taken for capital to accumulate and wages to adjust accordingly but also the absence of any extra stabilisation on the part of the government. The final simulation considers what happens if there is an increase in the real cost of capital. This is shown in Figure 10.6, which shows the proportionate change in the unemployment from base following an increase in the real interest rate. Employment rises as first as the system substitutes for the cheaper factor but in the long run employment falls since the NAIRU is a function of the real interest rate. Steady state unemployment is therefore also a function of the real interest rate. Thus this simulation is equivalent to a leftward shift of the aggregate supply curve; a new equilibrium is established but at a lower level of output and a higher level of unemployment.

144

JAMES NIXON AND GIOΥANNI URGA

Conclusion This chapter has estimated and simulated a consistent aggregate production structure for the supply side of the UK economy. We have paid particular attention to the restrictions required to ensure our system has a steady state solution and where the classical dichotomy holds. By estimating the unrestricted production set and using a flexible functional form for technology, we are able to make a distinction between those restrictions required for a steady state and those required for the neutrality of unemployment to the capital stock. We assume firms face adjustment costs in altering factor inputs, which leads us to estimate the system as a set of dynamic share equations and consistent dynamic cost function. The empirical results reported in this paper are encouraging: • We are able to estimate a fully dynamic model for the aggregate UK supply side that is theoretically consistent and has desirable economic properties. In particular the model has an identified steady state, with a constant equilibrium unemployment rate. Moreover the classical dichotomy is shown to hold, with an increase in the money stock eventually reflected in a proportionate change in the price level, while output and employment are unchanged. • The model has sensible non-unit elasticities of substitution, where the concavity requirements (for negative own price elasticities of demand) are met and is dynamically stable. When perturbed, the model therefore returns to a steady state growth path. • We find no strong evidence for increasing returns to scale once variations in factor utilisations are accounted for. With respect to capital stock neutrality, we have argued that the restrictions required for ‘super-neutrality’ will in general be non-linear combinations of the factor shares, which will themselves change over time. It is our view that the imposition of super neutrality is overly restrictive and requires a very specific response in real wages that is not consistent with the bargaining conception of real wages being a function of productivity. In the absence of super-neutrality equilibrium unemployment will depend on the factor share. In the context of a full system equilibrium, we therefore demonstrate how an increase in the real cost of capital, which leads to a reduced capital stock will increase equilibrium unemployment. We believe this to be of particular import for the current policy regime operating in Europe which we characterise as managing inflation through higher real interest rates. While this effectively curbs inflation by controlling aggregate demand, it also has an adverse supply side effect through its impact on relative factor prices and capital accumulation. This worsens the output—inflation trade-off, pulling in the production possibility frontier and leading to an increase in equilibrium unemployment. Thus the high unemployment experienced in Europe since the start of the 1980s is not in our view entirely a labour market phenomenon but fundamentally bound up with the impact of the policies that have been adopted in that time to combat inflation. Appendix 10.1: the dynamic cost function One way of getting a handle on dynamic adjustment is to assume firms seek to minimise quadratic adjustment costs, where the cost incurred may be direct such as construction or training costs, or else profit forgone incurred by producing at less than optimal scale. These adjustments costs themselves are likely to vary between factors, with capital the most expensive to adjust. This in turn is likely to be reflected in different speeds of adjustment. The most obvious way of dynamising our model is to borrow from the AIDS literature on modelling consumption. This will give us a dynamic model in factor shares. We therefore assume firms face additional costs Ci when adjusting each factor share (see Christofides, 1976). (31) where Ci (i=1, 2, 3) are conformable adjustment matrices. Minimising L* with respect to St we obtain generalised error feedback equations of the form: (32) where the dis-equilibria in n–1 factor shares at time t–1 influence the current-period adjustment and any particular share and G is in general a matrix. Anderson and Blundell (1983) show this is equivalent to a dynamic VARDL (1,1) process (33) where A+B+C=I, an identity matrix, and G=A, K=I–C=A+B. Since the sum of the changes in shares must be zero, i.e. (34) the columns of the K matrix must therefore also sum to zero. This implies that there are only n–1 independent dis-equilibrium shares; the K matrix is therefore of reduced rank (n–1)×(n–1). This means that the individual adjustment coefficients are not

INVESTMENT AND UNEMPLOYMENT

145

identified; rather, we are only able to obtain the ratios k11/k12 for example. Allen and Urga (1995) however suggest that the joint estimation with a suitably consistent dynamic cost function overcomes this problem. This is because the dynamic cost function utilises an extra constraint; namely, that the sum of the factor prices times the factor volumes has to equal total costs each period. They propose a particular form of dynamic cost function that can be estimated jointly with share equations of the form of (33) by imposing the mapping that the G matrix is a scalar m. This gives the following dynamic cost function that contains both equilibrium and disequilibrium terms and satisfies the integrability conditions between the cost function and the factor shares.

(35)

where bij are elements of the matrix B. The corresponding cost share equations will be of the form (36) where kij are elements of K and where K=mI+B. Appendix 10.2: restrictions on wages for super-neutrality To identify the sources of productivity we can substitute for costs in the wage equation using the full translog cost function. This will give us a wage equation in terms of factor prices and technology: (37) where Pi are the individual factor prices, v is the rate of technical progress and γi are coefficients on the individual factor prices. For unemployment to be independent of capital accumulation, there has to be (in Bean’s 1989 terminology) a sympathetic adjustment in wages to offset this effect. One possibility is the imposition of suitable cross equation restriction between the γ’s of our wage equation and the parameters of the cost function. However, these will themselves be non-linear functions that vary over time, again with factor shares. To demonstrate this, consider the Allen cross price elasticity of demand for labour with respect to an alternative factor price PF: (38) and the own price elasticity of labour is (39) Then for equilibrium unemployment rate to be independent of factor prices requires an offsetting adjustment in wages such that employment is constant. Equating dN in equations (38) and (39): (40) which, rearranging, gives: (41) Substituting from for the elasticities in terms of the factor share gives: (42)

146

JAMES NIXON AND GIOΥANNI URGA

where we can substitute further using the expression in Appendix 10.1 to get an expression in terms of parameters of the cost function and factor shares. This will therefore be the required restriction on the γ’s, such that (43)

Notes 1 We are grateful to Stephen Hall, Brian Henry, Andrew Sentance and Ron Smith for their comments and assistance. The authors would also like to extend their thanks to Chris Allen for his obvious contribution to our earlier work. Any errors in this version remain the responsibility of the current authors. Financial support from ESRC grant No. L116251013, ‘Macroeconomic Modelling and Policy Analysis in a Changing World’, is gratefully acknowledged. 2 Of course a cursory glance at Figure 10.2 indicates that the trend growth rate of GDP has been anything but constant. Part of this will be growth in labour supply, part of it will be catch up after a period of capital destruction such as a war (see Crafts and Toniolo, 1996, for a discussion of European growth after the Second World War). There then may still be residual changes in trend growth which are of central importance and may be explanations of unemployment in their own right. This is something we may wish to investigate in future; for the time being however we believe it is important to construct a model that has interpretable steady state properties before we consider altering the trend growth rate. 3 A proof of this proposition can be found in Barro and Sala-I-Martin (1995), pp. 54–55. 4 This is discussed in more detail in Allen and Nixon, 1997, who provide analyt ical solutions for a general supply side model 5 We chose a four factor production technology because it is important to allow for the large increases in commodity prices that have been experienced from time to time since the 1960s. In the case of oil prices for example, these have had a very significant effect on the demand for labour. 6 As an exact functional form, the translog cannot adequately represent a separable technology as a flexible second-order approximation. The set of constraints required for weak separability imposes strong restrictions on either the micro aggregation functions or the macro function (see Diewert, 1976, for a general discussion of aggregation, while Blackorby et al. discuss the restrictions). In order to avoid these restrictions, the weaker notion of a second-order approximation at a point has been adopted. It is not clear that this loss is trivial since the behaviour of the approximation away from the point of approximation will depend on the data set. Typically, this is not an issue when one is estimating point estimates of the elasticities of substitution but is more problematic when the translog is pressed into time series analysis. 7 We are grateful to Alan Manning for providing the derivations behind the relationships in his 1993 EJ paper, to which this section is obviously related. 8 Since this term will appear in the wage equation but not in the labour demand schedule, the W′W′ schedule will also be shifted by changes in the profit rate. Thus equilibrium unemployment is a function of the mark-up on marginal costs. One of the points made in our related papers on pricing (Henry, Nixon and Williams, 1997; Nixon and Williams, 1997) is that there appears to be evidence of a widening in producer margins in the 1990s, which mirrors the findings of Blanchard and Muet (1993) for France. Our argument here would imply that such an increase in profitability would lead to an increase in equilibrium unemployment. 9 Clearly, if this were different from the world rate of inflation our steady state we describe here would only be consistent with a depreciation in the nominal exchange rate. 10 We therefore make a distinction between the NAIRU and the natural rate: Layard, Nickell and Jackman (1991); Allen and Nixon (1997). 11 This is discussed in Nixon and Williams (1997).

References Allen, C.B. (1997) ‘A supply side model of the UK economy: an application of non-linear cointegration’, in Chris Allen and Stephen Hall (eds), Macroeconomic Modelling in a Changing World: Towards a Common Approach, London, Wiley. Allen, C.B. and Nixon, J. (1997) ‘Two concepts of the NAIRU’, in Chris Allen and Stephen Hall (eds), Macroeconomic Modelling in a Changing World: Towards a Common Approach, Wiley, London. Also published as London Business School, Centre for Economic Forecasting Discussion Paper DP 25–95, October 1995. Allen, C.B. and Urga, G. (1995) ‘Derivation and estimation of interrelated factor demands from a dynamic cost function’, London Business School, Centre for Economic Forecasting Discussion Paper 10–95. Alogoskoufis, G. and Manning, A. (1988) ‘On the persistence of unemployment’, Economic Policy, 2, pp. 427–469. Banerjee, A., Dolado, J., Galbraith, J.W. and Hendry, D.F. (1993) Co-integration, Error-Correction and the Econometric Analysis of Nonstationary Data, Oxford, Oxford University Press. Barro, R.J. and Sala-I-Martin, X. (1995) Economic Growth, New York, McGraw-Hill. Bean, C.R. (1989) ‘Capital shortage’, Economic Policy, 8, April, pp. 11–53. Berndt, E.R. and Wood, D.A. (1989) ‘Energy price shocks and productivity growth in US and UK manufacturing’, Oxford Review of Economic Policy, 2(3), pp. 1–26.

INVESTMENT AND UNEMPLOYMENT

147

Blackorby, C., Primont, D. and Russel, R. (1977) ‘On testing separability restrictions with flexible functional forms’, Journal of Econometrics, 5(2), pp. 195–209. Blanchard, O.J. and Muet, P.A. (1993) ‘Competitiveness through disinflation: an assessment of the French Macroeconomic Strategy’, Economic Policy, April, pp. 12–57. Blanchard, O. and Summers, L. (1986) ‘Hysteresis and the European unemployment problem’, NEBR Macroeconomics Annual, Vol. 1, pp. 15–89, Cambridge, MA, MIT Press. Brainard, W.C. and Tobin, J. (1968) ‘Econometric models: their problems and usefulness: pitfalls in financial model building’, American Economic Review, 58, pp. 99–122. Bruno, M. and Sachs, J. (1985) The Economics of Worldwide Stagflation, Cambridge, MA, Harvard University Press. Burnside, C., Eichenbaum, M. and Rebelo, S. (1995) ‘Capital utilisation and returns to scale’, NEBR Macroeconomics Annual, pp. 67–124, Cambridge, MA, MIT Press. Carruth A. and Oswald, A.J. (1989) Pay Determination and Industrial Prosperity, Oxford, Clarendon Press. Christensen, L.R., Jorgenson, D.W. and Lau, L.J. (1973) ‘Transcendental logarithmic production functions’, Review of Economics and Statistics, 55(1), pp. 28–45. Christofides, L.N. (1976) ‘Quadratic costs and multi-asset partial adjustment equations’, Applied Economics, 8, pp. 301–305. Crafts, N.R.F. and Toniolo, G. (1996) ‘Postwar growth: an overview’, in N.R.F.Crafts and G.Toniolo (eds), Economic Growth in Europe since 1945, Cambridge, Cambridge University Press. Darby, J. and Wren-Lewis, S. (1992) ‘Changing trends in international manufacturing productivity’, Scandinavian Journal of Economics, 94 (3), pp. 457–477. Darby, J. and Wren-Lewis, S. (1993) ‘Is there a cointegrating vector for wages’, Journal of Economic Studies, 20(1–2), pp. 87–115. Denny, M. and Fuss, M. (1977) ‘The use of approximation analysis to test for separability and the existence of consistent aggregates’, American Economic Review, 67 (3), pp. 404–418. Diewert, W.E. (1976) ‘Exact and superlative index numbers’, Journal of Econometrics, 4(2), pp. 115–146. Hall, S.G., and Nixon, J. (1997) ‘Controlling inflation: modelling monetary policy in the 1990s’, in Ghris Allen and Stephen Hall (eds), Macroeconomic Modelling in a Changing World: Towards a Common Approach, London, Wiley. Hall, S.G. and Nixon, J. (1999) ‘A new approach to modelling dynamic factor demands’, mimeo. Forthcoming as a London Business School, Centre for Economic Forecasting Discussion Paper. Henry, S.G.B., Nixon, J. and Williams, G.A. (1997) ‘Pricing behaviour in the UK’, London Business School, Centre for Economic Forecasting Discussion Paper, 04–97, March. Hutton, W. (1995) The State We’re In, London, Jonathan Cape. Kaldor, N. (1963) ‘Capital accumulation and economic growth’, in Friedrich Lutz and Douglas Hague (eds), Proceedings of a Conference Held by the International Economics Association, London, Macmillan. Layard R., Nickell, S. and Jackman, R. (1991) Unemployment: Macroeconomic Performance and the Labour Market, Oxford, Oxford University Press. Manning, A. (1992) ‘Wage setting, productivity growth and the equilibrium rate of unemployment’, LSE, CEP Discussion Paper, 65, March. Manning, A. (1993) ‘Wage bargaining and the Phillips curve: the identification and specification of aggregate wage equations’, The Economic Journal, 103(416), pp. 98–118. Manning, A. (1995) ‘Developments in labour market theory and their implications for macroeconomic policy’, Scottish Journal of Political Economy, 42(3), pp. 250–265. Mayes, D. and Young, G. (1994) ‘Improving the estimates of the UK capital stock’, National Institute Economic Review, 1(147), pp. 84–95. Nixon, J. and Williams, G.A. (1997) ‘Riding the roller-coaster: an analysis of the exchange rate pass through in the UK’, mimeo. Forthcoming as a London Business School, Centre for Economic Forecasting Discussion Paper. Pencavel J. (1985) ‘Wages and employment under trade unionism: microeconomic models and macroeconomic applications’, Scandinavian Journal of Economics, 87(2), pp. 197–225. Phillips, A.W. (1958) ‘The relation between unemployment and the rate of change of money wages, 1862–1957’, Economica, 34, pp. 254–281. Rowthorn, R. (1996) ‘Unemployment and capital-labour substitution’, University of Cambridge, paper presented to the CEPR/ESRC workshop on Unemployment Dynamics, Cambridge, UK. Urga, G. (1996) ‘Dynamic specification of factor demand equations: inter-fuel substitution in US industrial energy demand’, London Business School, Centre for Economic Forecasting Discussion Paper 19–96. July.

Part III The policy lessons

11 Overview A survey of key policy issues Ciaran Driver and Paul Temple

While the City’s unadventurous approach to long-term investment is partly to blame, UK plc is the prime culprit for not generating sufficiently attractive ideas. LEX, Financial Times, 21.5.1996 The UK is an excellent place for manufacturing but on the whole UK managers are downsizers. Allen Yurko, UK Chief Executive of Siebe, Financial Times, 21.5.1996 Investment returns are very high today, but industrialists seem proud to claim that they need to be higher still for them to invest more…The risk is that those who keep saying these things will start to believe them. This doesn’t improve their returns but it does keep investment down. Andrew Smithers, Chairman of fund advisers Smithers & Co., Evening Standard, 5.5.98 The British financial system is an astonishingly robust construction. In 20 years’ time, the researchers will still be uncovering its shortcomings. Will Hutton, Guardian, 20.6.1990 Introduction It is possible to construct a series of arguments denying an important role for capital investment in the formulation of economic policy. A robust defence of such a view can be based on diminishing returns to capital, the capturing of productivity gains by ‘insiders’, efficient forward markets for goods and services and an absence of positive externalities. Moreover, even if it were accepted that stimulating investment was a desirable objective, it is far from clear how policy should proceed. Which policy instruments for example are most likely to encourage a change of behaviour and induce additionality? In the UK, general disillusion in this regard with an industry policy which for many years formally targeted fixed investment, may account for some of the reticence of policy-makers to take on board the counter-arguments. In recent years they have preferred, even when paying lip-service to the desirability of investment, to concentrate on what might be called ‘hygiene’ factors. We may include under such a head, a stable macroeconomic climate, a tax regime which favours entrepreneurial endeavour and inward investment, and (a particular favourite of the new Labour administration), an emphasis on human capital formation. Can investment create jobs? The strongest case can be made for an investment policy where market imperfections cause a gap between social and private returns so that private calculation would give a biased judgement on the quantity of capital that would maximise growth. Many such potential imperfections have been discussed above. In our first overview we drew attention to the issue of liquidity constraints caused by information asymmetries, and to the role of risk in constraining investment. Private risk includes a premium for bankruptcy that should not figure in pooled or public risk (Meltzer 1989).1 In our second overview we discussed the role of endogenous growth and the role of externalities and spillovers. On the face of it, it seems strange that R&D is unquestionably accepted as carrying external benefits but the same is not held to be true of fixed investment. In both cases the gains from the activity are likely to be shared with other producers in downstream industries (and possibly upstream as well in the form of training and standards). In general the gains from investment will be split between the adopters of higher capital intensity and the user industries. Indeed, recent research suggests that the most important spillovers occur in the adoption or

150

CIARAN DRIVER AND PAUL TEMPLE

diffusion phase of R&D which generally involves fixed capital (Geroski 1991; Stoneman 1995). Where the form of capital accumulation is expansionary rather than capital deepening there is an even more certain spillover of benefit in that the identification of a new market is quintessentially something that cannot fully be appropriated but which provides valuable public information. In the remainder of this overview we discuss a number of policy issues by highlighting questions which the individual chapters address at least in part. We close with a section on practical policy proposals. One important source of a wedge between private and social rates of return would be if investment could generate jobs and reduce unemployment especially in the long term. In fact many continue to argue that there is no long-run relationship between investment and unemployment. The underlying argument rests on a stylised fact—roughly stable unemployment contrasting with big swings in investment intensity. The case here is that long-run time series evidence on the relative constancy of unemployment must imply a tendency for productivity gains to be siphoned off by the existing workforce. As Bean (1989) puts it this view ‘is little more than a statement that real wage aspirations tend to rise as the economy gets richer’. The story that is being told here is not, however, always clear. Does the process of compensating real wages occur at the micro-level through plant-level bargaining or at the macro-level through price falls?2 In the case of the latter, the effects are more complicated because there will be an expansion of real demand, and the mechanism whereby the gains accrue only to existing workers is unspecified. There is some evidence at the micro-level that firms’ profits do not increase as a result of capital deepening (Clayton and Carroll 1996). This may partly reflect the changed competitive conditions as a result of higher capital intensity and sunk costs. But it is not certain that the gains therefore go to the existing workforce. To the extent that there is a failure to appropriate profit gains it may be assumed that lower prices are spreading the productivity gains beyond the existing workforce. The extent to which that gets translated into higher demand and employment seems to us more complicated than is generally discussed.3 It hinges, for example on whether inflationary pressure is reduced by investment in capacity (Rowthorn 1996; Cosh et al. 1996; Driver and Shepherd, Chapter 12 below). It hinges also on the extent to which any induced expansionary investment contributes to solving mismatch unemployment, especially regional mismatch, which may be a significant component of the total. Finally it hinges on the split between real wage gains and reduction in standard and actual hours of work.4 Any increase in capital deepening is more likely to bring employment gains if it is accompanied by measures to contain wage pressure and encourage work sharing; and if it is accompanied by measures to encourage expansionary investment to accommodate the real income increase from lower prices. One of these key issues—the effect of capacity shortage on inflationary pressure is addressed in Ciaran Driver and David Shepherd (Chapter 12 below). They examine the behaviour of price in relation to cost in several UK manufacturing industries, allowing for the effect of separate labour and capital constraints. They find that prices are pro-cyclical and that in addition to (or as an alternative to) labour constraints, plant capacity shortage has a significant effect on inflation in some industries. Thus, capital investment may be an important contributor to price stability. This requires that monetary targeting should take into account the effect of monetary policy on capital accumulation. How effective has investment policy been? The issue of policy effectiveness in encouraging investment has been very much to the fore in the industrial policy debate in the UK. From the early post-war years until the 1980s, tax incentives for investment (and for a limited period direct investment grants) dominated the government’s assistance for industry. This pro-investment bias was removed in the middle 1980s, partly to avoid the distortionary impact of taxation and partly in the belief that the whole system failed to generate any additionality—that firms’ investments were little affected by the system of initial allowances and stock relief. Figure 11.1 (derived from Wren 1996) shows how various types of pro-investment policy measures contributed to total industrial assistance spending in the UK. Except for a brief period in which employment subsidies were of major importance investment subsidies dominated industrial assistance in its heyday. Today, at a national level, they are non-existent. Modern technology support, for example, is trivial by comparison. At their peak, in 1978, subsidies to investment (then mainly in the form of initial allowances and measured here in grant equivalent terms) totalled £4.3 billion in 1980 prices, some 10 per cent of fixed investment in that year.5 Empirical evidence relating to the impact of tax incentives has never been entirely clear cut, a theme which is taken up in Michael Sumner’s contribution (Chapter 13). He employs an aggregate approach based on a long annual time series for the UK and obtains a cointegrating relationship between gross investment, output and the ratio of manufacturing earnings index to the implicit price deflator for investment goods. Cointegration is confirmed for the pre-tax interest rate but adjusting it in a variety of ways to reflect the influence of tax incentives weakens this result either by reducing significance or by lowering the estimate of the output coefficient when the relationship is interpreted as an investment equation. The result here mirrors other results that only 52 per cent of UK manufacturing firms base their calculations on post-tax profits (CBI 1998). Estimation of the associated error correction equation (Sumner’s Table 13.4) shows two interesting results on the effectiveness of fiscal policy. First, announcement of policy appears capable of influencing the timing of investment. Second there is evidence of differences in effect of fiscal policy measures in different periods. The differences, Sumner argues, have

KEY POLICY ISSUES

151

Figure 11.1 Industrial support in the UK, 1946/7–1990/1. Source: Derived from Wren 1996, table 1. Notes: Calculations on a grant equivalent basis; data excludes regional and firm/industry assistance.

less to do with the type of policy pursued (allowances, grants, tax rates); rather, the effect simply reflects present value calculations. Importantly, however, the significant effect obtained is for an average measure over the time periods—ten years or more in each case. Neither changes in the parameters of policy nor changes in the long-interest rate used to discount the value of incentives are relevant. The conclusions are somewhat disturbing, but perhaps not altogether surprising in view of the large wedge between actual real interest rates and hurdle rates used by companies—see the discussion in Kate Barker’s chapter (14) below. Companies’ long-run reaction to fiscal regimes are such as to evaluate them at unchanged discount rates and seemingly to react to long-run average levels of incentive. Only in the case of the 1984 rise in the after tax price of capital was there a clear and marked effect on investment. Is there a problem with the way companies appraise investment proposals? These results on the way companies take account of incentives are illustrated further in the contribution by Kate Barker (Chapter 14). She notes that small changes in taxation are unlikely to affect investment much, and that large changes are risky. Barker makes three interesting points about high hurdle rates which have been widely seen as constraining investment —average hurdle rates are of the order of 20 per cent as opposed to 10 per cent for the cost of capital. First, she argues that high hurdle rates themselves are not sufficient to shelve projects if there are strategic reasons for investing—though it appears that in two-thirds of reporting firms this happens ‘seldom or never’ (CBI 1998). Second, in some cases the hurdle rate is simply varied to ration investments to a manageable number. Third, hurdle rates are similar for foreign and domestic firms in the auto sector. Taken together these arguments suggest that it is not the hurdle rates per se that are constraining investment but that they reflect constraints coming from other directions. Barker considers various constraints: skilled labour, finance, and attitude to growth. Managerial and technical skills still appear an issue, at least for those interviewed—mainly automotive and pharmaceutical firms. This echoes results from a previous study which identified in-house engineering capability, management time and labour goodwill as constraints on expansion (Morgan 1986). Whether these are truly to be regarded as constraints, or more a feature of the way companies are run is not entirely clear. Certainly in a lean company it is difficult to have sufficient slack to grow fast and this may appear as a labour constraint, rather than what it may be—a response to low growth aspirations. The question of finance links with our earlier discussion in Overview (Chapter 1). The companies interviewed expressed a sense that the financial markets looked short term and that better understanding and communication between finance and industry

152

CIARAN DRIVER AND PAUL TEMPLE

was important. Firms argued that pressure for higher dividends did not directly constrain them as the investment and dividend decisions are taken separately. However, as Barker notes the dividend decision does affect the availability of internal finance and this has featured as a constraint in the CBI survey. Certainly research suggest that companies in the UK tend to pay excessive dividends, either to fend off unwelcome takeovers (Mayer 1992; Mayer and Alexander 1990) or because the tax system has tended to favour this (Bond and Jenkinson 1996). The relationship of dividends to investment is addressed in Small (1996) who argues that the effect is unclear; although ‘constrained’ firms would be unable to undertake some profitable investment opportunities, ‘unconstrained’ firms would be free to do so instead. While there may be some truth in this, investment opportunities are often firm-specific since they require previously acquired options such as R&D to pursue them in the medium term. Whatever the strength of the effect, dividend constraint is likely to bite only for a subset of firms. Indeed many firms have been making large debt repayments and buying back shares, suggesting that the main constraint is self-imposed. This idea of a managerial-imposed internal constraint is mooted in the interviews. Some firms appear to accept a self-imposed limit on their investment levels—either a small multiple of annual depreciation or a steady pace of growth. As Barker remarks, foreignowned firms were more inclined to talk about growth as a specific target of corporate strategy. The chapters by Sumner and Barker interact with a long-running policy debate over and the effectiveness of tax vis a vis alternative policies. For some observers, high dividend pay-outs are important because of their impact on financially constrained firms and this is seen as correctable by a change in tax policy (Bond et al. 1996).6 Others have suggested more far-reaching reform of competition policy or company law (Mayer 1992; Wass 1990). We take the view here that whatever need there is to address the shorttermism debate, the problem of investment levels is wider than any problem associated with financial constraints. It is now clear from recent work at the Department of Trade and Industry that the productivity gap between the UK and Europe exists not just for a tale of under-performing small companies as was traditionally thought but applies to firms across the board. If this is the case it seems that the investment problem also applies to large firms as it does to the (generally) smaller financially constrained ones. This turns the spotlight on the firms themselves and highlights the fact that high dividend pay-outs may be the result rather than always the cause of low investment. This state of affairs might cause little worry if two conditions were satisfied. First, if companies pursued objective and rational criteria in investment decision-making. Second, if financial intermediation were reasonably effective in channelling funds back to start-up companies or those with profitable investment opportunities. Perhaps surprisingly, we doubt that either of these conditions are met. The first doubt concerns the way in which firms go about assessing investments and in communicating their plans with providers of finance. The evidence from the detailed CBI survey shows many puzzling features about the appraisal process. Discount rates differ between those using net present value and those using hurdle rates for no apparent reason. Average real rates inexplicably seem higher than average nominal rates. Short pay-back calculations seem common. Some firms are rationing on the basis of hurdle rates, though that can produce inconsistent results. The hurdle rates used are vastly higher than the cost of capital and while this might be rationalised it is unclear which explanation fits. The adjustment for risk appears difficult to explain and does not seem to be associated with firms’ betas. Dividends and capital investment are decided sequentially rather than in tandem. And finally the City firm Ernst and Young recently revealed that many company plans are evaluated by the City using three-year paybacks, and apparently without any discounting. All of this appears to indicate that companies and their financiers take a somewhat cavalier attitude to business school theory of net present value. This might indeed be a good thing because, as explained in Overview (Chapter 1), such calculations exclude option values. But there is little indication that companies, other than perhaps the oil giants, have taken option values into account.7 The result of this hiatus is a vacuum in decision-making which is then filled by a conventional or cultural response. In good times incrementalism will win out; when times are tough cost-cutting will triumph over all else and a general culture of downsizing will be hard to shake even years after it has ceased to be appropriate. Is this fanciful? An outsider’s view was provided some years ago by the US chief executive of one of Britain’s biggest engineering groups who said Britain was one of the most difficult places in the world to turn on the growth switch because of the general managerial culture.8 Many economists will regard it as axiomatic that firms are maximising profits. They will find it difficult to accept that what appears to be happening, in the UK at any rate, is a maximisation of the profit rate, rather than profit. Nevertheless, it is hard to square the observed facts with profit maximising behaviour. Modern theories of decision-making place an emphasis on systems of corporate governance and incentives available for managers. Here the story appears to be of incentives biased against growth, or at best towards growth by acquisition.9 This managerial culture is probably impervious to a few percentage points off the cost of capital. Only a determined cultural shift to growth-oriented policies is likely to produce results. This will require government departments to lead on a number of fronts. Most importantly, companies must be given an indication that changes in corporate governance which encourage risk

KEY POLICY ISSUES

153

taking expansionary investment will be regarded as a socially favourable activity. Second, there must be scope for companies in mature sectors to diversify out of these areas into more growth-centred activity. Is over-regulation the issue? An orientation to growth and innovation is often seen as going hand in hand with a light regulatory touch, whether in regard to capital flows, labour markets, or product markets. In recent years, the tide has swung somewhat against unthinking deregulation of capital flows and labour markets. Jonathan Michie [in this volume, Chapter 15] makes a case that innovation may be discouraged by overuse of temporary or casual labour. He also points to the influential role of McKinsey consultants in setting an agenda for industrial competitiveness in the UK and Europe. McKinsey tends to place considerable stress on product market competition and deregulation which is seen as one of the most important industry-level drivers of managerial decisions, including capital expenditure. However, as noted in a recent critique, the quantitative evidence for such simple relationships between competition and performance is rather thin; in general, industrial organisation theory suggests that growth is stimulated by a mixture of competition and frameworks for regulating competition (Van de Klundert and Potters 1997). Nevertheless, some aspects of the McKinsey approach may be a useful catalyst, though in a different form to that they recommend. There may well be a case for linking growth objectives and competition policy with the latter applied more stringently where growth targets seem too conservative. Does effective demand matter? The centrality of growth for investment is a theme picked up by the contribution in Chapter 16 from Ian Brinkley and Soterios Soteri of the Trades Union Congress (TUC) who draw attention to the UK’s historically low investment share and its consequences for the capital stock, especially in manufacturing industry. In the 1990s, they also find the low investment share in the UK a source of puzzlement, because of the decline in the wage share (and improvement in profitability) that they observe. The resolution of the paradox, they argue, is to be found in the UK’s macro-instability and in a corporate and financial culture favouring dividend pay-outs over investment. This of course assumes that the financial markets do have a bias against longer-term cash flows (for more on this see Chapter 5). A longer-term view could be encouraged they believe, by a new corporate governance framework, extending the role of non-executive directors, making closer involvement of institutional shareholders a legal requirement and expanding the share ownership of employees. While nobody would baulk at the contribution that greater stability would make to economic prospects in the UK, Brinkley and Soteri also suggest that there is a role for policy induced growth in stimulating the demand for capacity. One source of such an expansion in the UK is public investment, which has been in something like a free-fall in the 1990s—an eventuality not wholly explicable in terms of privatisation and new financing initiatives. It is of course clear that investment is a unique component in effective demand in that an expansion of investment contributes to both effective demand and to the capacity that can help to meet that demand. But how important is effective demand in the current context? When Keynes was developing the theory of effective demand in the 1920s and 1930s, the context was quite different to that of today. High rates of unemployment were combined with rather low rates of capacity utilisation. It was the absurdity of both these features coexisting with considerable social deprivation, that cried out for a new challenge to fiscal orthodoxy. Now, by contrast, it is a shortage of capacity—in its broadest sense, that may limit faster rates of growth, by stimulating inflationary pressure. In this alternative context, the actions of the authorities are crucial in establishing the way in which the business cycle interacts with longer-term economic trends. If inflation becomes the sole objective of macro-economic policy, and capacity utilisation one of the key indicators of incipient inflationary pressure, then the problem of how a faster rate of expansion in the medium term could ever come about is a very pertinent one. Fixed ideas, on the part of the authorities today, as to what rate of growth the economy is capable of sustaining are deeply disconcerting if faster capacity growth is to play a part in generating more employment. The conceptual apparatus of the NAIRU of course is just such an idea. The same goes for the idea of an ‘output gap’ or the difference between actual production and potential production. According to Grieve Smith (1996) for example, ‘It may be correct to think that, in the short term, keeping demand well in check will lead to a low rate of inflation: but in the longer term limiting the growth rate of capacity by limiting expectations of the growth rate of demand accentuates the danger of inflation.’ To get around the problem, Grieve Smith believes that the government must set targets for the growth of output and capacity in ways that alter firms’ perceptions of the risks involved in expanding capacity. Ultimately, it was almost certainly higher private investment which brought about full employment for three decades after the Second World War, but it would be naïve to suppose that such an expansion would have been possible without a general and credible commitment on the part of governments to maintain and enhance effective demand.

154

CIARAN DRIVER AND PAUL TEMPLE

Does the observed fall in investment—output ratios simply reflect high capital productivity? The falling investment ratios documented in our second overview (Chapter 7) can be turned on their head to argue that they merely reflect high capital (and total factor) productivity. It is sometimes argued that this justifies a lower rate of investment and savings because fewer total inputs are needed to produce the same output. It is important however to see the sources of this rise in capital productivity. It may be wrong to interpret such a rise in purely technological terms. It can, for example, reflect a rise in the capital share due to a shift in rents from workers to capital as a result of changed labour market conditions. In this case, employment and real wages could be lower due to a downward shift in labour demand (Blanchard 1997). Alternatively, a rise in capital productivity could occur from a conservative approach to expansion, keeping returns high. It would not be appropriate to interpret either of these phenomena as evidence that lower investment was socially desirable. It is also important to see the sectoral picture. The decade up to the mid-1990s was characterised by big losses in labour intensive manufacturing in Europe, as it attempted to improve productivity in competition with low-wage countries (Buigues and Jacquemin 1996). While this will have improved capital shares, there is a need for expansion in new markets to absorb the available workforce and this can hardly be achieved without increasing the share of investment in output. This shows the limitations of reading off capital requirements from an aggregate analysis of capital productivity. Only if the gains in capital productivity growth are sustainable as full employment is approached and also if the social rate of discount was met at previous levels of investment, could there be any complacency about lower investment output ratios as a consequence of higher capital productivity. The issues raised by lower investment–output ratios are explored by Jaewoo Lee in this volume (Chapter 17). He shows that the observed declines in European investment-output ratios relate to non-equipment investment and draws attention to the contrast between goods and service sectors, pointing to a number of (service) sector specific episodes to explain this contrast. To shed light on the implications of falling investment-output ratios, he explores what these experiences mean in the light of his measurement of capital productivity. Deriving an expression for equilibrium capital productivity (i.e. one that takes account of optimal factor ratios on the basis of wages and technology), he shows that measured capital productivities in Europe are converging on US levels from below. He attributes most of the change not to any alteration in the capital share or to real wage but to disembodied technological change (multi-factor productivity). Discussing the issue of whether high productivity today would justify devoting a lower level of resources to investment he distinguishes between a temporary shock to productivity and a permanent increase. Only if there is a permanent increase in productivity will lower investment necessarily be the optimal choice. In fact, the source of productivity change in Europe appears to be convergence on US levels, in which case Lee argues that, this gradual and sustained source of growth cannot explain the abrupt falls in investment-output ratios observed. More speculatively, he argues that the service sector specific nature of the changes in investment-output ratios point to causes in the product market rather than the market for labour or capital. How can investment be encouraged? To encourage investment a broad approach on several fronts is required. We have identified several. A first condition is a change in managerial culture, promoted by attention both to corporate governance, company law, and management education, so that growth becomes an acceptable and desirable objective for firms. Government measures here may include direct incentives or variation in the use of competition policy. There seems no reason not to encourage new entry if existing firms are not taking advantage of growth opportunities. Growth oriented policies do however raise the thorny issue of which sectors are appropriate for development. Questions of detailed industrial structure are not generally considered as suitable for the attention of governments or bureaucracies. A problem arises for growth prospects, however, when the existing structure is heavily oriented towards mature sectors. This leads on to the second question of finance and the ability of the financial system to recycle funds efficiently by intermediation. The City clearly believes that there is a problem here: the finance group BZW has for example argued that finance is in the wrong sectors. If this is so, one response would be to encourage diversification across sectors. Alternatively it may be desirable to change company law to encourage cross-shareholding (Mayer 1992). Growth in new sectors requires more than finance of course. Complementary factors including entrepreneurship and venture capital may also be required. But these potential constraints will be lessened to the extent that big firms are playing their part in initiating diversifying expansion. One way of prompting growth, is through public investment activity. In Overview (Chapter 7) we cited evidence for the beneficial growth effects of infrastructure spending. There may also be an argument on the demand side. The call for a demand boost requires some belief in multiple equilibria such that the economy is caught in a low-growth trap but would not necessarily return to that state if disturbed sufficiently from its present position. A game-theoretic model or something similar was developed by Scherer (1991) in relation to the US historical experience in the 1960s and 1970s. In this model the US producers were caught irretrievably off-guard by their overseas rivals so that their best reaction was a slow, low-cost,

KEY POLICY ISSUES

155

submissive product development effort. But Scherer notes that if it were possible for them to rejoin the race symmetrically on the next generation of products, accelerated schedules and higher spending might occur. It may be that the best way of encouraging companies on to a growth path would be an assured market for the next-generation products. Clearly, public policy here would be important to frame in a European context. Policy tools to encourage investment include maintaining stability and on lowering the cost of capital. On the first of these, it is important to note that stabilising one variable often tends to make another more variable and this puts a premium on knowing what kind of stability is most important for industry. If one goes by the resources devoted by business to forecasting, then it is the future real growth rate that is paramount and this is confirmed in work by Driver and Moreton (1991). Macroeconomic policy should therefore be framed with an eye to stabilising growth (see also Chapter 6). Specific measures to affect the cost of capital have been explored in recent years (Bond et al. 1996; Mayes and Young 1993). It may be that some approaches are ruled out by policymakers because of their considerable upfront budgetary cost and if that is the case it is imperative to seek forms of support that minimise dead-weight loss. One suggestion is for support to be concentrated on ex post payments to firms operating in industries where tight capacity pressure shows that investment has been validated by the market. Under this scheme, funds would only be released to companies with a superior investment rate relative to competitors. These criteria would reward risk-taking, rather than ameliorating risk and this may prove attractive for some companies (Driver 1996). Finally a major constraint on investment could occur if capacity is not maintained in the downturn and companies try to invest in a concerted fashion mid-way through the upswing. The results from Sumner’s analysis (Chapter 13) may be helpful here. Although he states that it is difficult to envisage a problem of demand management to which announcement-induced changing in timing would be a solution, there has in the past been consideration of counter-cyclical investment schemes which have tried to ensure that capital was put in place before the upturn.10 Given the evidence in Caballero et al. (1995) and elsewhere that investment is delayed due to irreversibility it seems sensible to spread some of it further back in the cycle. This not only spreads social risk but also ensures that bunching of investment orders with attendant inflationary pressure is avoided in the upswing. Firms would, of course, have to be compensated for the risk of lock-in to markets and technologies. Much of what we have said in these closing paragraphs has been speculative and intentionally so. Too often the problem of low investment has been highlighted, commented on and passed over. To make a difference it is not sufficient to identify a problem but to allocate the research resources to deal with it. While the increased academic interest in these matters highlighted at the outset of this book is welcome, more focus is needed at government level to guide policy on enterprise and business initiative. In particular, policy needs to get past first base in simply representing itself as ‘pro-business’ because that can connote either growth or complacency. The aim is surely to constitute business as an entrepreneurial partner in a compact for growth and jobs. Without this focused policy, neither a new Europe nor a new millennium can be expected to result in investment, growth and employment. Notes 1 For a review of the arguments on social and private cost of capital see Flemming and Mayer (1997). 2 If the capture of productivity gains by insiders is a problem there would seem to be a case for industry-wide wage bargaining (Sargent 1995 and see Overview (Chapter 7)). 3 It is difficult to model the effects of investment in isolation. See Young (1992) for a simulation of an investment boost using the National Institute model. 4 Where productivity gains are associated with greater work intensity we may expect workers to take at least some of the gains in lower work hours. In general, the effect on unemployment will depend on changes in employment and in the participation rate. 5 Of course the true cost of accelerated allowances is simply the forgone interest. 6 These authors are more confident about the elasticity of investment with respect to the cost of capital than Sumner’s contribution here. They cite previous work which suggests a long run elasticity of investment to the cost of capital of 0.5. 7 Robert Malpas, former managing director of BP cites a decision to build a computer control room for a refinery which managers had opposed using conventional NPV analysis but later justified on the grounds it would open up future product improvements (Financial Times 30.9.1991). 8 ‘Managers fail to find the growth switch’, Financial Times 17.7.1996. The company concerned was Siebe. 9 Alan Clements, former finance director of ICI suggested that higher hurdle rates might be due to share options and bonus systems (‘Put in a spin by financial advice’, Financial Times 24.7.1995). 10 The Accelerated Project Scheme of 1976 in the UK is an example.

References C.R.Bean (1989) ‘Capital shortage and permanent unemployment’, Economic Policy, April, 11–54.

156

CIARAN DRIVER AND PAUL TEMPLE

O.Blanchard (1997) ‘The medium run’, Brookings Papers on Economic Activity, 2, 89–158. S.Bond and T.Jenkinson (1996) ‘The assessment: investment performance and policy’, Oxford Review of Economic Policy, 12, 2, 1–29. S.Bond, M.P.Devereux and M.J.Gammie (1996) ‘Tax reform to promote investment’, Oxford Review of Economic Policy, 12, 2, 109–17. P.-A.Buigues and A.Jacquemin (1998) ‘Structural interdependence between the EU and the US’, in G Boyd (ed.) The Struggle for World Markets, London, Edward Elgar. R.J.Caballero, E.M.Engel and J.C.Haltiwanger (1995) ‘Plant-level adjustment and aggregate investment dynamics’, Brookings Papers on Economic Activity, 2, 1–54. T.Clayton and C.Carroll (1996) ‘Building business: evidence from Europe and North America, Panorama of EU Industry, Brussels. Confederation of British Industry (1998) ‘Target practice: how companies approach their key capital investment decisions’, Confederation of British Industry and Association of Consulting Actuaries, London. A.Cosh et al. (1996) in J.Michie and J.Grieve Smith (1996) Creating Industrial Capacity, Oxford, Oxford University Press, 52–74. C.Driver (1996) ‘Tightening the reins: the capacity stance of UK manufacturing firms 1976–95’, in J.Michie and J.Grieve Smith (1996), 75–92. C.Driver (2000) ‘Capacity utilisation and excess capacity’, Review of Industrial Organisation, forthcoming. C.Driver and D.Moreton (1991) ‘The influence of uncertainty on UK manufacturing investment’, Economic Journal, 101, 1452–9. J.Flemming and C.Mayer (1997) ‘The assessment: public-sector investment’, Oxford Review of Economic Policy, 13, 4, 1–11. P.Geroski (1991) ‘Innovation and the sectoral sources of UK productivity growth’, Economic Journal, 98, 375–90. J.Grieve Smith (1996) ‘Rebuilding industrial capacity’, in J.Michie and J.Grieve Smith (eds), 7–23. M.Kitson and J.Michie (1996) ‘Britain’s industrial performance since 1960: underinvestment and relative decline, The Economic Journal, 106, 434, 196– 212. R.Layard, S.J.Nickell and R.Jackman (1991) Unemployment, Macroeconomic Performance and the Labour Market, Oxford, Oxford University Press. C.Mayer (1992) ‘The financing of technology’, in I. Yates (ed.) Innovation, Investment and Survival, London, Royal Academy of Engineering. C.Mayer and I.Alexander (1990) ‘Banks and securities markets: corporate financing in Germany and the UK’, CEPR Discussion Paper 443. D.G.Mayes and G.Young (1993) ‘Industrial investment and economic policy’, Discussion Paper 56, London, National Institute of Economic and Social Research. A.H.Meltzer (1989) Keynes’s Monetary Theory, Cambridge, Cambridge University Press. J.Michie and J.Grieve Smith (eds) (1996) Creating Industrial Capacity, Oxford: Oxford University Press. E.J.Morgan (1986) Corporate Taxation and Investment, Aldershot, Gower. OXREP (1996) Special Issue on Investment, May. R.E.Rowthorn (1996) ‘Unemployment, wage bargaining and capital-labour substitution’, mimeo, University of Cambridge. J.R.Sargent (1995) ‘Roads to full employment’, National Institute Economic and Social Review, February, 74–89. F.M.Scherer (1991) ‘International R&D races: theory and evidence’, in L.G. Matteson and B.Stymne (eds) Corporate and Industry Strategies for Europe, Elsevier Science Publishers BV. I.Small (1996) Box in the article on dividends and investment, Bank of England Inflation Report, August. P.Stoneman (ed.) (1995) Handbook of the Economics of Innovation and Technological Change, Oxford, Blackwell. Th.Van de Klundert and J.Potters (1997) ‘McKinsey on capital productivity’, De Economist, 145, 1, 101–9. W.Vickery (1993) ‘Today’s tasks for economists’, American Economic Review, 83, 1, 1–11. Douglas Wass (1990) ‘Innovation and Industrial Strength’, PSI Journal, Summer. C.Wren (1996) ‘Grant equivalent expenditures on industrial subsidies in the postwar United Kingdom’, Oxford Bulktin of Economics and Statistics, 317–53. G.Young (1992) ‘Low investment on the supply side of the economy’, in I.Yates (ed.) Innovation, Investment and Survival, London, Royal Academy of Engineering.

12 Supply constraints and inflation Ciaran Driver and David Shepherd1

The behaviour of the price-cost relationship is of interest both to industrial economists and macroeconomists. Industrial economists have traditionally been concerned with the question of how the relationship between price and cost varies across industries and what these variations suggest about the impact of market structure on pricing behaviour. For macroeconomists, changes in price relative to cost, particularly over the cycle, convey useful information about the dynamics of wage and price adjustment during the inflation process. In this chapter we consider the problem mainly from the macroeconomic perspective, but using disaggregated data and allowing for microeconomic considerations. What we are concerned with in particular is the extent to which supply constraints, both labour and capital, affect the relationship between price and cost movements over time. To our knowledge, this is the first systematic attempt to test for the influences of different supply constraints on industrial pricing in the UK, though the subject has been addressed in general terms in Bean and Gavosto (1990).2 Supply constraints are often omitted from economic models because they tend to make analysis intractable. Nevertheless, we show that they are important determinants of price at industry level.3 Price theory Microeconomic theory suggests that the firm’s optimal price depends on average cost (or marginal cost) and the degree of market competition. A restrictive specification for the price of the ith firm producing homogenous goods under conditions of certain demand P(Q) is: (1) where c is average variable cost and λ is the conjectural variation term (which is zero for a Cournot solution). For n identical firms, the above equation may also be written as: (2) where ε is the market price elasticity of demand.4 Taking logs of (2) we obtain

Using a Taylor approximation (3) Expression 3 shows that price depends on cost conditions, represented by the first two terms on the right-hand side, and by competitive conditions, represented by the third term. The first term simply reflects average cost. The second term captures changes in average cost with respect to output; later in the chapter we distinguish between a traditional interpretation of this term and an interpretation based on supply constraints. In some specifications in the literature, the second term is omitted on the assumption of constant marginal cost. However, at peak utilisation rates, marginal and average cost are likely to rise as older plants are brought into operation (or as existing plant is operated at an increasingly less-efficient rate). In empirical work it is the last term in (3) that gives rise to most difficulty, because the conjectural variation term is generally not observable and because of the well-known problems in measuring concentration ratios.5 The literature contains several models which suggest that competition becomes more intense in periods of expansion, with the possible implication of anti-cyclical mark-ups. This may be because collusion is easier in a slump, because of lower costs of defection from collusive agreements in a boom, thin market effects in recession, greater product variety in expansion, or pro-cyclical probability of new entry (Cowling 1980; Weitzman 1982; Rotemberg and Saloner 1986). There is, however, no general agreement that competition, or lack of collusion, is pro-cyclical. Some authors have argued that price and mark-ups should be

158

SUPPLY CONSTRAINTS AND INFLATION

pro-cyclical because capacity constrained competition will tend to be Cournot, while excess capacity will lead to Bertrand pricing (Haskel et al. 1995).6 Empirical evidence on price The multitude of competing theories can be brought into sharper focus by studying the results of firm-based surveys that indicate which theories are most relevant in practice. Major studies have been carried out in the UK (Hall et al. 1996) and in the United States (Blinder 1991). One of the main results common to these studies is the importance of cost-based pricing, suggesting that responses to demand are muted. However, support for constant marginal cost pricing is markedly lower than for general cost-based pricing, suggesting that there may be a pro-cyclical element in price. Two other important factors mentioned in these surveys were long-term contracts and fear of price wars, both of which are likely to induce sluggish price behaviour. Interestingly, there was little support for theories of pro-cyclical elasticity.7 Econometric evidence concerning the cyclical behaviour of price is mixed. Some studies suggest that price varies in a procyclical manner. In other studies the evidence points to a mixed response, with contra-cyclical pricing at least for some industries or price behaviour that shows no strong cyclical pattern (Dreze and Bean 1990; Geroski 1992; Chan, Savage and Whittaker 1995).8 Previous work on the role of specific cyclical constraints has also yielded somewhat equivocal results. Bean and Gavosto (1990) explicitly discount the role of capital constraints on the grounds that the recorded incidence is lower than for demand or skilled-labour constraints. However, this argument does not allow for the different relative impacts of these constraints on pricing behaviour. Sentance and Emerson (1995) test the role of the plant capacity constraint for total manufacturing pricing and find a procyclical effect, though not at conventional significance levels.9 Given the importance of the issue of capacity constraints, and the mixed results of previous work, it seems desirable to look more closely at how different cyclical indicators affect the behaviour of the price–cost relationship. The distinctive features of our contribution are: first, we examine the behaviour of the relationship at a disaggregated level; second, we use a consistent data set in which all of the main variables included in the estimation are derived from the same source, from a survey of business intentions; third, we allow for the possibility that both capital and labour constraints may affect the relationship. Anticipating our results, we find that price moves in a pro-cyclical manner and that both labour constraints and capital constraints are important (independent) determinants of the price-cost relationship in different industries. In so far as we have been able to allow for the impact of conjectural variations and concentration, the results continue to support a significant role for a cyclical constraint term. The model The price equation we use for estimation purposes is straightforward and follows directly from (3).10 However, neither the second nor the third term on the right-hand side of (3) is easy to measure empirically. Writing t=cq, where t is total variable cost, the second term may be written as: where r is the ratio of marginal cost to average variable cost. Following standard procedure, we express t as a cubic cost function in q with parameter γi, i=0, 3, yielding: giving, for the second term in (3)11 where γ1, γ3>0 and γ2<0. This may be approximated by a quadratic expression in q with parameters фi, i=0, 2 with Given that the cost equations used above are short-term, they are conditional on fixed capacity and changes in q will therefore correspond to changes in capacity utilisation CU. As long as there is no rationing (i.e. absolute supply constraints) the coefficient on CU2 should thus be negative, reflecting the ф2 term above. Where rationing occurs, however, marginal cost will rise sharply with the implication that the ratio of marginal cost to average variable cost also rises steeply. Since capacity constraints are clearly more likely to be encountered at large values of capacity utilisation, the second term in (3) needs to contain, in addition to a quadratic expression in CU, a term reflecting the impact of the rationing constraint. We write this term as:

CIARAN DRIVER AND DAVID SHEPHERD

159

where CU* is some threshold level of capacity utilisation beyond which rationing occurs. Since CU* is not observed and since aggregation will smooth the threshold step change, we approximate the term D(CU) by a positive squared term in CU to indicate the steep rise in the ratio of marginal cost to average variable cost when rationing occurs. The effect of this modification is that the second term in (3) now contains as before a quadratic expression in CU, but the difference is that the sign of ф2 is now ambiguous. The third term in (3) represents the effect of competitive pressure on price. The components of this term are at best observed only imperfectly (for example, there is no agreement on the determinants of the conjectural variation term). As the assumptions needed to construct proxy variables for this term are of a different order than those required to construct measures of the production constraint, we initially exclude it but return to consider it later. In summary, our initial formulation following from (3) is that firms in each sector set price (p) (4) where 1>0 and 2>0 only if capacity constraints are not binding. While it would be desirable to estimate equation (4) directly, to get some idea of the long-run equilibrium relationship between price and cost, the nature of our data (discussed later) allows us to consider only the first difference form of the relationship: (5) Preliminary data considerations Our data source is the Industrial Trends Survey of the Confederation of British Industries (CBI) briefly described in Appendix 12.1. Disaggregated data are used because previous studies suggest considerable cross-industry variation in price setting (Hall et al. 1996). We utilise the responses to four questions, which are detailed in Appendix 12.1. Two of these questions relate to the behaviour of prices and costs. More specifically, the response to question 12 gives the percentage of firms who have raised domestic price over the previous four months. The response to question 11 gives the percentage for whom average costs have increased over the previous four months. Average cost in this survey is interpreted as average variable cost, specifically including materials, labour and energy in order of importance (Junankar 1989). A general indicator of capacity utilisation is given by the response to question 4, which records the percentage of firms who were not working below a satisfactory utilisation rate. Finally, the responses to question 14 give the percentages for whom output over the next four months was expected to be constrained by skilled labour or plant capacity shortages, respectively. The use of these variables will be described later. The data on capacity utilisation, labour shortage and capital shortage are thus in the form of cumulative percentages of firms who believed they were operating beyond some satisfactory utilisation rate or at some constrained level. In practice, it is quite possible that the time series of these cumulative expectations for each industry may be non-linearly related to timemovements in the industry mean utilisation rate or the mean level of labour and capital shortages. To proxy the form of this non-linear relationship it is therefore necessary to make some assumption about the distribution across firms of the underlying utilisation or shortage variables. If the distribution of capacity utilisation is approximately normal, a logit transformation of the data results in a linear proxy for time series movement in the utilisation and shortage variables (Minford et al. 1988).12 Similar issues arise with respect to our price and cost data. The data set is a time series of the percentages of firms responding that they had increased, kept unchanged, or reduced prices (and costs). As with any study that uses qualitative survey data on price or cost movements, we need to establish that there is a well-defined relationship between the qualitative responses (in this case the percentage responding that they had raised or lowered price) and the actual rate of price or cost change in the sector.13 Although the CBI survey does not record the magnitude of the price and cost changes, we would still expect to see a close relationship between the overall direction of the survey responses and the actual rate of price or cost inflation. The basic idea is that a higher rate of actual price (or cost) inflation is likely to be associated with a higher percentage of firms responding that they had increased prices (or costs) rather than reduced or left them unchanged. For our sample, the contemporaneous correlation coefficient between the two price variables (the percentage of firms responding ‘up’ and the actual rate of price inflation in the manufacturing sector) is 0.79, indicating that the correlation is indeed significant.14 Viewed on its own, however, the correlation coefficient does not provide sufficient information about the form of the relationship and a more formal analysis of the connection between the actual and survey data is required. This is contained in Appendix 2, where it is argued that data on the percentage responding ‘up’ to the question recording perceptions of price and cost movements are adequate proxies for the percentage change in price and cost.

160

SUPPLY CONSTRAINTS AND INFLATION

In summary, our basic data set consists of quarterly data for five variables covering the period 1976Q4 to 1997Q1. From questions 12 and 11 of the CBI survey we have the two untransformed ‘up’ variables on price and unit cost which we use as our indicators of price and cost movements, termed ΔP and ΔC respectively. The responses to questions 4 and 14, measured in logit form, provide three variables that indicate the degree of capacity utilisation (CU), the strength of the labour constraint (L) and the strength of the capital constraint (K). Because the price and cost variables are expressed in changes, for estimation purposes we shall be using the first differences of the capacity, labour and capital constraints (ΔCU, ΔL and ΔK). Following this approach, we obtain measures of the required variables for each of the nine main industrial groups designated in the CBI survey. Estimation and results The relationship we wish to estimate is of the form described by equation (5). The competition and collusion factors are omitted until later and the key question we are initially concerned with is the extent to which adjustments of price in response to cost movements are affected by cyclical variations in output relative to capacity. Our modelling strategy is as follows. First we estimate (5) with just the linear term in ΔCU. We then test for the addition of a quadratic term and note its sign. If the sign is not significantly negative, indicating the possible presence of capacity constraints we go on to estimate the effect of these constraints using survey-based measures of capital and labour constraints. A general lag structure is included to capture any dynamics in the adjustment process, arising from factors such as adjustment costs, and the presence of contracts or normal cost pricing (Geroski 1992; Hall et al. 1996). The equations we estimate are of the form: (6) where a variable addition test is to be performed on and (7) where L is the lag operator, with i=1 to 4 for the P and C terms and i=0 to 4 for the CU, L and K terms; et is an error term, which in view of the derivation of (5) may be of moving average form. Our approach in (7) is in contrast to the traditional Phillips curve analysis where it is usual to emphasise labour rather than capital as the key constraint on productive potential.15 Instead, we explicitly allow for the possibility that productive potential may be constrained by independent capital and labour constraints, and that changes in the price—cost relationship may occur when either or both of these constraints become operational. In effect, we assume that the opportunities for factor substitution are limited, particularly over the shorter run, and that there may be times when a firm responds to an increase in demand by raising prices relative to costs (with unchanged output) because an increase in capital utilisation is infeasible, even if there is no general shortage of labour. Similar reasoning of course applies when labour is scarce and capital is not. In an inflationary environment, variable costs may be rising because of underlying growth in money wages or material costs and the additional increase in price is then an indicator of the inflationary impulse arising from the capital shortage (or the labour shortage, if that is the binding constraint). Before reporting on the estimation, we need to check that the variables to be estimated are all stationary. Table 12.1 shows the results of stationarity tests on each of the five variables used in equations (6) and (7). The last three columns of Table 12.1 show that the first difference of the CU, L and K terms are clearly stationary. The ΔP and ΔC terms are both trended and stationarity tests here for the individual industries again indicate stationary series for most industries, using DF and ADF(4) tests. As it seems natural that the individual series will be integrated of the same order for all industries we employed a panel test with test statistics based on the average ADF across the panel (Im, Pesaran and Shin 1995). Our test statistics, which are distributed standard Normal under the null, are −5.85 for ΔP and −3.70 for ΔC. These panel tests indicate that both variables are stationary. Table 12.1 Stationarity tests: sample 1976Q4–1997Q1

TEX DF ADF4 CHEM DF ADF4 FDT DF ADF4 EIE DF

ΔP trended

ΔC trended

ΔCU no trend

ΔK no trend

ΔL no trend

4.45 4.16 4.51 3.99 6.70 3.95 4.91

3.53 3.37 3.71 3.6 4.28 2.80 3.28

10.42 4.53 11.65 3.54 13.08 5.32 11.45

11.57 4.06 11.32 4.23 13.94 4.98 13.25

9.27 4.55 11.30 4.09 12.69 4.63 14.45

CIARAN DRIVER AND DAVID SHEPHERD

ΔP trended

ΔC trended

ΔCU no trend

ΔK no trend

161

ΔL no trend

ADF4 2.27 1.94 3.64 6.59 3.84 ME DF 4.42 3.80 10.30 11.24 8.33 ADF4 4.40 3.43 4.52 4.31 4.34 MPRO DF 4.92 3.49 12.56 13.82 14.21 ADF4 3.91 3.02 3.82 4.69 3.85 MMAN DF 5.18 7.20 13.86 10.15 12.34 ADF4 3.44 2.86 5.08 4.60 3.20 MV DF 5.76 5.53 10.54 14.56 12.93 ADF4 2.91 2.75 5.11 4.73 3.93 OMDF 4.27 3.55 7.49 10.57 12.80 ADF4 4.31 4.18 3.60 3.32 3.19 Notes: Negative signs omitted. The 95 per cent critical values are approximately 2.9 without trend and 3.5 with trend. Mnemonics for the industry groups are given below: CHEM chemicals; EIE electrical and instrument engineering; FDT food, drink and tobacco; ME mechanical engineering; MMAN metal manufacture; MPRO metal products; MV motor vehicles and aerospace; OM other manufacturing (including plastics, paper and building products); TEX textiles.

While the lagged terms for price and cost in equations (6) and (7) may be regarded as independent of the current error term, this is not necessarily the case with the current cost term, and it may need to be instrumented. We report both OLS and Instrumental Variable (IV) results, where the instruments in the IV case are the one and two period lags of the expectations of the current cost variable, reported separately for the firms in the CBI survey.16 The validity of these instruments is checked using Sargan’s test for independence from the residuals of the regression. An additional concern we need to address is that the capacity and constraint terms may not be exogenous, possibly because of open economy effects. For example, if the domestic price in the manufacturing sector was largely determined by the world price (because of the tradeable nature of many of the goods) it is possible to imagine a situation in which a rise in the world price (caused by an increase in world demand) might induce both an increase in the domestic price and an increase in output, resulting in a higher degree of capacity utilisation. In this case, the direction of causation could be the reverse of the one hypothesised in this chapter. Alternatively, a rise in the world price might lead domestic firms to increase price, but to a smaller extent than the world price, and an increase in output might then occur because of the resulting improvement in competitiveness. Again this could imply a reversal of the direction of causation. We checked for the direction of causation using the Granger causality tests reported in Table 12.2. The results suggest no evidence of the direction of causality running from ΔP to ΔCU and for the majority of industries the results point in the opposite direction. The table reports F-tests on excluding a set of four lags on ΔCU(ΔP) in a regression of ΔP(ΔCU) which also includes four lags on the dependent variable. The results indicate that for all industries we can’t reject the hypothesis that ΔP does not Granger-cause ΔCU, and in the majority of cases we can reject the hypothesis that ΔCU does not Granger-cause ΔP. Turning to the estimates of equations (6) and (7), unrestricted estimation suggested that four lags on ΔC and ΔP were sufficient. The contemporaneous capacity utilisation and constraint terms were also entered along with four lags. For convenience of exposition, we present the sum of the lagged terms in each case along with the Wald exclusion test significance level. However, we report separately the contemporaneous cost term. The results for equation (6) are shown in Table 12.3 for IV estimation and the OLS results are given in Appendix 12.3.17 In general, the OLS results are robust to instrumentation and although the t-statistic on the current cost term falls, as expected, this term remains generally significant. The long-run coefficient is relatively stable, typically changing only by about 10 per cent between the OLS and IV estimates. The Sargan test for the independence of the instruments from the error term are satisfactory in all Table 12.2 Granger causality tests on ΔP and ΔC Industry

Significance level for exclusion of ΔCU terms

Significance level for exdusion of ΔP terms

TEX CHEM FDT EIE ME MPRO

0.00 0.11 0.57 0.00 0.00 0.07

0.71 0.87 0.70 0.18 0.85 0.65

162

SUPPLY CONSTRAINTS AND INFLATION

Industry

Significance level for exclusion of ΔCU terms

Significance level for exdusion of ΔP terms

MMAN MV OM

0.23 0.51 0.00

0.99 0.40 0.40

cases at the 5 per cent level and are not reported further. The other diagnostics are mostly acceptable, bearing in mind that the results are given for nine industries. There is little evidence of autocorrelation with only ME and TEX failing the LM(4) test at the 5 per cent level and none at the 1 per cent level. The functional form test is only failed (marginally) for MPRO. Nonnormality of the residuals is only a problem in the cases of EIE and MPRO and inspection of the residual histogram does not reveal this to be a major problem in either case. There is no evidence of heteroscedasticity, except for the MV case, which suffers from a shorter sample than the other industries. Parameter stability tests, saving 10 observations for out of sample prediction, were satisfactory. The long-run cost coefficient may be inferred from the coefficients on ΔC, ΣΔC(–i) and ΣΔP(–i). For example, for TEX, we find that the long-run coefficient is about 0.7. In most other industries, this figure is closer to unity. Our main interest is in the coefficients on the capacity utilisation terms. These are always positive and significant in 6 of the 9 industries. The summed coefficients in these 6 cases range from 0.20 to 0.79. Since the capacity utilisation term is the first difference of a logit, and the dependent variable is in first difference form, the coefficients can be regarded as loose approximations to the elasticities of price with respect to the index of capacity utilisation. The variable addition test (A-test) reported in the final column for each industry is a X2(5) test for the significance of the current and 4 lagged terms in This is reported along with the sign of the summed coefficients. In 6 out of the 9 industries, these terms are not significant, even at the 10 per cent level. Six of the 9 terms are, however sum-positive and of these, 2 (TEX and ME) are significantly sum-positive. The only significantly negative case is CHEM.18 We conclude from this that there is an a priori case for considering the possibility of production constraints on price as outlined earlier. Table 12.4 replicates the estimation procedure, except that in place of the general utilisation term we have included separate terms for labour constraints and capital constraints.19 The reasoning here is that, while these constraints are somewhat collinear at industry level with the capacity utilisation term, they offer a better proxy for the second term in (3) because they capture perceptions of supply constraints which are not included in the traditional cubic cost function representation. In general the two supply constraints (skilled labour and capital) are not themselves highly collinear so that it is possible to estimate the effects separately, as reported in Table 12.4. The diagnostics are again generally acceptable. The results show that the cost and lagged price coefficients are reasonably robust to the change in specification. There is a significant effect on price at the 5 per cent level, stemming from labour constraints, in CHEM, EIE, and OM. Capital constraints are significant in the metals industries (MPRO and MMAN) at 5 per cent and in ME at 10 per cent. For TEX, it appears that collinearity Table 12.3 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1, IV(2SLS) estimation Variable

Coefficient*

Diagnostics

TEX ΔC ΣΔC(−i) ΣΔP(−i) ΣΔCU(−i)

CONST 0.76 (5.53) −0.53 [0.00] 0.68 [0.00] 0.33 [0.00]

R2 0.80

CHEM ΔC ΣΔC(−i) ΣΔP(−i) ΣΔCU(−i)

CONST 0.96 (4.35) −0.44 [0.28] 0.32 [0.03] 0.14 [0.23]

FDT ΔC ΣΔC(−i) ΣΔP(−i) ΣΔCU(−i)

CONST 0.73 (2.82) −0.09 [0.57] 0.10 [0.61] 0.03 [0.61]

−0.98 (0.35) LM(4) [0.04] FF [0.18] N [0.78] H [0.94] A (+) [0.005] 0.83 (0.27) LM(4) [0.55] FF [0.49] N [0.91] H [0.98] A (−) [0.004] 3.72 (0.98) LM(4) [0.71] FF [0.63] N [0.51] H [0.16]

R2 0.74

R2 0.70

CIARAN DRIVER AND DAVID SHEPHERD

Variable

Coefficient*

163

Diagnostics

A (+)[0.710] −2.66 (1.41) R2 0.85 LM(4) [0.47] FF [0.26] N [0.03] H [0.82] A (−)[0.471] ME CONST −0.13 (0.06) R2 0.83 ΔC 0.83 (5.03) LM(4) [0.01] ΣΔC(−i) −0.53 [0.01] FF [0.22] ΣΔP(−i) 0.56 [0.00] N [0.97] ΣΔCU(−i) 0.32 [0.00] H [0.65] A (+)[0.018] MPRO CONST −9.26 (2.01) R2 0.69 ΔC 0.88 (4.41) LM(4) [0.14] ΣΔC(−i) −0.35 [0.69] FF [0.51] ΣΔP(−i) 0.54 [0.02] N [0.07] ΣΔCU(−i) 0.28 [0.11] H [0.43] A (+)[0.183] MMAN CONST 12.91 (1.61) R2 0.31 ΔC 1.00 (2.43) LM(4) [0.58] ΣΔC(−i) −0.21 [0.28] FF [0.14] ΣΔP(−i) 0.39 [0.19] N [0.78] ΣΔCU(−i) 0.20 [0.02] H [0.10] A (+) [0.742] MV CONST 3.45 (0.64) R2 0.33 ΔC 1.23 (2.00) LM(4) [0.36] ΣΔC(−i) −0.79 [0.07] FF [0.10] ΣΔP(−i) 0.39 [0.11] N [0.54] ΣΔCU(−i) 0.08 [0.26] H [0.01] A (−)[0.772] OM CONST −1.99 (0.87) R2 0.88 ΔC 1.00 (7.68) LM(4) [0.65] ΣΔC(–i) −0.63 [0.00] FF [0.82] ΣΔP(–i) 0.54 [0.01] N [0.42] ΣΔCU(–i) 0.26 [0.00] H [0.37] A (+) [0.490] Notes: The figures in round brackets are t-values. Wald exclusion test significance levels are given in square brackets. Classification changes dictate starting samples of 1978Q1 for MPRO and 1984Q2 for MV. The reported coefficients on ΔCU have been divided by 100 for ease of presentation. ΣΔC(–i), ΣΔP(–i), ΣΔCU(–i) represent the complete lag set of the relevant variables. The corresponding coefficients are summed in the third column and given along with the Wald exclusion test for the set of lagged variables (with significance levels in square brackets). Industry mnemonics are given in Table 12.1. Diagnostics: LM is the LM test for fourth-order serial correlation; FF is the Ramsey RESET test; N is the Jarque-Bera test for normality of residuals; H is a heteroscedasticity test using squared residuals; A is a variable addition joint test for ΔCU2(–i), i=0, 4, with the sign of the coefficient sum and the significance level given in brackets. EIE ΔC ΣΔC(−i) ΣΔP(−i) ΣΔCU(−i)

CONST 0.80 (3.81) −0.29 [0.18] 0.36 [0.22] 0.21 [0.01]

between ΔK and ΔL causes the constraints to appear insignificant and exclusion of both variables is strongly rejected by the data.20 To choose between these constraints we used non-nested tests for the OLS regression and found that the specification

164

SUPPLY CONSTRAINTS AND INFLATION

with ΔK was preferred by both the Akaike and Schwartz information criteria. We would take this as evidence that the capital constraint is also operational for this industry. In general, the significant capital constraints appear to arise in industries where capital is not lumpy (meaning that it can be added in discrete packages such as machine tools or textile machinery). Lumpy capital generally results in building ahead of demand and capacity constraints are accordingly less likely. There must be some concern that capacity constraints tend to occur in several industries that are an important part of the capital goods sector. The various results obtained so far may be summarised in a matrix (Table 12.5) that shows the significance of the utilisation or constraint variables in different industries. In terms of the cross-industry variation in results, we note that our set contains only two industries (FDT and MV) with no capacity utilisation or constraint effects in the IV estimation. It is not clear what explains these exceptions. In the case of MV, the sample is unusually short and OLS estimation for this industry does suggests a significant ΔK effect. As far as FDT is concerned, excise tax may be an important feature of pricing for this industry and since we have not included it in the price equation there may be some omitted variable bias. Table 12.4 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1, IV(2SLS) estimation TEX ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) CHEM ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) FDT ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) EIE ΔG ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) ME ΔG ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) MPRO ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) MMAN

Variable

Coefficient*

Diagnostics

CONST 0.79 (5.33) −0.52 [0.00] 0.72 [0.00] 0.20 [0.35] 0.12 [0.20] CONST 0.80 (4.14) −0.39 [0.24] 0.48 [0.01] 0.13 [0.38] 0.14 [0.05] CONST 0.78 (2.98) -0.17 [0.57] 0.30 [0.45] 0.04 [0.77] 0.15 [0.50] CONST 1.01 (3.97) −0.40 [0.10] 0.28 [0.47] 0.01 [0.64] 0.12 [0.04] CONST 0.77 (4.73) −0.47 [0.03] 0.63 [0.00] 0.17 [0.10] 0.09 [0.17] CONST 1.02 (6.31) −0.49 [0.01] 0.55 [0.00] 0.14 [0.02] 0.21 [0.14] CONST

−3.27 (1.14) LM(4) [0.06] FF [0.23] N [0.89] H [0.72]

R2 0.83

0.11 (0.04) R2 0.80 LM(4) [0.01] FF [0.99] N [0.72] H [0.17] 2.52 (0.64) LM(4) [0.79] FF [0.29] N [0.80] H [0.52]

R2 0.71

−3.53 (1.62) LM(4) [0.29] FF [0.09] N [0.79] H [0.90]

R2 0.83

−1.78 (0.79) LM(4) [0.22] FF [0.71] N [0.84] H [0.51]

R2 0.84

−9.63 (2.46) LM(4) [0.45] FF [0.01] N [0.12] H [0.41]

R2 0.78

13.22 (1.82)

R2 0.31

CIARAN DRIVER AND DAVID SHEPHERD

ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) MV ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i)

Variable

Coefficient*

0.60 (1.85) −0.36 [0.70] 0.24 [0.42] 0.31 [0.01] 0.21 [0.43] CONST 0.63 (2.26) −0.51 [0.11] 0.67 [0.05] −0.04 [0.17] 0.16 [0.43]

LM(4) [0.58] FF [0.14] N [0.78] H [0.10] 5.17 (1.37) LM(4) [0.04] FF [0.10] N [0.99] H [0.00]

165

Diagnostics

R2 0.69

CONST −3.13 (1.32) R2 0.89 1.01 (7.92) LM(4) [0.91] −0.50 [0.01] FF [0.93] 0.41 [0.08] N [0.53] 0.02 [0.89] H [0.10] 0.20 [0.04] Notes: The figures in round brackets are t-values. Wald exclusion test significance levels are given in square brackets. Other notes as for Table 12.2.

OM ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i)

Table 12.5 Significance of the utilisation and constraint variables Variables

IV Estimation

OLS estimation

TEXa TEXa a EIE MEa MPROa MMANa MVa OMa ΔK TEXb TEXc MEc MEa a MPRO MPROa a MMAN MMANa MVa ΔL CHEMa CHEMa a a EIE EIE OMa MEa MPROa OMa a Notes: Likelihood ratio (OLS) or Wald (IV) exclusion test significant at 5 per cent. b ΔK, ΔL jointly significant at 5 per cent but ΔK preferred—see text. c Significant at 10 per cent. ΔCU EIEa MEa MMANa OMa

Inclusion of the competitiveness term So far we have estimated the relationship described by equation (5) without taking account of the competition terms represented by the last term in (3). To test for the significance of this term we need data on changes in concentration and changes in conjectural variation. Given that all of our industries are exposed to a significant degree of import penetration, our approach to this problem is to assume that the conjectural variation term is determined by the strength of foreign competition.

166

SUPPLY CONSTRAINTS AND INFLATION

Since the threat from imports depends substantially on competitiveness, as measured by the real exchange rate, one way to proxy the conjectural variation term is in the form λ=λ (e), with λ′<0, where e is the real exchange rate index (measured by the sterling effective exchange rate index). As the exchange rate appreciates, foreign competition intensifies, and we would expect to see the conjectural variation term falling below the Cournot level of zero. Linearising, we obtain: Writing CR for concentration, the last term in (3) can be approximated by: This enters equation (5) in first difference form: As changes in the exchange rate show more volatility than the concentration ratio, we omit the last term. Including four lags for generality, but excluding the current exchange rate term due to possible simultaneity, we obtain the following additional lag sets to include in equation (6): where the lag is 0–4 for the first term, and 1–4 for the second. Both of these additional variables are stationary at the 5 per cent level. Due to unavoidable data limitations, it is only possible to estimate these additional terms for the period 1980Q1 to 1992Q4. The data on concentration by industry (i.e. the five firm concentration ratio) is only available over this period and it ceased to be published after 1992.21 In view of the truncated sample, we tested the new variables only on equation (6) to conserve degrees of freedom. We tested the additional terms by a joint exclusion test on the nine coefficients. For IV estimation none of the exclusion tests were significant. However, in view of the smaller sample, the advantage of IV estimation is muted and the OLS results may also be of interest.22 Using OLS, we found the exclusion tests to be significant at the 5 per cent level in four industries: TEX, ME, MMAN and OMAN. The results for those industries are given in Table 12.6. The summed coefficients over the lag structures for the new variables are signed correctly. The magnitude of the coefficients also seem plausible, even though the long-run levels effects of competition on price cannot be calculated, since our dependent variable is not in levels. For example, for TEX, a rise in CR of 1 per cent from its mean value would make a 12 per cent difference to the dependent variable at its mean value, amounting to perhaps an additional 0.25 percentage point in the rate of price increase. An appreciation of the exchange rate of 1 per cent at the mean value of CR would take about 1.0 percentage points off the rate of increase in prices, Table 12.6 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1980Q1–1992Q4 TEX ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) ΣΔCR(–i) ΣCR*Δe(–i) ME ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) ΣΔCR(–i) ΣCR*Δe(–i) MMAN ΔC ΣΔC(–1) ΣΔP(–i) ΣΔCU(–i) ΣΔCR(–i) ΣCR*Δe(–i)

Variable

Coefficient (t-value)

Diagnostics

CONST 0.51 (3.07) −0.24 [0.06] 0.63 0.00] 0.34 0.00] 0.69 0.09 −0.03 0.00] CONST 0.54 (3.76) −0.44 [0.01] 0.53 0.00] 0.30 0.00] 1.58 0.27] −0.02 0.01] CONST 0.36 (2.09) −0.12 [0.36] 0.51 [0.02] 0.28 [0.01] 1.49 [0.11] −0.01 [0.00]

1.13 (0.16) LM(4) [0.38] FF [0.08] N [0.76] H [0.69]

R2 0.81

8.81 (0.19) LM(4) [0.19] FF [0.60] N [0.63] H [0.99]

R2 0.82

2.94 (0.21) LM(4) [0.47] FF [0.05] N [0.87] H [0.45]

R2 0.69

CIARAN DRIVER AND DAVID SHEPHERD

Variable

Coefficient (t-value)

167

Diagnostics

OM CONST 3.30 (0.58) R2 0.84 ΔC 0.73 (4.19) LM(4) [0.77] ΣΔC(–i) −0.52 [0.00] FF [0.19] ΣΔP(–i) 0.61 [0.01] N [0.20] ΣΔCU(–i) 0.16 [0.03] H [0.88] ΣΔCR(–i) 6.49 [0.04] ΣCR*Δe(–i) −0.02 [0.06] Notes: The figures in round brackets are t-values. Likelihood exclusion test significance levels are given in square brackets. The last two terms are always jointly significant at the 5 per cent level for these industries.

ignoring any indirect effects through changes in input costs. The latter effect is rather less for the other industries in Table 12.6. In each of the cases reported in Table 12.6, the exclusion tests on the ΔCU terms also continued to be significant at the 5 per cent level and the coefficients were not markedly different than before. This suggests that whatever explanatory power the competition terms may have, their omission does not invalidate the analysis reported in Tables 12.3 and 12.4. Conclusions We have considered the extent to which different supply constraints affect the behaviour of the price-cost relationship in the UK manufacturing sector. Our results suggest that price generally moves in a procyclical manner relative to cost. The result on procyclicality, while not uncontroversial, is in line with a number of recent studies cited earlier. Our results allow for an interpretation of the cyclical movement. Standard cost theory suggests that the effect of capacity utilisation, in so far as it captures variation in cost with output, should be concave quadratic in the absence of supply constraints. This hypothesis was rejected for eight out of nine industries studied. Satisfactory results were obtained by directly entering survey-based capacity constraints as determinants of price. It was found that both labour and capital constraints each had significant independent effects on price in a number of industries. This result was found to hold for both OLS and IV estimation. Over a shorter time period, it was possible to include a proxy for thestate of competition or collusion. In this case, we found that the impact of the cyclical variable remained significant in the presence of the competition variable in the four industries where the latter appeared to be significant. Expressed in a practical context, it appears that both capital and labour constraints have been sources of inflationary pressure in the economy, in the sense that they are factors which have constrained supply relative to demand. This conclusion highlights the importance of an adequate response of fixed investment to expected future demand, alongside any measures designed to increase the effective availability of labour. It also suggests that any attempt to measure the elusive ‘output gap’ should focus on the availability of both capital and labour resources. Appendix 12.1: data sources Detail of the survey questions used The principal data source for this paper is the Industrial Trends Survey carried out by the main UK employers organisation, the Confederation of British Industry, and described in Junankar (1989). This data base is the most reputable of UK survey sources. It has been published on a regular basis since 1958 and has been widely used by economists. The responses in the survey are weighted by net output with the weights being regularly updated. The survey sample is chosen to be representative and is not confined to CBI members; over 1500 responses are received each quarter. The exact wording of the survey questions extracted from the CBI Industrial Trends Survey is given below for the questions used in the text. Question 4 Is your present level of output below capacity (i.e., are you working below a satisfactory full rate of operation?)

168

SUPPLY CONSTRAINTS AND INFLATION

Question 11 Excluding seasonal variation what has been the trend over the past four months with regard to average costs per unit of output? What are the expected trends for the next four months? Question 12 Excluding seasonal variation what has been the trend over the past four months with regard to average prices at which domestic orders are booked? Question 14 What factors are likely to limit your output over the next four months: please tick the most important factor or factors? Variable definitions ΔP ΔC ΔCU ΔK ΔL CR e gp

Percentage of respondents replying ‘up’ to Question 12. Percentage of respondents replying ‘up’ to Question 11. First difference of the logit of the percentage of respondents replying ‘no’ to Question 4. First difference of the logit of the percentage of respondents checking ‘plant capacity’ for Question 14. First difference of the percentage of respondents checking ‘skilled labour’ for Question 14. Proportion of sales accounted for by the five largest firms in each 3-digit industry, weighted by sales and summed to correspond to CBI Industry Groups. Sterling real effective exchange rate index (Bank of England). The rate of growth of the producer price index for UK manufacturing (UK Economic Trends). Appendix 12.2: relationship between CBI qualitative data and the inflation rate for total manufacturing

The traditional approach to the problem of translating qualitative data into quantitative data has been to assume a point estimate for the meaning of ‘up’ and ‘down’ (Anderson 1952; Theil 1952; Lee 1993). Recent work has tended to assume that the relationship between the actual and survey data is inherently non-linear. The rationale behind this assumption is that the rate of inflation (Δp) can be represented as a weighted sum of upward price movements (Δp+) and downward movements (Δp−), with the weights corresponding to the percentages of firms replying up or down. A linear relationship is then valid only if the meaning of ‘up’ is constant across firms and over time.23 Pesaran (1984) assumes that the ‘up’ replies correspond to a point estimate that depends non-linearly on the actual rate of inflation. In this case, the relationship can be represented as: (A2/1) The alternative linear form is: (A2/2) where gp represents the actual percentage change in manufacturing prices, R and F indicate ‘up’ and ‘down’ responses, and e is an error term. Pesaran shows that all of the relevant information from the qualitative data is contained in the R variable (the percentage responding up) and this is confirmed for our sample period using CBI information on the ups and downs for total manufacturing and the official (ONS) series for the index of producer prices in the manufacturing sector. However, while Pesaran found in favour of the non-linear form for the period 1958–81, we could not confirm this result for our 1976–97 sample. Using a linear relationship, with ML estimation of a first order autoregressive error structure, we obtain a satisfactory equation for the quarterly change in manufacturing prices as shown below.24

(A2/3)

CIARAN DRIVER AND DAVID SHEPHERD

169

R_bar_squared=0.73; DW=2.17; σ=0.006; Mean Dependent Variable =0.016; t-values in brackets. Inspection of the residual correlogram revealed it to be flat and the residual histogram was approximately Normal. We attempted to test the non-linear form of the relationship by direct non-linear ML methods, but the results were very poor. As an alternative, we logged both sides of (A2/1), ignoring the F term, and approximated log(1–λR) by a Taylor expansion approximation. The resulting equation was estimated by ML methods with an autoregressive error structure and we failed to find any significance for the (1–λR) term. We conclude that while the Pesaran approach is theoretically attractive, it does not appear to perform better than the linear model for our sample. A possible explanation is that while notions about the meaning of R may not be constant over time in the minds of respondents, they do not change quickly in response to movements in actual inflation. Finally, we note that while the linear model was deemed inadequate by Pesaran because its parameters were unable to represent the full range of observed inflation, this is much less of a problem with our sample. The coefficient on R in equation (A2/3) is 0.068. In conjunction with a maximum R of 0.73, this gives a maximum quarterly inflation rate of about 0.05, compared with an observed maximum of 0.058.25 Our intention in this matter has been to arrive at a representation of the survey data responses that corresponds with observed price and cost inflation in the UK manufacturing sector. The data analysis discussed above suggests that for our sample it is reasonable to take the untransformed R responses as appropriate measures of price and cost inflation in the sector. As it turns out, this procedure is also convenient because problems connected with data matching would make it difficult to check any appropriate transformation for the disaggregated industries.26 Appendix 12.3. OLS results Table 12.A.1 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1 TEX ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) CHEM ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) FDT ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) EIE ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) ME ΔC ΣΔC(–i) ΣΔP(–i) ΣΔCU(–i) MPRO ΔC ΣΔC(–i) ΣΔP(–i)

Variable

Coefficient*

Diagnostics

CONST 0.58 (6.11) −0.41 [0.00] 0.73 [0.00] 0.37 [0.00] CONST 0.76 (6.81) −0.27 [0.24] 0.38 [0.00] 0.14 [0.18] CONST 0.62 (6.09) 0.00 [0.32] 0.11 [0.35] 0.05 [0.50] CONST 0.59 (6.141) −0.16 [0.16] 0.45 [0.03] 0.21 [0.00] CONST 0.61 (6.12) 0.39 [0.01] 0.66 [0.00] 0.36 [0.00] CONST 0.54 (4.59) −0.04 [0.85] 0.54 [0.00]

−0.03 (0.01) LM(4) [0.05] FF [0.69] N [0.95] H [0.75] 1.77 (0.61) LM(4) [0.14] FF [0.28] N [0.62] H [0.46] 4.33 (1.23) LM(4) [0.69] FF [0.98] N [0.45] H [0.14] −1.94 (1.14) LM(4) [0.31] FF [0.95] N [0.03] H [0.69] 0.27 (0.13) LM(4) [0.01] FF [0.34] N [0.98] H [0.27] −8.88 (2.06) LM(4) [0.16] FF [0.27] N [0.62]

R2 0.81

R2 0.76

R2 0.71

R2 0.86

R2 0.84

R2 0.73

170

SUPPLY CONSTRAINTS AND INFLATION

ΣΔCU(–i)

Variable

Coefficient*

0.36 [0.00]

H [0.66]

Variable

Coefficient*

Diagnostics

Diagnostics

MMAN CONST 15.51 (2.37) R2 0.52 ΔC 0.40 (3.40) LM(4) [0.16] ΣΔC(–i) −0.33 [0.22] FF [0.44] ΣΔP(–i) 0.41 [0.01] N [0.45] ΣΔCU(–i) 0.27 [0.00] H [0.57] MV CONST 4.51 (1.27) R2 0.71 ΔC 0.44 (3.63) LM(4) [0.96] ΣΔC(–i) −0.38 [0.01] FF [0.08] ΣΔP(–i) 0.77 [0.00] N [0.55] ΣΔCU(–i) 0.09 [0.2] H [0.57] OM CONST 1.56 (0.69) R2 0.88 ΔG 0.89 (9.20) LM(4) [0.59] ΣΔC(–i) −0.55 [0.0] FF [0.02] ΣΔP(–i) 0.57 [0.00] N [0.37] ΣΔCU(–i) 0.27 [0.00] H [0.32] Notes: The figures in round brackets are t-values. Likelihood exclusion test significance levels are given in square brackets. Classification changes dictate starting samples of 1978Q1 for MPRO and 1984Q2 for MV. Other notes as for Table 12.3. Table12.A.2 Dependent variable: percentage responding ‘up’ to price question (ΔP), sample 1976Q4–1997Q1 TEX ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) CHEM ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) FDT ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i) EIE ΔC ΣΔC(–i) ΣΔP(–i) ΣΔK(–i) ΣΔL(–i)

Variable

Coeffident (t-value)

Diagnostics

CONST 0.59 (5.85) −0.37 [0.00] 0.76 [0.00] 0.24 [0.06] 0.11 [0.15] CONST 0.73 (7.47) −0.38 [0.08] 0.50 [0.00] 0.13 [0.23] 0.14 [0.02] CONST 0.53 (4.91) 0.02 [0.18] 0.21 [0.04] 0.05 [0.35] 0.05 [0.17] CONST 0.66 (6.32) −0.16 [0.08] 0.38 [0.11] −0.02 [0.65] 0.12 [0.01]

−2.40 (0.88) LM(4) [0.19] FF [0.37] N [0.94] H [0.84]

R2 0.84

0.40 (0.15) LM(4) [0.01] FF [0.21] N [0.66] H [0.08]

R2 0.80

3.96 (1.12) LM(4) [0.91] FF [0.93] N [0.30] H [0.15]

R2 0.74

−2.37 (1.27) LM(4) [0.45] FF [0.92] N [0.77] H [0.90]

R2 0.86

CIARAN DRIVER AND DAVID SHEPHERD

Variable

Coefficient (t-value)

171

Diagnostics

ME CONST −1.53 (0.70) R2 0.85 ΔC 0.62 (6.04) LM(4) [0.43] ΣΔC(–i) −0.38 [0.01] FF [0.22] ΣΔP(–i) 0.70 [0.00] N [0.80] ΣΔK(–i) 0.19 [0.03] H [0.18] ΔL(–i) 0.09 [0.10] MPRO CONST −7.25 (2.13) R2 0.83 ΔC 0.66 (6.82) LM(4) [0.50] ΣΔC(–i) −0.23 [0.03] FF [0.27] ΣΔP(–i) 0.63 [0.00] N [0.17] ΣΔK(–i) 0.16 [0.00] H [0.21] ΣΔL(–i) 0.20 [0.04] MMAN CONST 14.69 (2.18) R2 0.59 ΔC 0.34 (2.97) LM(4) [0.17] ΣΔC(–i) −0.13 [0.70] FF [0.06] ΣΔP(–i) 0.23 [0.33] N [0.28] ΣΔK(–i) 0.35 [0.00] H [0.46] ΣΔL(–i) 0.21 [0.31] MV CONST 5.17 (1.37) R2 0.71 ΔC 0.41 (3.32) LM(4) [0.28] ΣΔC(–i) −0.40 [0.02] FF [0.36] ΣΔP(–i) 0.78 [0.00] N [0.45] ΣΔK(–i) 0.01 [0.03] H [0.00] ΣΔL(–i) 0.12 [0.19] OM CONST −2.70 (1.16) R2 0.89 ΔC 0.92 (9.49) LM(4) [0.90] ΣΔC(–i) −0.45 [0.00] FF [0.07] ΣΔP(–i) 0.45 [0.01] N [0.40] ΣΔK(–i) 0.03 [0.78] H [0.10] ΣΔL(–i) 0.19 [0.01] Notes: The figures in round brackets are t-values. Likelihood exclusion test significance levels are given in square brackets. Classification changes dictate starting samples of 1978Q1 for MPRO and 1984Q2 for MV. Other notes as for Table 12.3.

Notes 1 We would like to thank Stephen Hall and participants in a Bank of England seminar for helpful comments on a preliminary draft of this chapter. Any remaining errors are ours. 2 Temple and Urga (1997) address the effects of separate supply constraints on manufacturing imports. 3 Survey evidence from the CBI database referenced later in the chapter also shows that constraints are important. For example, averaged over the eighty or so quarters in our samples, one in six respondents cite the capital stock as a constraint on production. 4 Hay and Morris (1991) remind us that the cyclical behaviour of price becomes very much more complex once we allow for uncertainty with respect to cost or demand. See Driver et al. (1996) for a model of price behaviour under uncertainty with supply constraints. 5 A particular complication for an open economy is that some allowance may have to be made for the impact of foreign competition. 6 From a different perspective, Green and Porter (1984) argue that if output is not strongly pro-cyclical in the face of unobservable industry demand then price should be pro-cyclical. 7 A further finding in the UK survey suggested that in response to demand booms, the primary response would be to increase labour hours. An increase in price was, however, next in importance, ahead of the decision to add capacity. 8 There is also a raft of studies on the cyclical behaviour of the markup (Haskel and Martin 1995; Small 1998). 9 The price response to cost under rationing presumably depends on whether firms wish to exploit a position of temporary excess demand. This depends in turn on considerations such as limit pricing, and the relative attractiveness of immediate cash flows and

172

10 11 12

13 14

15

16 17 18

19 20 21

22 23

24

25 26

SUPPLY CONSTRAINTS AND INFLATION

long-run relationships with customers (Eichner 1976). The observed price under rationing also depends on whether the supplier can choose which markets to ration. It is also similar to the kind used in macroeconomic models (see, for example, Romer 1996). It may easily be shown that if γ1 is normalised to unity, the term (r–1) rises from zero at the q value corresponding to minimum average variable costs (−γ2/2γ3) to a maximum at q=−2/γ2(1+(1+1/γ3)0.5 Specifically, if the distribution of utilisation across firms is Sech-squared (an approximation to the normal), the observed data (CS) on the proportion working above a satisfactory level of utilisation corresponds to the integral of the Sech-square density function from the satisfactory threshold to infinity. Thus CS=1/(1+exp(a−bU)) where the argument of the exponential term in U is a linear measure of capacity utilisation that can be recovered by taking the logit of CS. Driver (1986) shows that CS has a normal distribution. It could be argued that, even if such a relationship is unknown, a study of the factors determining the qualitative answers is of interest in itself. The actual rate of price change in manufacturing, detailed in Appendix 12.1, is effectively an index of the price of gross output. However, it excludes intra-industry transactions and for this reason, and because the CBI data are somewhat biased towards larger firms, we would not expect an exact correlation between the corresponding time series. This assumes that factor substitution allows full employment to be reached with any capital stock. Under these conditions, the availability of labour is the main determinant of whether firms have capacity to meet the current level of demand, and the size of the capital stock exercises its influence principally via the marginal product of labour, which determines the feasible path of real wages. Specifically, the instruments are one-period and two-period lags on the one-period ahead forecasts for the ΔC variable, obtained from the same survey data source, recorded as the second part of the answer to question 11, reproduced in Appendix 12.1. We tested for moving average error structures, as implied by the differencing of equation (3). However, estimation results were not superior to the OLS or IV results reported in the text and occasionally the estimation procedure failed to converge. In chemicals, unit material cost may fall sharply with output, implying a relatively large γ2 term in the cost equations. From note 11, this may indicate that (r–1) approaches a maximum plateau at a relatively low q level, suggesting a strongly convex term in CU in equation (5). Strictly speaking we should exclude GHEM from this second set of estimations as we found no a priori evidence of constraints in this sector. However, as this was the only such industry, we cannot be sure that the result was not obtained by chance. The R-bar-squared falls by less than 1 percentage point if either of these on its own is omitted, but falls by 12 points if both are omitted. We are grateful to Professor Steve Davies for making available the data at 3-digit level and for providing us with industry weightings to construct the ratio at the required level of aggregation, matching the CBI industry groups. The correspondence used to arrive at the industry series is provided in a version of Lee (1993), published as a working paper by the University of Leicester. It should also be noted that the small sample distribution of the Wald statistic is unknown. Pesaran (1984) writes the growth rate of the survey variable (e.g. price) as: where the superscripts denote a rise or fall in the variable and the Wi are the weights of the firms in total manufacturing. He then shows that the balance of the ‘ups’ and ‘downs’ may be interpreted as growth rates under the assumption of a common interpretation of ‘up’ or ‘down’ across respondents. A less restrictive assumption is that Xi is distributed uniformly with symmetrical supports about zero. Carlson and Parkin (1975) derive a growth rate from the raw Xi assuming a normal distribution. Foster and Gregory (1977) criticise the normality assumption for producing bias in contexts where asymmetry might be expected. The balance measure is probably most appropriate where there is not a substantial deterministic trend. Where the ‘ups’ and ‘downs’ are collinear, it may be more satisfactory to omit one of the variables as in Pesaran (1984), Thomas (1995). See also Cunningham (1997) for a review of quantifying survey data. Use of the quarterly series is problematic because the survey data refers to the previous four months. This is a hangover from the period when the survey was carried out three times a year and we suspect that many respondents respond in terms of the time since the last survey. In the reported regression we have entered a single lead on the survey response, as this is published close to the beginning of the quarter. The equation is also satisfactory if the contemporaneous value is used, but slightly inferior to that in the text. For completeness, we also checked the behaviour of the unit cost variable against the behaviour of unit labour costs in manufacturing. Once again, the linear model with ‘ups’ seemed more appropriate than the non-linear model. Even for the total sample this problem arises since the coverage of the CBI Survey is biased towards larger firms.

References Anderson, O., Jr (1952) ‘The Business Test of the IFO-Institution for Economic Research, Munich and its Theoretical Model’, Revue de l’Institut International de Statistique, 20, 1–25. Bean, C. and Gavosto, A. (1990) ‘Outsiders, Capacity Shortage and Unemployment in the United Kingdom’, in J.H.Dreze and C.Bean (eds), Europe’s Unemployment Problem, Cambridge, MA, MIT Press. Blinder, A. (1991) ‘Why are Prices Sticky? Preliminary Results from an Interview Study’, American Economic Review, May, 89–96. Carlson, J.A. and Parkin, M. (1975) ‘lnflation Expectations’, Economica, 42, 123–38. Chan, A., Savage, D. and Whittaker, R. (1995). The New Treasury Model’, Treasury Working Paper, 70. Charemza, W.W. and Deadman, D.F. (1992) New Directions in Econometric Practice, Aldershot, Edward Elgar. Cowling, K. (1980) Monopoly Capitalism, London, Macmillan.

CIARAN DRIVER AND DAVID SHEPHERD

173

Cunningham, A. (1997) ‘Quantifying Survey Data’, Bank of England Quarterly Bulletin, August, 292–300. Dreze, J.H. and Bean, C. (1990) ‘Europe’s Unemployment Problem: Introduction and Synthesis’, in J.H.Dreze and C.Bean (eds), Europe’s Unemployment Problem, Cambridge, MA, MIT Press. Driver, C. (1986) ‘Transformation of the CBI Capacity Utilisation Series: Theory and Evidence’, Oxford Bulletin of Economics and Statistics, 48(4), 339–52. Driver, C., Abubacker, S. and Argiris, G. (1996) ‘Capacity Choice under Monopoly, Flexible Price and Demand Uncertainty’, Southern Economic Journal, 63(2), 526–32. Eichner, A. (1976) The Megacorp and Oligopoly, New York, Basic Books. Foster, J. and Gregory, M. (1977) ‘Inflation Expectations: the Use of Quantitative Survey Data’, Applied Economics, 9, 319–29. Geroski, P.A. (1992) ‘Price Dynamics in UK Manufacturing: a Microeconomic View’, Economica, 59, 403–19. Green, E.J. and Porter, R.H. (1984) ‘Non-cooperative Collusion Under Imperfect Price Information’, Econometrica, 52, 87–100. Hall, S., Walsh, M. and Yates, T. (1996) ‘How do UK Companies Set Prices?’, Bank of England Quarterly Bulletin, 36(2), 180–92. Haskel, J., Martin, C. and Small, I. (1995) ‘Price, Marginal Cost and the Business Cycle’, Oxford Bulletin of Economics and Statistics, 57, 1, 25–41. Hay, D.A. and Morris, D.J. (1991) Industrial Economics and Organisation, Oxford, Oxford University Press. Im, K.S., Pesaran, M.H. and Shin, Y. (1995) ‘Testing for Unit Roots in Dynamic Heterogeneous Panels’, WP 9526, DAE, University of Cambridge, revised June 1996. Junankar, S. (1989) ‘How Do Companies Respond to the Industrial Trends Survey?’, paper given to CBI Conference, Centrepoint, 6 November. Lee, K. (1993) ‘Formation of Price and Cost Inflation Expectations in British Manufacturing Industries: a Multi-Sectoral Analysis’, Economic Journal, 104, 372–85. Minford, M., Wall, M. and Wren-Lewis, S. (1988) ‘Manufacturing Capacity: a Measure Derived from Survey Data Using the Kalman Filter’, Discusion Paper 146, London, National Institute of Economic and Social Research. Nickell, S.J. (1978) The Investment Decision of Firms, Cambridge, Cambridge University Press. Pesaran, M.H. (1984) ‘Expectations Formation and Macroeconometric Modelling’, in P. Malgrance and P.A. Muet (eds), Contemporary Macroeconomic Modelling, Oxford, Basil Blackwell. Romer, D. (1996) Advanced Macroeconomics, New York, McGraw-Hill. Rotemberg, J. and Saloner, G. (1986) ‘A Super Game-Theoretic Model of Price Wars During Booms’, American Economic Review, 76, 390–407. Sentance, A. and Emerson, R. (1995) ‘Manufacturing Capacity and Investment’, Report to the Foundation for Manufacturing and Industry, August. Small, I. (1998) ‘The Cyclicality of Mark-ups and Profit Margins: Some Evidence for Manufacturing and Services’, Bank of England Quarterly Review, August, 267–73. Temple, P. and Urga, G. (1997) ‘The Competitiveness of UK Manufacturing: Evidence from Imports’, Oxford Economic Papers, 49(2), 207–27. Theil, H. (1952) ‘On the Time Shape of Economic Microvariables and the Munich Business Test’, Revue de L’institut International de Statistique, 20, 105–20. Thomas, D.G. (1995) ‘Output Expectations within Manufacturing Industry’, Applied Economics, 27, 403–8. Weitzman, M. (1982) ‘Increasing Returns and the Foundations of Unemployment Theory’, Economic Journal, 92, 787–804.

13 Long-run effects of investment incentives Michael Sumner

Introduction Three stylised facts summarise the fiscal treatment of investment during the last half century. The first is the variety of instruments deployed: successive Chancellors of the Exchequer have shown much ingenuity in designing new policy tools, and new labels for old ones. Second, if these various instruments can legitimately be represented on a uniform scale by computing their effects on the present value of capital costs, the stance as well as the form of fiscal policy has varied substantially: in 1958 the post-tax cost of capital (for investment in machinery) exceeded its pre-tax level by 26 per cent; in 1967, the net effect of the tax system and investment grants (in manufacturing) was to reduce the pre-tax cost of capital by 9 per cent. Finally, there has been a marked reduction in the frequency of change. Fiscal provisions affecting investment were altered in fourteen of the eighteen years ending in 1973. Thereafter one major reform was initiated in 1984, and there were two trivial changes in the early 1990s. The question which naturally arises is, so what? Are policy-induced changes in the post-tax cost of capital reflected in the time-profile of investment? Does the form as well as the present value of policy instruments matter? A student of the voluminous literature which had accumulated by the end of the period of policy activism concluded, ‘Perhaps the most that can be claimed…is that the studies discussed…do not support any allegation that the incentives have been wholly ineffective. It is difficult to go much further’ (Lund 1976, p. 261). It is equally difficult to draw a more definite conclusion from the subsequent literature. With the decline of activism academic attention has refocused on particular episodes, notably the announcement effects of the 1984 reform (Robson 1989; Sumner 1992), and on microeconomic issues at the level of the individual firm (e.g. Blundell et al. 1996). The object of this exercise is to provide answers to these questions at the macroeconomic level by estimating a simple model of investment over the longest feasible period. A simple model The first stage of the estimation is based on the standard neoclassical model of cost minimisation conditional on output. By adopting Bean’s (1981) convenient approximations to eliminate the capital stock along an equilibrium growth path, investment (in plant and machinery by manufacturing industry, I) can be expressed as a log-linear function of (manufacturing) output (Y) and relative factor prices before (RP) or after tax (RP*Fi). The measurement of fiscal policy’s effect on relative prices is described in the following section. The second stage is an error-correction model which incorporates deviations from the equilibrium growth path and allows the role of additional, short-term influences on investment to be assessed. Candidates for consideration include capacity utilisation (CU) and ex post real interest rates, measured in terms of wage inflation (RRW) or the rate of change of capital goods prices (RRPK). As an alternative to conventional present-value measures of fiscal impact on relative factor prices, a set of dummy variables (Di), which distinguish between different phases of fiscal policy according to the principal instruments employed, is also examined. Fuller definitions of the variables are given in the data appendix. The assumptions about their time-series properties implicit in this description of the model are considered explicitly in a subsequent section. Measurement Several summary measures of fiscal influence on the investment decision have been described in the literature. The most frequently used index incorporates the present value of all tax allowances on a unit investment discounted at a current nominal interest rate, conventionally the long gilt rate to reflect the low risk attached to future tax reductions (FR). A British survey (CBI 1994) echoes US findings by Summers (1987) in reporting that the discount rates used in investment appraisal

MICHAEL SUMNER

175

are rarely revised in line with market movements, hence an alternative is to use a fixed rate, here 10 per cent (FF). Both these measures will understate the marginal tax rate on investment if, as both surveys suggest, discount rates are markedly higher than market rates. The limiting representation of the latter possibility is that future allowances are simply ignored (King 1972), so that only first-year allowances enter the calculations (FFY). The cost of capital conventionally includes a term in expected capital gains, which depend in part on expected future fiscal provisions (e.g. Bond et al. 1993); and on several occasions the authorities have announced explicitly temporary or future permanent changes in those provisions. At the end of 1966 investment grants were increased for the following two years. The Chancellor announced in the 1972 budget a switch in the following year from the classical to the imputation form of corporation tax, implying tax relief at a higher rate if investment were delayed. Phased reductions in both capital allowances and the corporate tax rate were announced in 1984. Finally, a one-year increase in the first-year allowance was introduced in 1992. These announcements created incentives to change the timing of investment even for firms which, because of a fixed proportions technology or entrepreneurial incompetence, were beyond the neoclassical pale. The expected cost changes implied by the announcements (DFi) were therefore included in the error correction model as separate variables instead of being incorporated in the measures of fiscal influence. Separate accounting for the intertemporal substitution effect of the announcements still leaves the cost of capital abnormally low in years such as 1967–68 and, less obviously, in 1984, when the tax rate on future returns was lower than that applied to the first-year allowance for capital costs. These blips imply a temporarily stronger incentive to substitute capital for labour. To allow for the possibility that this incentive was for any reason inoperative, the blips were removed (from the fixedrate series) to produce a fourth measure of fiscal impact (FA). There are no grounds for assuming that all unannounced changes in fiscal policy were unanticipated. During the period of activism when change was almost an annual event, there would be little to lose and potentially much to gain by advancing or retarding expenditure before the budget in accordance with contemporary economic indicators and official pronouncements on them; the results reported by Feldstein and Flemming (1971), who estimated a larger response to fiscal policy than to other components of the rental price of capital services, are certainly consistent with this possibility. To minimise the consequential risk of finding a spurious association between investment and the fiscal measures, annual data are used in order to smooth out any such speculative change in the timing of expenditure. The sample period begins in 1955 to exclude the supply constraints of the earlier postwar years. Summary statistics on the variability of and associations between the four fiscal indicators are provided in Table 13.1. Integration and Cointegration The Dickey-Fuller tests (augmented where appropriate) reported in Table 13.2 confirm the assumptions made above that investment, output and pretax relative factor prices are all I(1) variables, while capacity utilisation and the two measures of real interest rates are stationary. The status of the fiscal Table 13.1 Measures of fiscal policy Variable

Coefficient of variation %

Correlation coefficient

FF

FFY

FA

FR FF FFY FA

9.3 8.3 18.4 7.3

0.89

0.11 0.41

Table 13.2 Order of integration Variable

Null hypothesis I(1)

I(2)

I Y RP CU RRW RRPK FR

−2.21 −1.97 0.54 −4.18 −3.50 −3.62 −2.93

−6.87 −4.82 −4.17

−7.25

0.80 0.89 0.55

176

LONG-RUN EFFECTS OF INVESTMENT INCENTIVES

Variable

Null hypothesis I(1)

I(2)

FF −2.67 −7.37 FFY −1.61 −5.72 FA −2.10 −7.06 RP*FR −0.92 −7.72 RP*FF −0.92 −7.79 RP*FFY −1.65 −5.23 RP*FA −0.87 −7.25 Notes: a All variables except RRW and RRPK are entered as logarithms. b Data for 1955–94 except for CU, which begins in 1958. c Critical value at 5 per cent level is –2.94.

policy measures is less clear. These variables are necessarily bounded, and in practice their fluctuations lie well within those bounds; but on a literal interpretation of the test statistics they appear as I(1). For present purposes, however, the ambiguity is immaterial, for the post-tax measures of relative prices are all definitely I(1). Application of the Johansen procedure, with an optimal lag in the VAR of two, indicated (at most) one cointegrating vector among the I(1) components of the model, with an adjustment matrix consistent with its interpretation as an investment function. The numerical results for the alternative measures of relative factor prices are summarised in Table 13.3. The null of non-cointegration is comfortably rejected using the pre-tax ratio. The output elasticity differs insignificantly from unity and hence is consistent with constant returns to scale. The price elasticity is credible, and significantly different from zero. The inclusion of either FR or FF in the price ratio reduces the test statistics and weakens confidence in cointegration. The additions also yield less plausible estimates of the output elasticity. The two remaining measures of fiscal influence do not reduce the test statistics, but they do reduce the estimated output elasticity even further. If the pre-tax price ratio and the fiscal contribution are entered separately to relax the constraint of equal coefficients, the null of non-cointegration cannot be rejected with even 90 per cent confidence using FA. The inference of cointegration survives the decomposition using FFY, but this variable is highly insignificant, with a p-value for the chi-squared exclusion test of 0.24. In summary, measuring relative prices on an after-tax basis is not necessary Table 13.3 Cointegration tests Relative price measure

Maximum eigenvalue test Trace test Y

Normalised parameters

Relative price

RP 26.5 35.6 0.91 0.25 RP*FR 24.2 33.7 0.84 0.35 RP*FF 24.6 34.2 0.77 0.36 RP*FFY 27.7 43.2 0.62 0.39 RP*FA 28.4 37.6 0.74 0.38 Notes: a Critical values (incorporating the Cheung-Lai (1993) adjustment) for the maximum eigenvalue and trace tests of the null hypothesis of no cointegrating vector at the 95 per cent (90 per cent) condidence level are 24.4 (21.6) and 34.4 (31.0) respectively. b Variables entered as logarithms. c Estimation period is 1955–94 with maximum lag in the VAR of 2.

for cointegration; on the contrary, it weakens the performance of the model. Any long-run influence of fiscal policy on the investment decision is not captured by any of the measures employed here. Error-correction models The second stage of the estimation process used the lagged residuals from the cointegrating regression which included pre-tax prices as the error-correction term (ECT) throughout. Alternatives from the other equations listed in Table 13.3 were checked but, as in the first stage, the other specifications all produced inferior results. Other negative findings are failure to detect any role for either version of real interest rates or for any of the fiscal policy measures, whether entered as levels or first differences, at any lag. Few of these variables, in any combination, were correctly signed, and none had a t-ratio even approaching unity. Unsuccessful variations included eliminating negative observations on ex post real interest rates, splitting each measure into its separate components, and interacting the fiscal variables with capacity utilisation to allow for the possibility of differences in response over the cycle.

MICHAEL SUMNER

177

The positive outcome of the exercise is reported in Table 13.4. The error-correction term itself is always highly significant, confirming the inference of cointegration. The other variables included in all versions are capacity utilisation, lagged increases in output, relative factor price acceleration, the expected changes in capital costs implied by announcements of fiscal change (DF), and a lagged dependent variable. The relative price term is the pretax measure; inclusion of any of the fiscal indices again weakened the results. Announcements of fiscal change, however, are strongly significant. Those included in the DF series relate to 1968, 1972 and 1984–85, all involving substantial differences in cost between years. The episode in 1992–93 was on a much smaller scale, and any effects would have been limited to rescheduling during 1993, hence its omission. Table 13.4 Error correction models ECT (5.03) CU (4.02) dRP—dRP(–1) (4.13) Max. (dY(–1), 0) (1.90) dI(–1) (2.37) DF (3.30) Intercept (3.48) D 1958–72

−0.56 (4.37) 0.12 (4.88) 0.43 (4.39) 0.78 (1.83) 0.24 (2.19) 0.63 (3.81) 1.24

−0.48 (6.97) 0.16 (6.26) 0.43 (4.42) 0.72 (1.96) 0.21 (2.08) 0.72 (4.60)

−0.41 0.17 0.42 0.74 0.17 0.79

0.87 (2.34)

D 1973–83

0.90 (2.44)

D 1984–94

0.85 (2.26)

FA3

0.64 (4.39) 0.89 3.94 −1.80 0.37 0.46 0.45

R2 0.87 0.89 SER (%) 4.18 3.86 Durbin h 0.02 −1.49 Functional form 0.38 0.50 Normality 0.36 0.55 Heteroscedasticity 0.84 0.62 Parameter stability 0.42 0.73 Chow 0.48 0.80 Notes: a t-ratios in parentheses. b Last five rows are prob-values of the relevant F or chi-square (for normality) statistic. c Break date for predictive failure and Chow tests is end–1982. d d denotes first-difference operator.

The baseline results of the first equation are supplemented in the second by dummy variables representing qualitative differences in fiscal policy. The initial specification consisted of five dummies, representing the period of extreme activism ending in 1965, when the investment allowance was the principal instrument; the era of investment grants in 1966–70; a short interlude of lower valued initial allowances in 1971–2; the period when full deduction of cost in the year of purchase was allowed, terminated in the 1984 budget; and the residual period of low tax rates and low allowances. The dummies were then combined to reduce the standard error of the regression, with the three dummies included in the second equation as the result. Somewhat surprisingly, the differences in the choice of policy instrument during the period of activism do not show up at all in retrospect, though they provoked much contemporary comment.

178

LONG-RUN EFFECTS OF INVESTMENT INCENTIVES

The coefficients of the dummy variables are (just) significantly different on a Wald test, but they are strikingly similar to the ordering of the mean values of the fixed-rate measures of F (FF and FA) during the three sub-periods distinguished. The third equation of Table 13.4 illustrates by substituting the sub-period means of FA (FA3) for the dummies. A natural interpretation of these results is that qualitative differences between the various policy instruments are not, after all, important. The common currency of present value is a legitimate representation of business perceptions of fiscal policy; but the calculation of those values is insensitive to market interest rates, for the substitution cannot be made using the mean values of FR. For an analogous reason, the extreme myopia represented by the view that only first-year allowances matter is also ruled out. The averaging process carries the obvious implication that the calculation of perceived present value is also insensitive to unannounced changes in the fiscal parameters themselves: activism is indistinguishable from noise. A simple means of tentatively quantifying the effects of fiscal policy is to compare the regimes represented by the second and third dummy variables in Table 13.4. Each regime was introduced with an assurance of long-term stability and each proved durable in the event, so that uncertainty about future policy changes seems unlikely to have affected business response. The 1984 reform raised the relative after-tax price of capital by about 10 per cent (in terms of FF) and, comparing the coefficients of the dummies, investment was 6 per cent lower than otherwise after the reform. The implied elasticiy is considerably higher than that of pre-tax prices. Whether the difference reflects the relative permanence of the tax change, a difference in dispersion across firms, or reliance on a single instance of change are questions which require more data. Conclusions It is no easier to reach firm and positive conclusions about the effects of fiscal policy on investment than it was twenty years ago. The difficulty of so doing cannot be attributed to a fixed proportions technology, for pretax factor prices have welldetermined effects, or to the limited variations in fiscal instruments since the 1970s. On the contrary, year-to-year variations appear to be totally discounted unless they are announced in advance. Announced policy changes do have an effect on the timing of expenditure, but it would be difficult to envisage a problem of demand management to which the rescheduling caused by an announcement would constitute a solution. Data appendix I Y RP CU RPW RRPK DF

Fi

Manufacturing investment in plant and machinery at 1990 prices. Index of manufacturing production. Ratio of index of earnings in manufacturing to implicit deflator for I. CBI index of capacity utilisation. Twenty year bond rate minus rate of wage inflation. Twenty year bond rate minus rate of change of implicit deflator for I. Expected change in capital costs caused by announcements of future changes in fiscal parameters. Values are 9.6 per cent in 1968, −16.7 per cent in 1972, 13 per cent and 7.4 per cent in 1984–5. The value associated with the 1992 announcement was much smaller, at 1.9 per cent, and any response would have been reversed within 1993. (1–t)/(1–Zi), where t is the corporate tax rate and Zi is the present value of cash grants and tax allowances on manufacturing plant and equipment, discounted at the long bond rate (for FR), at a fixed rate of 10 per cent (FF), including only grants and first-year allowances (FFY), and excluding explicitly transitory changes (FA).

Values for CU and Fi are at mid-year; annual averages for all other variables. Sources: Fiscal provisions from Melliss and Richardson (1976), updated using budget statements; CU from CBI; all other series from standard ONS sources. References Bean, C.R. (1981) An econometric model of manufacturing investment in the UK, Economic Journal, 91, 106–21. Blundell, R.W., Bond, S.R. and Meghir, C. (1996) Econometric models of company investment, in Matyas, L. and Sevestre, P. (eds), The Econometrics of Panel Data (2nd edn), Kluwer Academic Publishers, Dordrecht. Bond, S.R., Denny, K. and Devereux, M.P. (1993) Capital allowances and the impact of corporation tax on investment in the UK, Fiscal Studies, 14(2), 1–14. CBI (1994) Realistic Returns: How Do Manufacturers Assess New Investment? CBI, London.

MICHAEL SUMNER

179

Cheung, Y.W. and Lai, K.S. (1993) Finite sample sizes of Johansen’s likelihood ratio tests for cointegration, Oxford Bulletin of Economics and Statistics, 55, 313–28. Feldstein, M.S. and Flemming, J. (1971) Tax policy, corporate saving and investment behaviour, Review of Economic Studies, 38, 415–34. King, M.A. (1972) Taxation and investment incentives in a vintage investment model, Journal of Public Economics, 1, 121–47. Lund, P.J. (1976) The econometric assessment of the impact of investment incentives, in Whiting, A. (ed.), The Economics of Industrial Subsidies, HMSO, London. Melliss, C.L. and Richardson, P.W. (1976) Value of investment incentives for manufacturing industry 1946–1974, in Whiting, A. (ed.), The Economics of Industrial Subsidies, HMSO, London. Robson, M.H. (1989) Measuring the cost of capital when taxes are changing with foresight, Journal of Public Economics, 40, 261–92. Summers, L.H. (1987) Investment incentives and the discounting of depreciation allowances, in Feldstein, M. (ed.), The Effects of Taxation on Capital Accumulation, Chicago University Press, Chicago. Sumner, M.T. (1992) Fiscal policy, seasonality, and intertemporal substitution of investment spending in the UK, Journal of Public Economics, 49, 123–34.

14 Investment policy and the employers’ perspective Kate Barker

Background One of the pieces of received wisdom which permeates a good deal of commentary on the UK economy is that the level of investment here is in some sense too low. This view has certainly been a major theme of CBI lobbying for many years, and the problem is believed to have several dimensions. For example, there is concern about the level of public capital investment, especially in infrastructure, as well as about the private business sector. We have in fact argued both that there is not enough investment, and that changes in government policy, in the areas of investment incentives and/or of saving, are necessary to resolve this problem. It would have to be admitted, however, that in the early 1990s these two propositions generally fell on stony ground in government circles. There were brief flurries of interest—for example Norman Lamont introduced a temporary increase in first-year capital allowances (effective between November 1992 and October 1993). And Stephen Dorrell, when Financial Secretary to the Treasury, launched a wide-ranging and rather useful debate in 1994, although under the uninspiring title of the Industrial Finance Initiative. Unfortunately, this became blighted by a wave of concern that the exercise presaged an attack on dividends, and the interesting questions which Mr Dorrell asked were not at that time answered. Further, it became clear in May 1996 that there was a lack of conviction, as far as the government of the day was concerned, that there was a UK investment problem. At this time, the Cabinet Office produced a paper arguing that business investment in the UK compared well with that of our major competitors. It was against this unfriendly policy stance that the CBI considered taking a fresh approach to the investment question in 1996. The issue remains relevant in 1998, when there is a greater interest shown by government, as a higher rate of investment is likely to prove part of the answer to the low productivity problem. The first step in our work, which has continued since 1996, was therefore to consider the evidence on UK investment. This question has of course been looked at many times, and from many angles. The evidence with regard to the numbers is covered very adequately elsewhere, and I do not want to repeat the data here. However, I will restate the main conclusions from our reworking of the figures. The UK’s capital stock is lower, per capita, than that of our main competitors, although capital productivity seems to be relatively high. This low level of capital reflects a poor investment performance over the last economic cycle. However, looked at relative to GDP, business investment has been comparable to that of other industrialised countries since the late 1980s. The situation is, therefore, not all bad. But it does leave open the possibility that encouraging a higher level of capital investment could boost the UK’s GDP growth rate, at least temporarily. This would certainly be worthwhile, if it enabled the closure (on a sustainable basis) of the gap in GDP per head which still persists between the UK on the one hand, and Germany and France on the other. Of course, raising our growth rate is not likely to be achieved solely by a policy focus on business investment. There are also concerns about the adequacy of public capital investment, and a clear role for improved skills as well, if the potential additions to the capital stock are to be fully exploited. This article looks at these issues only in the context of firms’ perceptions of factors restraining investment, rather than doing these large and important topics justice in their own right. The company perspective The CBI wanted to take a slightly different approach to this issue, adding value from the business standpoint, rather than the more theoretical. With the support of McKinsey consultants, we conducted a series of detailed interviews with companies about their investment practices, and how they approached major capital spend decisions. These were mostly held around mid-1996. Before embarking on these conversations, we established a set of hypotheses about what factors might restrain UK business investment, under three broad headings:

POLICY: THE EMPLOYERS’ PERSPECTIVE

181

• Structural factors (such as economic instability, or financial market pressures for short-term returns). • Factors internal to the firm, such as irrationally high hurdle rates on capital projects, or management incentive schemes which tend to penalise managers with high investment levels. • Factors due to the actions of government, such as the inadequacy of capital allowances, or investment grants; or the structure of UK taxation. In particular, some commentators believed that the tax structure existing in 1996 encouraged too high a level of dividends, at the expense of companies’ ability to finance new investment readily from their retained earnings. In the course of this study, we talked to a total of 30 companies in some detail. This at first sight sounds rather unimpressive, but since we tried hard to see more than one person in each firm, and to see them separately, this involved over 50 interviews, all of at least an hour’s length and many longer. In conducting the interviews, we sought as far as possible to ask questions in an open manner, rather than asking ‘leading questions’ based on our hypotheses. The CBI is very grateful to those who gave up their time so generously to help this work along. The conclusions are, of course, couched in general terms as the companies would not wish to see their views identified. In August–September 1998, the companies were contacted again by post, and their views sought in particular on whether any of their original views had changed over the two year period since the first interviews. Despite the major changes which have taken place (for example, the appreciation of sterling and Gordon Brown’s major reform of corporation tax), the original participants have in almost all cases been happy to support their earlier views. The interviewees were drawn initially from three broad sectors. First, automotive components, which is an industry involving a wide range of technology and skills (and which also raises interesting questions about the need to locate close to your customer). It is also true to say that this is a sector which has had to struggle against the background of the roller-coaster events over the course of the 1990s, both in the UK and in the wider European automotive market. The second sector was pharmaceuticals, selected because it is widely perceived as a UK success story and we hoped that it would offer a useful contrast. Like automotive components, companies compete on a global scale but products are even more important. Finally, came food retailing, selected to represent the service sector. Again, this was deliberately chosen as a sector of probable UK strength, rather than weakness. However, unfortunately this sector proved to be a difficult one from the viewpoint of this exercise. Competitive conditions between food retailers are so strong that very few companies were prepared to talk to us, and therefore it was even more difficult to draw clear and robust conclusions than proved to be the case in the other two sectors. The best sample in this study was from the automotive components sector, and this should be borne in mind when looking at the reported outcome. Before moving on to talk about the findings from the interview programme, I should mention that a second disappointment (though probably rather less of a surprise) was how few foreign companies were prepared to spend time discussing this project. Of the 30 firms canvassed, 10 were foreign-owned companies, but only 3 were foreign firms in the sense that the interviews referred entirely to experience and practice outside the UK. For this reason, only passing reference is made to the comments from these interviews. The other 7 were UK subsidiaries of foreign-owned companies—with either US or German parents— as we were sticking to the conventional approach in much of this kind of research, which compares the UK with these two countries. A very interesting follow-up of a project of this type might be to look at France, where institutions and policies have been different again. The outcome there has been an impressive performance in terms of economic stability, and manufacturing sector productivity. Structural factors The two structural factors which figured in our hypotheses were economic instability and short-term financial market pressures. Very little mention was made in the interviews of economic instability. This is perhaps surprising in view of the UK’s recent record. It is also rather at odds with the importance which companies responding to the CBI survey of manufacturing have accorded over many years to uncertainty about demand as a factor restraining investment. When reference was made to economic uncertainty in the UK, it was generally in a historic context, not as a factor for the mid-1990s. Indeed, most interviews started with a comment about how much the investment climate had improved in the UK over the previous 10 to 15 years. A perception that the UK financial markets had a short-term perspective, and were too focused on earnings (rather than looking at a wider range of financial variables) was much more common. However, a surprising remark made a number of times was about communication with the City. Firms often felt that their businesses were not fully understood, but some rather perversely indicated that communication with analysts was very time-consuming. Others (the majority) considered that investing time in communication with equity analysts brought long-term benefits.

182

KATE BARKER

Smaller firms expressed more concern about the impact of the financial markets. A number of interviewees seemed to be rather risk averse, although equally others complained about the shortage of venture capital funds. The UK’s historic economic instability seems to have led to worries about taking on too much risk on both sides of the market for risk finance. The other unexpected feature was that a number of companies stated that one main factor restraining their growth was a shortage of the right skills in the UK, at a relatively high level. This took different forms: in the automotive sector there was concern over the supply of growth-oriented managers, and individuals with the right combination of management and technical skills to carry projects forward. In the pharmaceutical sector, the concern is over the supply of scientists of the right quality. Any difficulties here could pose a significant risk, and be a reason for shifting the location of R&D facilities. In what seemed like a neat confirmation of Napoleon’s famous put-down of England, the food retailing sector did not share this concern. Factors internal to firms The argument that UK firms invest too little because they set their target, or hurdle rates for projects too high was given little support by the interviews. Since these were conducted, the CBI has also run a second postal survey on companies’ hurdle rates, updating a previous survey in 1994 (CBI, Target Practice, CBI and Association of Consulting Actuaries, London, July 1998). This was based on 326 respondents, across all sectors of industry. The main conclusion of the second survey was that there was little evidence that firms had actually reduced their target rates much in the mid-1990s, despite the changes in policy which were designed to provide a more stable economic environment. The comments below draw on both the outcome of the interviews and the results of the 1998 survey: • First, and perhaps most importantly, no firm said that their level of investment was constrained by potential projects failing to attain hurdle rates. Virtually all the interviewees explained that their hurdle rates would be set aside if a project had a compelling strategic rationale (which presumably means that the return would be higher if the cost of not undertaking the project could be adequately evaluated). This was broadly confirmed in the survey results, as only 16 per cent of firms said that they never undertook projects which fell below their hurdle rate. • Second, most companies talked about the need to choose among projects, even at the existing ‘high’ hurdle rates, due to other criteria which determined how much investment they wished to undertake. So, the hurdle rate may be set to prevent more projects coming forward that the board feel they can deal with–either in management or cash-flow terms, or to avoid having to raise the firm’s level of borrowing. The perception that it was not hurdle rates as such which held back capital expenditure was as true at the plant manager level as at the more senior level. On the contrary, one company had increased the hurdle rate to reduce the number of projects coming forward. Again, this conclusion also emerged from the survey results, in which the vast majority of firms reported that projects which attained the hurdle rates could still be rejected. • Third, and here this refers only to the automotive component sector, there was no indication that the foreign-owned or foreign-based companies used significantly different hurdle rates or pay-back periods. One or two companies talked about a different attitude to investment, however, with a greater preparedness in other countries to consider very long pay-backs for strategic projects. This was related to the perceived financial market pressure in the UK to produce short-term profits, and also to the UK’s greater economic instability. Privately owned UK companies, however, did not refer to these issues to such an extent. The 1998 survey results also showed a mixed picture. Of those respondents with international operations, 40 per cent used different criteria in different countries. The reasons most often cited for this were different inflation rates, interest rates, or political risk judgements. Management incentives were raised in most of the discussions. Most firms had recognised the risk of setting up an incentive scheme which encouraged a short-term approach, and had sought ways to avoid this. However, a point which was not among our prior considerations also emerged, which was that several firms have some kind of pre-set limit to their investment levels. A number of companies talked about investing over the longer-term the same amount, or a small multiple of, their annual depreciation. Others referred to a wish to keep growth at a steady pace, and it was notable that among the sample, the foreign-owned firms were more inclined to talk about growth as a specific target of corporate strategy. Government policy issues As indicated, the questions which were used sought to avoid suggesting any particular policy tool or taxation change to the interviewee. The most obvious area for comment is investment incentives, either through capital allowances or investment grants. It was noticeable that few companies picked out capital allowances spontaneously, while rather more talked about

POLICY: THE EMPLOYERS’ PERSPECTIVE

183

investment grants as a key factor in location decisions—though generally of a second-order nature. Many other factors are taken into account in picking a site, of which the location of customers is undoubtedly the most significant. The final issue of dividends and dividend taxation is clearly complex, going beyond the scope of an investigation based on an interview technique. However, quite a number of companies were at pains to stress that investment decisions and dividend decisions are not taken at the same time, and therefore the argument that pressure for high dividends restrains investment did not apply. But clearly the dividend decision does have an impact on the availability of internal finance for firms—although it was rarely mentioned as a specific constraining factor. Of course, since we met with these companies, the structure of corporation tax in the UK has been radically changed. In the survey on target rates, companies were asked if the abolition of ACT tax credits on dividends for the exempt funds, and the associated cut in the corporation tax rate, would affect the availability of investment funds. 90 per cent thought that the reform would make no difference, and a further 8 per cent that it would have an adverse impact. It will be some time before it is apparent whether or not this negative view of the reform is justified. (The Treasury’s own view was that the change should encourage a longer-term view, and therefore a better quality of investment decision—but as the new corporation tax regime will not be fully in place until the early 2000s, this too will not be confirmed or rejected for some years.) Policy implications What do these conclusions mean for policy? As has been mentioned, the UK has moved on since the survey was carried out. Not only has there been a major change in the tax regime, but there has also been a new focus for government policy—the improvement of productivity. In these circumstances, it is much more likely that policy-makers will be sympathetic to proposals which have as their goal an increase in the UK’s investment rate. There are three main messages which I would draw out as being of most relevance to the present policy debate. The first is that that there is very obviously no single compelling policy which is going to change the rate of business investment in the UK, and it is unlikely that small changes in the structure of taxation could do more than make marginal differences to investment decisions, particularly for larger companies. Significant shifts can, of course, have a greater impact. But big tax changes can bring difficulties of their own. They can be disruptive to businesses, produce unintended consequences, and in itself uncertainty about taxation policy is a negative factor when firms are putting together plans. In the context of the Labour government, this suggests that it may be wrong to be optimistic about what will result from the changes to the taxation system which have been implemented. Part of the reason for this caution is that firms themselves do not perceive the changes as having been beneficial, and therefore are not likely to look at their investment decisions in any very new way. There may, however, be some changes which would be of assistance to companies who have particular issues. One such group is high-tech businesses—the tax treatment of R&D is already under review, and on the evidence of our enquiries this is potentially useful. There is perhaps a point on investment grants. These can certainly change location decisions, once a project has the goahead in principle. And while the price of attracting investment can sometimes be questioned, clearly plants which attract a local supply base may have a high value. We were struck by how often the location of the customer drove automotive component location decisions. The second main conclusion is that, while the financing system in the UK is not perfect, especially with regard to growth companies (this project did not look in detail at the questions of finance for growth businesses, where the CBI has undertaken much separate work), the view that unduly high hurdle rates are being driven by the demands of investors has not so far been substantiated. The hurdle rate is not the sole criteria, or often even the prime one, in an investment decision. Even if it were, these interviews, and the CBI survey, both suggest that firms based outside the UK do not necessarily set lower target rates. However, quite a lot of frustration about communication between firms and their investors was manifested during our conversations. Efforts to improve information flows, in both directions, seem to be required. The third conclusion is not really surprising. It is that policy cannot seek to influence investment in isolation. The overall management of the economy has to permit a higher growth rate to develop, and this will require a wide set of supply-side policies. Labour shortages were mentioned at every level, even though at the time of the interviews the UK labour market was not especially tight. But the really significant shortfalls seemed to be technical and scientific skills, or for top flight management material. Increasing the economy’s capacity to grow is a slow process which requires several lines of coherent attack. The investment response cannot be adequate if these further conditions are not also met. Until the level of skills is capable of producing higher productivity, wages in the UK will continue to be relatively low, and there will be little incentive to install capital intensive production facilities. The CBI is always involved in studying the evidence on investment, and on productivity more widely. This interview project provided us with some fresh insights into how firms view the UK as a location, and what barriers there might be to doing business here. The conclusions suggest that low investment is one of a number of factors which seem to impede UK

184

KATE BARKER

growth. Action to raise the investment rate directly—better capital allowances, R&D tax incentives, or persuading firms to reduce hurdle rates are therefore unlikely to succeed unless other reforms are made at the same time.

15 The capacity to tackle unemployment Jonathan Michie

Since the late 1970s unemployment has risen dramatically in Europe. A key obstacle to the restoration of full employment is the erosion of industrial capacity which has taken place to the point where it is no longer adequate for employing the available workforce, even if the necessary demand for the output were there.1 Rebuilding that capacity is therefore essential if demand is to be allowed to increase sufficiently to bring levels of unemployment significantly below those prevailing in recent years. The policy implication for Government is that measures to stimulate investment have an important role to play in reducing unemployment. Moreover, measures to improve the quantity or quality of the labour force, or efficiency in the use of labour, may lead to higher unemployment unless they are accompanied by more investment in physical capital. Of course there is no automatic link between the size of the capital stock and the number of people employed. But the two are certainly related. If firms are close to full capacity working but are reluctant to invest in additional capacity, they will be disinclined to bid for new orders that they might not have the capacity to meet. In this case the output of the firm may turn out to be lower than it could have been, and likewise its employment. And this phenomenon may at times be quite general across an economy, resulting in lower domestic output and employment and a worse balance of trade. By causing capacity to be scrapped and deterring new investment, deflationary policies aimed at damping down wage demands may have the long-term effect of making the inflation and unemployment dilemma worse. The economy is left in a poorer position to absorb increased costs than would otherwise have been the case. Yes, it hurts. But no, it does not necessarily work.2 The problem is that Governments see unemployment as a labour market issue—caused by wage-fixing institutions, the role of welfare benefits or the quality and motivation of the workforce. Other potentially important issues, such as the impact of capital formation on employment, tend to be rather neglected. And it is true that economic orthodoxy currently holds that investment has little or no long-run effect on employment. The problem of job-creation is seen primarily as a matter of encouraging more employment on whatever capital stock happens to exist at the time. Thus in the highly influential work of Layard, Nickell and Jackman (1991), investment in new capital stock generates higher wages which lead to a loss of employment on existing equipment which exactly offsets the extra jobs created on new equipment. Hence no net increase in employment. But this result assumes a substitutability between capital and labour far greater than that which exists in reality, requiring that a 1 per cent increase in real wages reduces employment by at least 2.5 per cent and probably more. Yet even using their own econometric estimates, it was found by Rowthorn (1996) that any such reduction turns out to be well under 1 per cent. When this unrealistic assumption is dropped none of the major conclusions of the mainstream literature on unemployment survive. Instead, we are faced with the real world where increased investment can create faster growth of output and employment. And expectations about future demand and profitability are central to firms’ decisions about expanding capacity.3 Firms need to be confident that additional production will be profitable and that demand will grow at such a rate as to justify any expansion of their capacity. But experience in recent years has made managers cautious on both counts. Thus, using the CBI Industrial Trends Survey, Driver (1996) demonstrates that since the 1970s firms have become more willing to risk losing orders due to having insufficient capacity. Using the required rate of return in relation to the cost of capital, there is evidence of growing caution about the risks of installing additional capacity. Means must be found to tackle this. In particular the operation of corporate taxation needs to be reconsidered. The elimination of free depreciation and the reduction of corporate tax rates have adversely altered the balance of risk between profiting from expansion and losing from over-capacity. The key to sustainable growth and a return to full employment is to implement these sort of supply-side policies alongside macroeconomic measures to ensure a continuous expansion of profitable production opportunities. This requires sustained low interest rates and a competitive exchange rate—a mixture the UK had thankfully stumbled into in the mid-1990s. It was always likely that one or either of these favourable circumstances would be sacrificed as the price of entry into a European single currency. Ironically, though, it was actually the election of a Labour Government which put paid to both—handing interest rate, and therefore exchange rate policy over to an unaccountable Bank of England that promptly raised interest rates repeatedly, causing sterling to become overvalued, and pushing the economy toward recession.

186

JONATHAN MICHIE

It is therefore vital that the Government’s supply-side policies on competitiveness and Welfare to Work are combined with a renewed commitment to a balanced policy agenda, with fiscal, monetary and exchange rate policy all being used to target sustained economic growth, alongside policies to tackle poverty, inequality, and insecurity.4 Unfortunately, the abandonment by Government of monetary and exchange rate policy has been accompanied by an ultraorthodox approach to competitiveness. This was exemplified by the use of taxpayers’ money to commission a report from McKinsey consultants, the results of which were as unhelpful as they were predictable. After almost twenty years of being subjected to deregulatory policies, the British economy was diagnosed as over-regulated. The proof, according to McKinsey’s was that almost half of Britain’s hotels are almost 100 years old, as against only 3 per cent in the US. The fact that few of America’s hotels were operating in the nineteenth century, though, has less to do with regulation and competitiveness than it has to do with that country’s history. Likewise, the fact that a large number of Britain’s hotels are more than 100 years old is not caused by over-regulation, nor is it a particular problem for Britain’s tourist industry or for the country’s general economic competitiveness.5 As suggested above, we would argue that the main reason for lower productivity in Britain is, rather, inadequate investment. This in turn is caused largely by the short-termism that is fostered by a number of factors. First, the City of London, and more generally the relationship between Britain’s industrial and service sector on the one hand, and the financial sector on the other, tends to create high hurdle rates for investors, a much higher use of overdraft facilities that are recallable at any time, as against long-term loans, than is the case in most other countries, and excessive takeover activity, threatening the position of any managers who do invest long-term in ways that are not reflected immediately in share prices.6 Second, UK company law, amongst other things facilitates the problem of excessive takeover activity, for example by obliging company directors to recommend any takeover bid to shareholders if it represents a good financial offer relative to the share price, regardless of what long-term damage the directors suspect that the predator might do to the business, or that the takeover might do to the economy.7 And third, the economic instability that, ironically, has often been exacerbated by Government attempts to target nominal variables such as the money supply or the exchange rate, has allowed the real economy to suffer as a result.8 Flexibility Alongside deregulation, the other buzzword of supply-side policy since the late 1970s has been ‘flexibility’, and again there has been a disappointing failure to break with the past by the Blair Labour Government. Disappointing because a much more nuanced approach to the issue is required if it is to be usefully used as a policy goal. Instead the impression has been given that it you want to tackle unemployment you can forget about lack of demand, inadequate investment levels, high interest rates or an overvalued currency —as if they are the old answers, and New Labour wants new ones. If you want to be at the cutting edge, think flexibility. Who could possibly object? Who wants to be inflexible? This new agenda sounds compelling provided no one asks what is actually meant by ‘flexibility’. Does it mean training the workforce so they can undertake a variety of tasks? And allowing employees to arrange their hours of work around childcare and other responsibilities? Or does it mean undermining the protection offered by maximum hours legislation, by allowing an employer to agree with the individual security guard that they will carry on working 80 hours a week after all? And encouraging a hire-and-fire mentality which increases labour turnover and makes it that much more likely that a worker you train today will be working for a rival firm two years down the line, thus making it less likely that firms will invest in such training? The term ‘flexibility’ is often used but rarely defined. It is used to cover a huge range of policies and practices—including all the examples above and many more, some of which may improve working conditions and may also encourage firms to take the high road of upgrading skills and work practices, and investing in new product and process innovations. But many of which most certainly do not. So where does this leave us? Is the Labour Government right to be trumpeting the virtues of having a flexible labour market, or are they carrying on up a Thatcherite blind alley? The first point is simply to make the distinction between the different types of policies and practices that are encompassed within the overall term ‘flexibility’. Some are positive and some negative. Recognising this would be a step forward. The next step would be to consider whether ‘positive’ and ‘negative’ flexibility has the same or different effects on corporate performance and economic outcomes. Because even though negative flexibility may be bad for the wellbeing of employees, perhaps it’s necessary to compete internationally? This next step has been attempted by Michie and Sheehan (1999), that reports research into what effect different forms of flexible practices actually have in reality. First, we grouped together the sort of flexible work practices that would be generally welcomed by workers and trade unions—using teamwork, having an implicit employment security pledge, having increased job assignment flexibility, regular information sharing between workers and management, and so on. We found that the use of such work practices has a positive

THE CAPACITY TO TACKLE UNEMPLOYMENT

187

and significant effect on the probability of firms innovating. For example, firms that incorporate at least one of the above practices are 40 per cent more likely to innovate compared to firms that use no innovative work practices. We then looked at the alternative approach to flexibility, of using seasonal, temporary, casual, fixed-term and part-time employees. The use of this sort of ‘flexibility’ was found to have a significantly negative effect on the probability of the firm innovating.9 Labour was elected on a platform emphasising the need for increased education, training, R&D and innovation. The latest research suggests that this is consistent with certain types of ‘flexibility’ but not others. Greater clarity is thus called for—in policy debate and legislative practice. Conclusion The chapter by Barker (14) argues that there is ‘a clear role for improved skills…if the potential additions to the capital stock are to be fully exploited’. We would concur, and add that key to improving skills is to take the high road variant of labour flexibility discussed above, rather than the low road, hire-and-fire version of flexibility that actually inhibits corporate investment in improved skills. A similar point could be made about Barker’s support for favourable tax treatment of R&D (where she reports that ‘on the evidence of our enquiries this is potentially useful’).10 Again, increased labour turnover will reduce the payback that companies get from investment in R&D and innovation, as much of the benefit accrues over time through the tacit knowledge acquired by the firm’s workers involved in and working with the new processes. There are thus a range of supply-side policies that could and should be pursued, some of which have been discussed above, and others of which (such as investment in public infrastructure, and reforming corporate governance) are discussed elsewhere in this volume.11 But as discussed above, this needs to be combined with active macro-economic policies to ensure overall coherence. Welfare to Work requires work to be made available, not just the withdrawal of welfare. Interest rate policy needs to take account of its effect on investment levels—not only through the cost of capital but also through its effects on consumer demand, on which investment decisions depend. And so on. The point here is not to repeat the various arguments, but rather to stress, in conclusion, that investment levels are indeed crucially important for economic growth and living standards, and that to foster high and growing levels of investment requires demand and supply policies to be actively pursued in a coherent and conscious fashion. Notes 1 For the evidence of UK underinvestment and consequently low capital stock see Brinkley and Soteri (this volume) and, in more detail, Kitson and Michie (1996). 2 This is argued in detail by Michie and Wilkinson (1995). 3 Thus in Barker’s discussion of investment incentives and capital allowances being generally of a second-order nature for companies, she reports that this is due to many other factors being taken into account in picking a site, ‘of which the location of customers is undoubtedly the most significant’ (Barker, this volume). 4 Watson and Hay 1998) provide an excellent discussion of the problem—‘for problem it is’ (p. 408)—of productive investment in Britain, and the need within this context for the Labour Government to ‘transcend the pervasive neoliberal paradigm that has come to circumscribe the parameters of what is considered politically and economically feasible, possible and desirable in Britain as elsewhere’. (p. 408). 5 Even The Economist, commenting on the McKinsey Report, asked ‘does it really matter that much to the economy as a whole if British hotels are smaller and older than American ones?’ (31 October 1998). 6 See for example Hutton (1995) and Lee (1996). 7 This was certainly the legal advice given to the Director of Manchester United plc, Greg Dyke, to the effect that he was legally obliged to recommend the 1998 BSkyB bid to take over Manchester United to shareholders of Manchester United, regardless of whether he personally thought that the takeover would be good for Manchester United (as reported by Mr Dyke to the November 1998 Annual General Meeting of Manchester United shareholders). 8 According to Adair Turner, Director General of Britain’s Confederation of British Industry (CBI), the main reason for the lower productivity in the UK ‘is that the sharp ups and downs of the British economy have made managers more risk-averse than bosses in other countries’ (cited in The Economist, 31 October 1998, p. 40). This is very much in line with what is argued in Driver and Michie (1998). 9 Our results also indicated that trade union recognition has a positive and significant effect on the probability of firms innovating. Perhaps the government should bear this in mind in the face of those companies who lobby against trade union recognition. What have they got to hide, other than their own bad practices? 10 Barker, this volume.

188

JONATHAN MICHIE

11 On public investment, Barker (this volume) refers to the ‘concerns about the adequacy of public capital investment’ and Brinkley and Soteri (this volume) advocate public investment as one of their key areas for action. For a detailed and costed case for boosting public investment as part of an expansionary policy to tackle unemployment, see Kitson et al. (1997). Brinkley and Soteri (this volume) also advocate a new corporate governance framework as a key area for action.

References Driver, C. (1996), ‘Tightening the Reins: The Capacity Stance of UK Manufacturing Firms 1976–1995’, in J.Michie and J.Grieve Smith (eds), Creating Industrial Capadty: Towards Full Employment, Oxford: Oxford University Press. Driver, C. and Michie, J. (1998), ‘Managerial Culture and the Capacity Stance of Firms’, ESRC Centre for Research on Innovation and Competitiveness Working Paper, University of Manchester. Hutton, W. (1995), The State We’re In, London: Jonathan Cape. Kitson, M. and Michie, J. (1996), ‘Britain’s Industrial Performance Since 1960: Underinvestment and Relative Decline’, The Economic Journal, Volume 106, Number 434 January), pp. 196–212. Kitson, M., Michie, J. and Sutherland, H. (1997), ‘The Fiscal and Distributional Implications of Job Generation’, Cambridge Journal of Economics, Volume 21, Number 1 (January), pp. 103–120. Layard, R., Nickell, S.J. and Jackman, R. (1991), Unemployment: Macroeconomic Performance and the Labour Market, Oxford: Oxford University Press. Lee, S. (1996), ‘Finance for Industry’, in J.Michie and J.Grieve Smith (eds), Creating Industrial Capacity: Towards Full Employment, Oxford: Oxford University Press. McKinsey Consultants (1998), Driving Productivity and Growth in the UK Economy, October, Report produced for HM Treasury, London. Michie, J. and Sheehan, M. (1999), ‘HRM Practices, R&D Expenditure and Innovative Investment’, Industrial & Corporate Change, forthcoming. Michie, J. and Wilkinson, F. (1995), ‘Wages, Government Policy and Unemployment’, Review of Political Economy, Volume 7, Number 2 (April), pp. 33–149. Rowthorn, R. (1996), ‘Unemployment, Wage Bargaining and Capital-Labour Substitution’, ESRC Centre for Business Research Working Paper, WP38, Cambridge. Watson, M. and Hay, C. (1998), ‘In the Dedicated Pursuit of Dedicated Capital: Restoring an Indigenous Investment Ethic to British Capitalism’, New Political Economy, Volume 3, Number 3 (November), pp. 407–426.

16 The UK’s investment problem Ian Brinkley and Soterios Soteri

Introduction One of the most persistent worries about the long term performance of the British economy has been under-investment in both physical and human capital. In the 1980s and 1990s market based solutions were tried. These failed to raise the investment share, the long term growth rate, or significantly close the industrial productivity gap with countries like France and Germany. As the economy enters the new century, concerns about investment in capacity and people have returned to dominate the policy agenda. The present investment share, according to the Treasury, is barely able to sustain the assumed long run growth rate. It is possible to disagree with the Treasury about the long term growth potential of the economy. It is also true that improving capital productivity can for a time offset a low investment share. But it is hard to disagree with the Treasury’s conclusion in the March 1998 Red Book: If the UK is to raise its trend growth rate and catch up with other major countries in terms of GDP per head, then it seems almost certain that some considerable increase in the investment—GDP ratio will be required. In the UK and EU where the share of investment to GDP has declined more sharply since the 1960s, economic growth has halved. A key concern for the UK in 1998 is that the economic slowdown which is taking place will rapidly erode the benefits of the investment growth experienced since 1994. The traditional focus on physical investment has been supplemented with growing attention to investment in people. Worries about investment in the skills of British workers has a very long history. However, many of the central policy debates in the 1980s and 1990s were dominated by the view that investment would be supplied as the returns to individuals from education and skills rose. At the same time ‘flexibility’ was hijacked to mean widening out the wage distribution, weakening collective bargaining and creating ‘hire and fire’ labour markets. In the policy debates today employability has replaced flexibility, not just in name but because of a renewed concern about both the levels and quality of national investment in both education and training. Many have seen investment in people taking over as one of the key determinants of whether the UK economy can perform well in the next century. The shift in labour markets across the industrialised world towards high skill white collar employment certainly suggests there must be some truth in this. How much truth is still a matter of debate—not least because statistics are in short supply and because the links between overall growth and job performance and investment in education and training are not straightforward. Improving employability is clearly not a panacea for economic success and high employment. Equally, an economic growth and employment strategy that did not include a strong commitment to human capital investment would hardly be credible. The investment question is not just on the UK’s economic policy agenda. The investment share has declined through much of Europe, limiting the long term growth potential and reducing the rate of employment creation across much of the European Union. Making capacity expanding investment and creating high skill workforces is a priority within the EU’s Broad Economic Guidelines and the Amsterdam Treaty. The article looks at the investment record and some of the key alleged causes and suggests a policy agenda for tackling the investment question into the next century. The investment record The UK’s investment share has been in decline for many years. In the 1960s and 1970s the economy invested between 18 and 19 per cent of GDP. In the 1980s this declined to 17.5 per cent, and again in the 1990s to 16 per cent. Despite a longer and

190

THE UK’S INVESTMENT PROBLEM

Figure 16.1 Private sector and business investment (percentage of GDP using 1990 constant prices). Source: ONS.

stronger economic recovery in the 1990s, latest figures show the UK remains rooted towards the bottom of the European investment league. Between 1990 and 1997 investment hardly grew at all (see Table 16.1). Table 16.1 Investment shares compared, 1960–1997 UK

EU

US

1960–97 17.8 21.5 18.4 1960–69 18 23 18.3 1970–79 19.2 23 19.1 1980–89 17.5 20.2 19 1990–97 16.1 19.1 16.8 Sources: European Commission Annual Economic Report 1997 and OECD main economic indicators.

Japan 31 31.6 33.1 29.1 29.8

Since the 1960s there has been a general reduction in the advanced economies’ investment to GDP ratios, but this should not be of any consolation to the UK. GDP per head is lower in the UK and this has translated into a consistently lower level of investment per worker. The cumulative effect of lower UK investment per head has resulted in UK employees having a lower level of productive capital at their disposal. Capital stock estimates are difficult to estimate—but estimates that do exist show the level of the UK’s capital stock per worker are so much lower than other major countries–50 per cent lower according to one source quoted by the government1—that one cannot avoid the conclusion that UK employees have a considerably lower level of capital at their disposal. Some comfort might be had from the faster rates of business growth and the rise in the business share in GDP in the 1980s, when private investment growth was very strong. But there are two important caveats. First, investment responded to the unsustainable boom and in many cases may have over-invested in assets which stood idle when the boom turned to bust. Indeed it is possible that some of the investment famine of the 1990s has its counter-part in the investment boom of the 1980s (see Figure 16.1). Second, there was a major transfer of assets of corporations and other bodies from the public to the private sector. It is probably impossible to say what business investment would have been without these policies, but if privatisation was supposed to help raise total investment in the economy then it failed. Investment weakness in the 1990s is not simply a matter of over-indulgence in an earlier economic cycle or selling off the family silver. The investment question has a strong sectoral dimension, with weaknesses in manufacturing and the public sector and a better performance in the service sector. One of the key reasons for the UK’s low capital stock is the slow rate at which the UK manufacturing sector has been investing relative to the service sector. Indeed, whilst the manufacturing sector’s gross capital stock has grown by 9 per cent since the 1980s the gross capital stock of the service sector has increased by 55 per cent (Figure 16.2). Even more disconcerting, the UK’s manufacturing stock which has grown at about 1 per cent per year since the 1970s compares poorly with other advanced industrial countries. During the 1980s and early 1990s, for which broadly comparable data are available from the OECD, the

IAN BRINKLEY AND SOTERIOS SOTERI

191

Figure 16.2 Gross capital stock (£bn at 1990 prices). Source: ONS.

Figure 16.3 Public investment as share of national income (net public sector investment as percentage of GDP). Sources: HMT July 1998; TUC.

manufacturing stock in the UK was growing at half the rate of the US. The gap with the rest of the EU was nearly as large and that with Japan was bigger still.2 Cuts in public investment go far deeper than simply transfers to the private sector, particularly once the steam ran out of the privatisation programme. Public sector capital investment fell by 44 per cent between 1994 and 1997, a fall roughly equivalent to 1 per cent of GDP (Figure 16.3). Recent Treasury figures show net public investment fell from around 3 per cent of GDP at the end of the 1970s to less than 1 per cent by the end of the 1980s. Figures show net public investment down to 0.8 per cent of GDP. However, it is not all gloom and doom. In contrast to manufacturing, service sector investment has been much stronger, with the service sector capital stock growing by 55 per cent since the 1980s. Services are also internationally traded and with two-thirds of non-dwelling investment accounted for by the service sector no policy can afford to ignore what is happening in the service areas of the economy. Indeed, the strong service investment performance may well have helped hold price inflation down in the present recovery by expanding capacity and adapting new technologies. An article by James Nixon in the London Business School Economic Outlook in February 1998 makes the point that the sustained rate of investment growth since 1994 helps explain the better UK inflation performance over the 1990s economic recovery, despite the significant fall in unemployment.

192

THE UK’S INVESTMENT PROBLEM

Profitability, investment and the wage share Poor investment performance has in the past been linked to poor profitability. The economic debate, especially in Europe, has long featured calls for wage moderation, so that real wages running below productivity growth could be translated into a higher investment share in the economy, allowing faster growth and higher employment. This argument is an important one which should be taken seriously. However, as the TUC report for the London G8 Trade Union Summit in April 1998, Productivity and Social Partnership, showed, the experience of the 1990s also suggests it must also be treated with caution. In the long run, real wage growth and productivity growth have broadly moved in line across most industrialised countries. However, in the 1990s this changed—real wages grew more slowly than productivity, whether by social partnership design in search of economic success or simply by changing the balance of bargaining power in the labour market. Productivity grew nearly twice as fast across the EU as did real wages—in the UK productivity grew by about 2.2 per cent, real wages by 1.2 per cent. This seems to have been primarily a European experience—wages and productivity grew in line in Japan and the US —although in the latter real wage growth was largely sustained by longer working hours (see Table 16.2). The emergence of a real wage–profitability gap naturally increased both profitability and the profit share within the European economies, but few economies have seen such a marked change as the UK. International comparisons of profitability are fraught with problems, and both the EU and the OECD issue strong health warnings about comparing levels rather than trends. However, the trend is clear enough. Comparing 1990 and 1997 rates of return in the UK increased by 9 percentage points to reach a 30 year high. This compares with rises of less than 2 percentage points in Germany and 0.5 per cent in France. Indeed, the UK outpaced the US, where profitability also increased strongly, by 5 percentage points over the same period. The fall in the wage share between 1990 and 1997 was more marked in the UK than in the EU, but what is startling is that the UK investment share declined to a greater extent. The percentage changes tell their own story: comparing 1990 and 1997 the investment share fell 18 per cent, the wage share fell by 6 per cent, and profitability increased by 37 per cent. The investment share also fell in Germany, Italy, and France—but not in the US, where the investment share increased (see Table 16.3). Table 16.2 Productivity and real wage growth (annual percentage change) in the 1990s US

Japan

EU15

UK

Productivity growth 1.1 1.0 2.0 2.2 Real wage growth 1.3 1.0 1.1 1.2 Source: EU Commission. Note productivity is GDP per person, real wage is compensation per employee including employer social security, using GDP deflator. Table 16.3 Wages, investment and profits, 1990–1997 Wage share (1990=100)

Investment share (1990=100)

US 101 105 Japan 101 92 EU15 96 88 UK 94 82 Source: Productivity & Social Partnership: TUC 1998.

Profitability (1990=100) 121 84 109 137

Growth and instability The comparative reluctance for business to invest in the 1990s has puzzled many observers, given that financial incentives to invest have been highly favourable for several years. In many European countries, the underlying reason may simply be that in the face of low domestic demand and poor growth, new investment is not needed. In the 1990s economic growth in the US has averaged 2.6 per cent, in the EU 1.4 per cent with even slower growth in countries like Italy. This is well below the performance of the 1980s, hardly a decade of high growth itself. The UK’s earlier emergence from recession and stronger growth performance has pushed up the average performance to 1.7 per cent. Not only has growth been low it has been uncertain (Figure 16.4). European economies have been subject to a stop-start recovery, with— according to the EU Commission—the 1993 recovery brought to a halt in 1995–96 by international currency unrest and monetary tightening (not to mention excessive fiscal tightening in many European economies). The European recovery has now restarted, the European Commission has been putting out some optimistic forecasts—but nothing is guaranteed.

IAN BRINKLEY AND SOTERIOS SOTERI

193

Figure 16.4 Investment over two recoveries, 1981–1992=100 (1990 prices). Source: ONS.

However, the UK has proved itself even more unstable than the rest of Europe. The OECD’s recent survey of the UK economy found that of all the G7 economies the UK had the highest or second highest degree of variability in GDP inflation, real short term interest rates and the real exchange rate between 1980 and 1997. Investment in the UK falls more sharply during economic downturns in comparison to other countries because the UK has experienced deeper recessions than elsewhere. For example, during the economic downturn in the early 1990s, the cumulative fall in output was 2.5 per cent in the UK, 0.6 per cent in the US, 1.2 per cent in Germany and 1.5 per cent in France. At times when economies start to experience an acute slowdown, this creates a greater degree of uncertainty over the future level of demand for firms products and acts a major disincentive for firms to undertake investment expenditure which is to be recouped by future sales. In response, firms in the UK, typically, tend to reduce investment sharply and this creates a further negative expenditure shock to the economy which leads to lower output growth. The reaction of investment expenditure to lower economic growth compounds the initial slowdown and creates a downward spiral with lower output and investment feeding off each other. Even without the OECD’s confirmation, many businesses with direct experience of the roller-coaster UK economy will have become cautious about investing until they can see the recovery is durable. Not only does this encourage caution, but it may be a factor in the tendency for UK business to set notoriously high rates of return for new projects compared with other major industrialised countries. Those in the manufacturing sector have experienced just about every variation on the theme of exchange rate policy, including benign neglect, and have faced chronic instability. The most recent instalment has been the exchange rate’s effective appreciation of 25 per cent since the summer of 1996, with movements up or down subject to the markets’ guessing game with the Bank’s Monetary Policy Committee about future interest rates. Economic instability has played a part in deterring long term physical investment, but the creation of ‘hire and fire’ labour markets has played a role in deterring investment in human capital. Work by the OECD certainly suggests that employment in the UK responds more quickly to changes in output than it did in the past or is the case in other European countries. The hand of the flexible labour market can also be seen in the shortages in the IT industry, notorious for encouraging short term contracts, subcontracting and freelancing and the lack of job security when demand was relatively low. The economic price is of course paid when demand expands and those with the skills take the chance to make as much as possible in the short term knowing they have no long term job security. Similarly in the construction industry shortages of skilled workers regularly occur as workload instability and poor pay and job insecurity makes established workers leave and deters new entrants. But other institutional and structural changes have caused under-investment in people: • One is the decline in collective bargaining. The high-productivity high-involvement economies of Europe also have very high levels of collective bargaining coverage and stronger social partnership arrangements. Research studies show that the chances of getting training are higher in recognised companies and higher controlling for other characteristics—if someone is a union member.

194

THE UK’S INVESTMENT PROBLEM

• Another is the development of the so-called ‘flexible’ workforce. Although the extent of structural change in the labour market is often over-stated, the shift to part-time work, the expansion in self-employment in the 1980s and the growth of temporary and other insecure work forms in the 1990s have done nothing to encourage more investment in the workforce. Nor has the encouragement of the ‘bargain basement’ labour market: the low paid are another major group who miss out on training. In the debate about the alleged economic benefits of flexible workforces, the very substantial economic costs also need to be taken into account, and under-investment in people is very clearly one of them. Corporate culture and short-termism As well as the instability in the macro-economic climate, the UK may well have had to contend with a corporate culture and financial market structure that encourages dividend payments over long term investment. The productivity-wage gap did not translate into higher investment or reduced working hours but into the pockets of shareholders. Dividend payments remained high even in the recession years, and in the recovery dividend payments have risen nearly twice as strongly as investment. Between 1992 and 1997 dividend payments by industrial and commercial companies have risen by 67 per cent compared with investment of 27 per cent (both figures in current prices). The way our major corporations are governed has a critical influence over the investment decisions they make. Directors of public companies are, by and large, only accountable to their shareholders. Few individual shareholders take an active interest in company strategy and even fewer have enough influence to challenge the board of director decisions. Institutional investors, who dominate the market and do have influence, seldom exercise it. As long as dividend payments and share values are kept up, directors face comparatively little effective accountability. Indeed, many directors have a vested interest in keeping share price healthy because share options and holdings play such a major part of their renumeration packages. Moreover, the constant threat of takeover activity provides an added incentive to maintain high share values—both to defend hostile bids or to ensure a good price in the face of uncontested or successful bids. High take-over and merger rates have been defended almost on Darwinian principles, with only the economically fittest surviving. Sadly studies of the longer term effects of takeovers can find little improvement in economic performance. Tackling the investment shortfall The persistent problem of under-investment in both physical and human capital in the UK economy will not be resolved in the short term: a significant increase in the investment to GDP ratio will have to reverse a downward trend established well over twenty years ago. However, nor is it an impossible task. Moreover, it is not a task for Government alone: the social partners must also address the issue. There is no single solution to the problem: what is required is a co-ordinated approach across a number of policy areas. Some of the key areas for action are set out below. Growth Growth remains an important determinant for investment and employment. All the incentives in the world will not encourage business to invest if it thinks there is no market for the extra capacity. The priority in Europe is clearly to sustain the recovery, and for the UK to stabilise the economic growth rate and steer the economy away from a damaging recession. The key trick for policy makers will be to make sure the economy grows sufficiently strongly to encourage investment without setting off the inflationary boom of the late 1980s. Stability The Blair Government has strongly emphasised stability; for example, the March 1998 Red Book said: ‘Economic stability is vital if the Government is to achieve its central objective of high and stable levels of growth and employment. Instability in the recent past has contributed to the UK’s poor growth performance, not least by holding back the long-term investment that is the foundation for a successful economy.’3 A higher degree of economic stability would certainly encourage more investment to take place than would otherwise be the case but stability must be around high rates of growth. Moreover, restoring stability in the labour market to encourage investment in people is equally important. Many recent labour market reforms— the Fairness at Work proposals, the Minimum Wage, the implementation of the working time Directive—will all help. New Deal and other measures to strengthen the training base are also welcome developments. But equally it cannot be left to Government. There is a strong responsibility on the social partners in the workplace to develop and strengthen the existing collective bargaining agenda on training and skills development.

IAN BRINKLEY AND SOTERIOS SOTERI

195

Exchange rate policy If the low manufacturing investment performance of the UK is to be addressed in a meaningful manner, then an explicit exchange rate policy will be needed. As pointed out above, exchange rate instability is not only a major problem for manufacturing, now being pushed into yet another recession by the high pound, it has been the Achilles’ heel of the UK economy. The sooner the government makes an explicit effort to join economic and monetary union (EMU) the better. A sustainable target rate for entry into EMU should be set between DM2.50–2.60 and the Bank of England should be required take this into account when setting interest rates. Active fiscal policy The TUC shares the Government’s aim of medium term economic stability, but this also requires economic stability in the short term—otherwise the goal of medium term stability will be forever just out of reach. Monetary policy alone is not enough: active short term fiscal policy will be required. There is some sign the Government may be moving in this direction, with the idea of shorter term ‘constrained discretion’ in addition to medium term policy rules and plans. The TUC has suggested in a policy statement, Economic Policy and Social Partnership, that the adoption of an active counter-cyclical fiscal policy which acts in the initial stages of an economic slowdown will limit the degree to which output falls and help break the downward spiral of lower output expectations and declining investment expenditure and job losses. Public investment The decline in public investment has been a key cause of the low share of national income going to investment in the UK. The government’s plan to double the share of public investment in GDP over the next three years is a very welcome first step. But it also has to be recognised that this is from a small base, and even a doubling leaves net public investment as a share of GDP at roughly the level of the mid-1990s and far behind the rest of Europe. Reliance on asset sales and privatisation to fund the programme and the development of genuine public-private partnerships is one potential way forward, although none provide a ‘free lunch’ and there are also disadvantages as well as potential gains. The TUC believes there is scope to give public corporations more freedom to borrow form the markets for worthwhile projects. The slight easing signalled by the Government in the 1998 Fiscal and Economic Strategy report for existing corporations is a step in the right direction, but must be built on and extended to new institutions, for example, local housing corporations. Corporate governance The TUC would recommend that a new corporate governance framework be introduced and based on a more open and transparent system with partnership and dialogue between company directors, shareholders and employees acting as its core foundations. The framework should, amongst other factors, include: • a legal requirement for institutional shareholders to vote their shares at companies’ annual general meetings; • the expansion of employee share ownership schemes which will provide workers who depend on the company for their livelihood, and will naturally take a long-term view, a voice to express their concerns; • an expansion of the suggestions of the Cadbury and Greenbury committees’ recommendation to extend the role of nonexecutive directors as bringing independent voices to the board, by including employee nominated representatives. Social partnership The TUC is healthily sceptical of the claim that unemployment cannot fall below a pre-determined limit without inflationary pressures taking off, but equally there must be some point where further falls in the unemployment rate cannot be readily achieved by demand and investment expansion alone. In order to achieve full employment the question of wage inflation alongside investment, productivity and labour market reform must be addressed. As set out in the 1997 policy statement Partners for Progress, the TUC acknowledged that the collective actions of the social partners was a critical element in determining how far the goals of real wage increases, low inflation and high employment and investment can be met. The UK is unusual in Europe in having no mechanism through which the social partners can discuss the macro-economic consequences of their actions. Indeed, most other European countries regard social partnership as essential in tacking the enormous economic, technological and social changes affecting all the advanced industrialised economies and in meeting the growth and employment objectives of the Amsterdam Treaty.

196

THE UK’S INVESTMENT PROBLEM

Conclusion The UK has suffered from continuously low levels of investment since at least the 1960s and this has resulted in the UK possessing a lower level of capital stock per person than other advanced industrial countries. The capital stock gap which exists is a key cause of the UK’s lower GDP per head in comparison to other countries like France and Germany and the UK needs to raise the share of national income going to investment if this gap is to be narrowed. The reversal of the UK’s declining investment to GDP ratio will not be an easy task to accomplish, but neither is it an impossible one. The government can do much to help reverse the UK’s poor investment performance, not least by raising the declining level of public sector investment, but to raise their chances of success a co-ordinated approach which involves the social partners across a number of policy issues is required. This chapter has put forward a number of measures to help tackle the UK’s low level of investment, but recognises that there is no short-term panacea to rectifying a problem which is endemic in the UK. A long-term strategy focusing on the need to achieve a more stable economic environment alongside changes to the corporate governance system are essential if the UK is to raise its long-term investment performance in physical and human capital. However, it should not be forgotten that the long term is by definition a continuous stream of short-term time horizons and therefore what happens in the present affects where we end up in the long term. If the government wishes to create a more stable economy and encourage long-term investment it must therefore also respond to current economic pressures. A more active fiscal policy geared to steering the economy away from—or at least limiting the extent of—oncoming recessions is crucial if long-term investment behaviour is to be encouraged. For the manufacturing sector, the need to dampen the highly volatile behaviour of the UK’s exchange rate over the short term is also of high importance to encouraging manufacturing investment to come forward. The adoption of a more active fiscal policy and a commitment to dampen exchange rate fluctuations to achieve long-term stability would be complemented by dialogue with the social partners. The provision of a mechanism through which the social partners can discuss the needs of the economy with the government would allow a more co-ordinated macro-economic response to short-term pressures which would help limit inflationary pressures building up and preventing growth which is so vital to encouraging long-term investment to come forward. Notes 1 See Financial Statement and Budget Report 1998, p. 57, HM Treasury. 2 See ‘Manufacturing: time to take (capital) stock’, Andrew Glyn, Financial Times 08/07/98. 3 See page 20, para. 2.01.

17 Investment and capital productivity in Europe and the US Jaewoo Lee1

Introduction Since the 1970s, an intriguing contrast arose in unemployments between the US and European countries. Unemployment rates have hovered around two totally different levels across the Atlantic (Bean 1994). From the viewpoint of applied research and policy, such differences among countries are both intellectually demanding and practically important. They challenge us to revisit theoretical premises that we otherwise take for granted. In this chapter, I discuss another transatlantic contrast, in terms of the physical investment. Table 17.1 shows the average growth rate in the gross fixed investment for the US and four European countries, for periods 1975–95 and 1980–95. For both sample periods, the growth rates of investment in the US are much higher than those in the UK, France and Italy. The US rates are higher also than the rates of Germany, though by a smaller margin. Since the mid-1970s, investment in European countries seems to have stagnated, relative to the US. Although investment is a highly cyclical variable and thus can be growing slowly during a cyclical downturn without causing a long-term problem, the slowdown in the growth of investment for over a decade can be a source of concern for the long-term prosperity of an economy. Physical investment is the building block of the capital stock which makes the economic growth possible. Levine and Renelt (1992), for example, identify physical investment as the most robust determinant of long-run economic growth. Stagnant investment, therefore, can have a dire consequence for long-term economic well-being and, as such, merits the full attention of policy makers and economists alike. Share of investment Growth rates of investment do not convey the information on the ‘level’ of investment. To get a measure of comparison that retains information on both the level and change of investment, controlling for the different sizes of the economy at the same time, I look at the ratio of investment to its output. The comparison qualifies the sense of stagnation imparted by Table 17.1. Table 17.1 Average growth rates in gross fixed investment US

UK

France

Germany

Italy

1975–95 2.87 1.43 0.95 2.06 0.50 1980–95 2.25 1.65 1.25 1.91 0.68 Notes: The growth rates are average annual rates. They are calculated on the basis of the OECD data. The data ends in 92 for Germany and in 94 for Italy.

Aggregate Investment Figure 17.1 shows the ratio of gross fixed investment to GDP of each country, on the basis of the data from OECD.2 The figure conveys two messages. For one, growth rates of investment change in the same broad direction as in Table 17.1. The ratio of investment to GDP declined sharply in France, Germany and Italy since the mid-1970s and has not recovered yet. While Germany shows some contrast to other countries in terms of the growth rates of investment in Table 17.1, the movement of its ratio of investment is very similar to those of France and Italy. Rather, the movement of the ratio for the UK stands out among European countries. The ratio in UK kept plunging till the early 1980s but bounced back from the mid-1980s. For the whole period from 1975 to 1995, however, the decreasing investment in the early 1980s seems to have undermined the brisk performance during later years, thus generating the low growth rates of investment in Table 17.1. Overall, we get the sense of a relative decline in the intensity of investment for four European countries, except for a late rebound in the UK. This pattern fits well with the image of stagnant European investment that is imparted by Table 17.1.

198

JAEWOO LEE

Figure 17.1 Gross fixed income. Source: OECD.

On the other hand, the level of the ratio of investment is much higher in three Continental countries than in the US. The ratio in the UK is also slightly higher than that in the US for most years. In terms of the levels of investment controlled for the size of the economy, then, it is the US that has had stagnant investments. While the European countries had been putting twice as large shares of output in investment as the US until the mid-1970s, they reduced the share of output put into investment since then. Still, European countries have been investing a larger share of output than the US have been doing. Such a decrease in the order of the multiple of European investments over the US investment, in itself, is not necessarily a cause of concern. Such a decrease can be a natural outcome of Europe’s re-converging to the long-run steady state level of capital stock. Postponing further discussion of this until later, we look at more facts in this section. Sectoral investment To check for the possibility that investment performances have varied across different sectors of the economy, this section compares the investment shares across sectors. Before going into a sectoral division, we compare the performance of the two categories of aggregate investment: equipment vs. non-equipment. Figure 17.2 shows the ratio of investments in machinery and equipment relative to GDP. We see that equipment investment has not declined in Europe, unlike the total investment. Whatever is the source of decrease in the European investment, it does not relate so much to equipment investment as to nonequipment investment. Figure 17.3, that plots the movement of non-equipment investment, shows directly that the non-equipment investment is the major source of the contrast in the investment to output ratio between the US and European countries. This adds to the importance of examining the data at a more disaggregated level. I thus turn to the OECD Intersectoral Database of October 1995. The dataset allows a comparison of investment at various industry levels. Investment and output are gross fixed capital formation and GDP by industry, evaluated at the 1985 national currency unit. Since I will compare the ratio of investment to output, denomination in the national currency poses no problem. Rather than jumping to the industry level comparison, I first organize different industries into two sectors: goods-producing and service-producing sectors. The goods-producing sector includes agriculture, manufacturing, mining and construction. The service-producing sector includes electricity, gas and water, transport, storage and communication, finance, insurance, real estate, etc., community, social and personal services, and producers of government services. This division into goods and services is motivated by the previous contrast between equipment and non-equipment investments. The importance of equipment investment would be higher in goods-producing industries than in service-producing industries. In addition, the service-producing sector is becoming more and more important, most conspicuously in the employment share. In the case of the US, more than 80 per cent of 1995 employment belongs to the service-producing sector. Nor do the two sectors exhibit identical characteristics, as was explored in Lee (1997). This division turns out to be useful for the analysis of investment as well. Figure 17.4 plots the investment—output ratio for the two sectors. In the service-producing sector of France and Germany, the investment—output ratio has fallen sharply and stayed at low level since the mid-1970s. In the goods-producing sector of both countries, investment—output ratio has fallen little, at least much less than in the service-producing sector. In Italy, on the other hand, it is the investment—output ratio in

CAPITAL PRODUCTIVITY IN EUROPE AND THE US

199

Figure 17.2 Investment in machinery and equipment. Source: OECD.

the goods-producing sector that has fallen and stayed low since the mid-1970s. Investment-output ratio in the serviceproducing sector had a dip through the late 1970s and the early 1980s, but fully rebounded by the mid-1980s. In UK services, the ratio had a dip in the early or mid-1980s, but recovered afterwards.3 I then examine movements of investment—output ratio for individual industries, and sometimes dividing industries to finer categories when a strong pattern is observed. After straightforward comparisons of many industries, I identify several cases where the investment—output ratio exhibits a strong contrast with those of other countries and at the same time is persistent in the sense that the pattern seems to be more than cyclical fluctuations in terms of its duration. Table 17.2 contains the five-year average numbers for the industries which show contrasting patterns. In community, social and personal services, the UK exhibits a large singular decrease in investment—output ratio. It falls from 0.64 in the early 1970s to 0.33 in the late 1980s. In most other countries, the ratio stays constant or increases slightly. In electricity, gas and water, transport, storage and communication, the investment—output ratio has fallen most dramatically in the UK, but also definitively in France and Germany during the period from the mid-1970s to the late 1980s. In real estate and business services, the ratio has fallen greatly in France and Germany during the same period.4 Finally, there is an episode of a strong increase in the ratio of investment to output. In financial institutions and insurance in the US, the invest ment—output ratio has increased since the late 1970s. The margin of increase, from 0.08 to 0.24, is much higher than that of any other country. In the 1990s, as a result, its absolute value of investment–output ratio in finance and insurance is second only to that of the UK, which was high all through the sample period.

200

JAEWOO LEE

Figure 17.3 Non-equipment investment. Source: OECD.

Starting with contrasting the growth rates of total investment among five countries, we came to five episodes of conspicuous movement in investment to output ratio. Although the contrast at the aggregate level cannot have been driven solely by the five patterns just mentioned, the four patterns excluding the US episode are an important part of the contrast at the aggregate level, because they move in the same direction as the investment—output

CAPITAL PRODUCTIVITY IN EUROPE AND THE US

201

Figure 17.4 Share of investment: goods versus services. Source: OECD.

ratio at the more aggregate level of goods and services sectors. A successful explanation of the decline in the investment— output ratio in Europe, therefore, should be able to account for the decline in the four European cases just identified. This becomes one criterion to compare various possible explanations of the decline in the aggregate investment. Table 17.2 Sectoral investment shares US

UK

France

Community, social and personal services 1966–70 0.09 n.a. n.a. 1971–75 0.09 0.64 0.12 1976–80 0.10 0.45 0.11 1981–85 0.12 0.39 0.11 1986–90 0.14 0.33 0.15 Transport, storage and communication 1966–70 0.30 n.a. n.a. 1971–75 0.28 0.32 0.41 1976–80 0.28 0.26 0.42 1981–85 0.21 0.21 0.30

Germany

Italy

0.14 0.18 0.24 0.25 0.25

n.a. 0.10 0.09 0.10 0.13

0.39 0.46 0.34 0.32

n .a. 0.46 0.43 0.38

US

UK

France

Electricity, gas and water 0.32 n.a. 0.90 0.31 0.90 0.66 0.26 0.50 0.66 0.30 0.41 0.55 0.30 0.38 0.33 Real estate and business services 0.45 n.a. n.a. 0.43 0.45 0.89 0.39 0.38 0.72 0.32 0.33 0.58

Germany

Italy

0.54 0.55 0.39 0.42 0.35

n.a. 0.17 0.18 0.22 0.24

1.40 1.27 0.96 0.81

n.a. n.a. n.a. n.a.

202

JAEWOO LEE

US

UK

France

Germany

Italy

US

UK

France

Germany

Italy

1986–90 0.18 0.25 0.25 0.31 0.44 0.34 0.42 0.52 0.67 n.a. Financial institutions and insurance 1966–70 0.07 n.a. n.a. 0.11 n.a. 1971–75 0.08 0.27 0.08 0.10 0.08 1976–80 0.11 0.32 0.07 0.08 0.07 1981–85 0.17 0.45 0.08 0.09 0.08 1986–90 0.24 0.38 0.11 0.08 0.08 Notes: The data are from the Intersectoral Database of OECD. The numbers are the ratio of the gross fixed capital formation and value added, by each sector.

Capital productivity It is not necessarily desirable to have a high investment—output ratio. The most dramatic counter-example is the experience of Asian countries. The financial crisis in Asian countries that started in the latter half of 1997 has been widely attributed to the excessively high investment that has been financed by foreign capital. Japan, which has had a very high ratio of investment to output, has also been experiencing a long slump. What matters is the investment which is productive enough to justify the cost incurred. I thus explore what is implied for capital productivities by the contrast in the investment experiences between Europe and the US. As a first step, I develop a simple approximation to productivities of capital. Measurement Consider a competitive firm that possesses a constant returns to scale technology. The current profit at period t is, with the output as the numeraire, (1) (2) Regardless of the specifics of C(K,I) function, the firm optimizes on labor in the same way. After the optimal employment of labor, the current production function can be written as: (3) Defining (4) the production function can be rewritten as (5) In this formulation, Ht can be viewed as the equilibrium capital productivity. This capital productivity, however, incorporates optimal employment of labor in it, and thus is not a purely technological parameter.5 It depends on capital intensity α, wage rate wt, price Pt, and the level of technology At. Since all determinants of this capital productivity are variables that change slowly, especially relative to fluctuations in output, investment, or employment, the capital productivity will also change slowly. This observation helps us to quantify this capital productivity in a simple manner. Dividing both sides by Yt and noting that we have

Since the change in H is much smaller than the change in the output or the capital stock, we assume that (6) Using also that

CAPITAL PRODUCTIVITY IN EUROPE AND THE US

203

Figure 17.5 Capital productivity. Source: OECD.

we can rewrite equation (6) as: (7) We can thus quantify Ht in equation (5).6 We track the evolution of capital productivities for the five countries. This approximation measures the movements in capital productivities not so much annually as over a longer horizon. Figure 17.5 shows two patterns. The capital productivities are higher in the US than in European countries, but they converge. The capital productivity falls in the US and increases in France, Germany and Italy. Sources of the change Equation (4) suggests that there are three sources for the long-run movement in capital productivities. Taking log of equation (4), we get (8) First, the change in the capital share can help to explain the contrast. An increase in the capital share increases the capital productivity. Burda and Sachs (1988) compare the growth in the capital—labor ratios from 1961 to 1986 between Germany and the US. Although the capital—labor ratio is not identical to the capital share α, the two should be well correlated. The capital —labor ratio grows at a higher rate in Germany for all sectors during the 1960s. During the 1971–86 period, the difference narrows, but still the ratio grows faster in Germany for most sectors. Blanchard (1997) reports the increase in the capital share for Continental European countries. In terms of Figure 17.1, Continental countries were investing more, and thus the increase in the capital share is a natural outcome of high investment. But, Blanchard’s calculation shows that capital share increased by 0.1 at most, which would contribute to the increase of log Ht by a little more than 0.05 through the first term of equation (8). Given the change in log Ht by more than 0.2, the increase in capital share accounts for less than 25 per cent of the increase in capital productivities. Second, the real wage can contribute to the pattern. A decrease in the real wage increases the capital productivity. This works in the wrong direction for the US. Wages have been declining at the same time as the capital productivity has been decreasing. To summarize the average tendency of the sample countries, we run the following regression: (9)

204

JAEWOO LEE

The estimated value of β1 is 0.43 with the standard error of 0.18. The labor market elements do not seem to have contributed to the observed change in capital productivities. It may be inappropriate to assume that the labor market is competitive as in the derivation of equation (8), especially in the presence of so many discussions on the frictions in European labor markets as the culprits of high unemployment. The bias that may be induced by this assumption, however, does not seem to contribute to explaining our finding. If the labor market were highly inflexible in Europe relative to the US, labor would be hired suboptimally in Europe. The suboptimal hiring of labor, then, would decrease the capital productivity in Europe rather than increase it. Finally, then, the rest belong to the change in the multifactor productivity. An increase in the multifactor productivity increases capital productivity. The convergence of productivity among OECD countries have been confirmed by many economists, including Bernard and Jones (1996) as a recent one. Their paper especially reports a finding that squares well with the bigger contrast in the pattern of investment–output ratio in the service sector. They find that the convergence in the aggregate productivities was driven by the convergence not so much in manufacturing industries as in non-manufacturing industries. This lends credence to the interpretation that much of the contrasting investment patterns may be due to the contrasting productivity performances. Investment and productivity Since the low investment-to-GDP ratio can be interpreted as high capital productivities, there does not have to be an immediate concern when the output grows at a reasonable rate. In terms of the long-run implications of the low investment, however, the currently high capital productivity cannot be a source of contentment. It is important to ask what it implies about the productivity at a longer horizon. In addressing the question, I do not rely on models of investment that are aimed at explaining investments at cyclical frequencies. The observed shift over decades requires more than a purely cyclical frame of reference, because ten years seems to be long enough for cyclical divergences from structural equilibrium to balance out. Nor does the involved time horizon seem long enough to lend itself to the analysis by purely growth-theoretic models. This, in fact, may again fall in the domain of ‘medium run’ that has been mostly neglected in macroeconomic analysis. As a way to strike some balance between the two approaches, I seek recourse to a neoclassical framework that is broadly adopted in models of cycles and growth alike, namely the notion that consumption decisions are made optimally in the intertemporal context. More specifically, I start out by focusing on the interaction between (optimal) savings and the profile of expected productivities. Let us first take the time profile of productivity to be exogenous, and consider if the high current productivity always justifies low current investment.7One possibility is that a high current level of productivity is caused by a permanent increase in productivity. To attain the same level of consumption, less capital stock and thus investment are sufficient. If consumption is maintained at the same level as under the lower productivity, some extra resources are freed up. Unless people are extremely short-horizoned, the extra resources will be appropriately allocated between present consumption and future consumption. The part allocated for future consumption is investment. As far as a significant part of the extra resource is allocated for the present consumption, the ratio of investment to output will be lower when the productivity is higher than when the productivity is lower. In this case, the lower investment is the optimal choice, except for possible quantitative disputes on the desirable extent of the decrease in the ratio of investment to output. If the high current productivity accompanies a changing time profile of productivity, however, the decrease in investment can be concerning. Let us consider the case when the future productivity is expected to be lower than the current productivity. Here again there are two possibilities. First consider the case when there is a downward adjustment in the anticipated future productivity. Then, the future consumption becomes relatively more costly in terms of the current consumption. The substitution effect tends to tilt the consumption in the direction of increasing the current consumption, and the income effect tends to decrease both the current and the future consumptions. If the substitution effect dominates, the current investment can fall. Still, such a fall in investment is accompanied by the overall decrease in the consumption (production) possibility, and is not welcome. Alternatively, consider the case when there is a temporary boost to the productivity, namely when the productivity is higher now than both in the past and in the future. Although the same substitution effect tends to tilt the consumption profile in the direction of increasing the current consumption more than future consumption, the benefit of the temporarily high productivity is spread out over all horizons to increase the consumption of all periods, possibly at a lower investment than before. If the ratio of investment falls because of this effect, here again is little cause for concern. We can then ask whether the particularly low investment in the sectors identified in the previous section is caused by a temporary or permanent increase in productivity in those sectors. Given the result of productivity convergence literature, however, the productivity in those sectors has been gradually improving or catching up with the US productivities. These are not an abrupt increase in the productivity that is permanent or decade-long though temporary. The sustained and gradual increase in the productivity can hardly explain the drastic fall in the investment ratio experienced in these sectors.

CAPITAL PRODUCTIVITY IN EUROPE AND THE US

205

Hence, the decrease in investment in these industries is best understood as coming from an abrupt downward revision of anticipated productivities. To consider what might have instigated such a downward revision of productivities, we now consider what can explain the productivity change, namely the determinant of productivity. Since we would like to understand how people form expectations of productivity prospects to be realized in the future, it is natural to refer to our understanding of economic growth. Recent literature on endogenous growth, surveyed in Aghion and Howitt (1998) for example, offers a very rich set of possibilities for explaining the productivity differential. Most of those explanations, however, are not sectorspecific, and thus are little likely to be able to pinpoint the causes of revisions in sector-specific productivities. The economywide number of researchers and scientists is little likely to have a disproportionately negative effect on the sectors identified earlier. Nor are conditions in markets for labor or finance likely to have a disproportionate effect on the three service sectors, because similar factor market conditions are faced by all sectors. This then points us to the product market as the major cause of the difference. The product-market characteristics of the three service sectors could have led to the expectation of lower productivity for these sectors, thereby lowering investment. Aghion and Howitt (1998, chapter 7) offer several frameworks under which the competition in the product markets can affect the adoption of technologies and thus productivity. More generally, unfavorable product-market characteristics would decrease the profitability of future operations and thus depress the incentive to invest. While it is beyond the scope of this chapter to explore the specific product-market characteristics that contributed to reducing the investment incentive in these industries, there are suggested evidence on the limiting characteristics of the output market from the McKinsey report. Burda and Sachs (1988) also suggest looking to the product market for differences in the capital—labor ratio between the manufacturing and services industries. One remaining challenge, however, would be to find what product-market characteristics brought about the apparently abrupt change in the pattern of investment in the mid-1970s. Whether the abruptness is due to a change in the market conditions at that time or due to external shocks that had differential effects on industries according to the product-market characteristics, there seems to be a need for an extensive exploration for the working of the service sector at large. Although there has been much research on the productivity measurement of the service sector, little research has been done on the working of service industries, despite the informed interest in it.8 More research is in order, to understand how external shocks, the market characteristics, and even the internal organizations, would affect the performance of the service sector, not only from the long-run perspective as in the productivity literature, but also and more from the medium-run perspective. In other words, a more macroeconomic or general-equilibrium perspective seems in order. Conclusion Between the US and some European countries, the movement in aggregate investment seems comparable in manufacturing industries. The contrast has been largest in non-manufacturing industries. As much as the low investment—output ratio is a concern, the effort to understand its cause or to formulate a solution should be directed towards understanding the service sector better. Although there has been increased interest on the workings of industries that belong to the service sector, the research has been mostly from a partial-equilibrium perspective. Little has been offered in the way of a general-equilibrium analysis that enables us to analyse the decrease in investment—output ratio as was seen in some European countries. A successful explanation would require us to have a better macroeconomic, or general-equilibrium, understanding of the serviceproducing industries. Such an understanding will then shed light on the interaction among product markets, investment, and technological progress. Notes 1 I thank, without implicating, Michael Burda, Ciaran Driver, Kaku Furuya and Paul Temple for comments. 2 The base year is 1980 for France and Germany, 1985 for Italy, 1987 for the US and 1990 for the UK. 3 This finding is comparable to the contrast that Burda and Sachs (1988) found in terms of the capital—labor ratio of manufacturing and services, between the US and Germany. In particular, refer to their table 7. While they interpret this contrast in the growth of the capital—labor ratio from the viewpoint of sluggish employment growth in Germany, there is no precluding a related contrast in terms of investment. More generally construed, their result lends support to a difference between manufacturing and services. 4 Early ratios of investment to GDP for Germany are larger than 1. This is possible, in an accounting sense, if the industry borrows heavily from other industries. I have not found out whether the industry was indeed investing this heavily or the statistics are inconsistent because of some unrecorded revisions. 5 There are a few implicit assumptions in this measure of capital productivities. First, since perfectly competitive firms are assumed, strategic pricing/production decisions by firms in the imperfectly competitive market are ignored. A significant change in market imperfection, as would be captured in the changes in mark-up, will thus compromise the accuracy of this measure of capital productivity. Second, a competitive labor market and full capacity utilization are assumed. Again, cyclical deviations of them from

206

JAEWOO LEE

the assumed benchmark will contaminate this measure. Therefore, the measure will better approximate the actual capital productivity for the time horizon over which such variation averages out. 6 This approximation is simpler to implement than the alternative of constructing the capital stock variable. The alternative approximation can be derived, using log Kt+1−log Kt≈It/Kt, as

7 No distinction is made between the marginal and the average productivities, because the productivity (At) enters linearly into the capital productivity (Ht) in equation (4). 8 As an example, refer to Martin (1998).

References Philippe Aghion and Peter Howitt, Endogenous Growth Theory, 1998, MIT Press: Cambridge, MA. Charles Bean, ‘European Unemployment: A Survey,’ Journal of Economic Literature 32(2): 573–619, 1994. Andrew Bernard and Charles Jones, ‘Comparing Apples to Oranges: Productivity Convergence and Measurement across Industries and Countries,’ American Economic Review 86(5) (1996), 1216–1238. Oliver Blanchard, ‘The Medium Run,’ Brookings Papers on Economic Activity 1997:2. Michael Burda and Jeffrey Sachs, ‘Assessing High Unemployment in the Federal Republic of Germany,’ The World Economy 11(4) (1988), 543–563. Jaewoo Lee, ‘Do Services Temper Fluctuations?’ 1997, UCI mimeo. Ross Levine and David Renelt, ‘A Sensitivity Analysis of Cross-Country Growth Regressions,’ American Economic Review 82(4) (1992), 942–963. McKinsey Global Institute, Removing Barriers to Growth and Employment in France and Germany, March 1997. Peter Martin, ‘Revolution Again,’ The Financial Times, 4 June 1998.

Index

Note: page numbers in italics denote tables or figures where these are separated from their textual reference Abbegglen, J. 126 Abel, A.B. 3, 6, 7, 9, 11, 12, 14, 16, 78, 80, 94, 183 Abel-Blanchard method 6, 7, 9, 11 Abel-Hartman framework 82, 104 Abramovitz, M. 164 adjustment cost 73, 74n5; asymmetric 79, 82; convex 75n16; flexible accelerator 6 adverse selection 87 Agarwal, J. 203 agency cost function 64–5 agency premium parameter 73 agency problem: see principal-agent problem aggregate investment 182, 330, 331 Aghion, Philippe 167, 340 Aiginger, K. 12, 13, 17 Albach, H. 122 Alexander, I. 127, 256 Allen, C.B. 243 Allen elasticities 234, 235 Amsterdam Treaty 317, 327 Anderson 243 Anderson, O. Jr 283 Anderson, T. 71 Aoki, M. 116 arbitrage 124–5, 128 ARCH estimations 155–6 ARCH models 139–40, 145 Arellano, M. 71, 89 Arrow, K.J. 165 Aschauer, D.A. 184, 189 asset lives 100 automotive components 303, 304, 305

Barran, F. 53, 54, 59, 64, 73 Barrell, R. 172, 202, 205, 207, 215, 216 Barry, F. 202, 210 Bean, C.R. 5, 174, 223, 252, 266, 268, 293, 329, 342 Beaudry, P.M. 85–6 Beckett, Margaret 106 Belgium: FDI 202; inward-investment growth 207 Berle, A.A. 106, 117, 129 Bernard, Andrew 338 Bertola, G. 82 Bhaduri, A. 4, 5 Bhattacharya, Rina 15, 20 Blanchard, O. 5, 7, 8, 9, 11, 237, 260, 338 Blinder, A. 267 Blomström, M. 210 Blue Book Industry concentration 96 Blundell 243 Blundell, R. 8, 122, 292 Bollerslev, T. 139 Bombach, G. 164 Bond, S. 51, 55, 57, 61, 64, 68, 71, 89, 122, 256, 262, 293 borrowing costs model 53–4 Bosworth, B. 8 Bradley, J. 202, 210 Brinkley, Ian 258, 259 Brown, Gordon 303 Buigues, P.-A. 260 building investment: cross-country analysis 187–92; and equipment investment 188–9; and growth 184, 194; and output 188, 195; timescales 185, 187–8 Burda, Michael 338, 341 Burnside, C. 232 business cycle, diffusion of technology 136 business sector investment 318 BZW 261

Balasubramanyam, V.N. 210 Ball, M. 185 Baltagi, B.H. 71 bankruptcy 252 banks: debt/equity holders 114–16; Germany 114–15, 116, 129; Japan 115, 116, 122, 129 bargaining behaviour 31, 38, 175, 227–8, 323–4; see also wage bargaining Barker, Kate 255

Caballero, R.J. 9, 20, 79, 82–3, 84, 93, 95, 138, 262 Cable, J.R. 115 Cadbury Committee 106, 327 Callen, T. 38, 40 capacity utilisation 41; increasing 310;

207

208

INDEX

price 269, 270, 271; profits 121; sectoral 86; and unemployment 259 capital: cost of 88, 97; irreversibility 15, 23n20; and labour 177, 310; liberalisation of flows 5; lumpiness 277; marginal revenue product 5, 17; physical 316, 325; shadow value 7, 8; shortage of 270, 272 capital-debt ratio 57 capital-labour ratio 164, 221, 223, 224, 338, 341, 342n3 capital constraints 9–10, 277 capital costs 41, 120; investment 38; policy tools 262; social/private 263n1; unemployment 240, 241 capital deepening 163–4, 165, 175–6, 253 capital gearing 36, 41 capital market hypothesis 115 capital productivity 223, 260–1, 335–8, 342n5, 337 capital stock 192–4, 195; adjustment 126; employment 173, 174, 175, 309; growth rate/output 9; investment 118, 170, 258; machinery and equipment 168, 169; neutrality 242; physical investment 329; technical progress 223; UK 302, 319 Carroll, C. 253 Carruth, A. 231 cash-flow 32, 35–6, 51, 53, 74 catch-up, technology 164 CBI surveys 255; business expectations 38; database 20, 283–7; demand uncertainty 95; hurdle rates 305; Industrial Trends Survey 48, 86, 87, 97, 270, 282, 310; interviews of companies 302–4; investment appraisal 293; investment levels 302–3; panel data 79; share prices 126, 127, 128; ‘Target Practice’ 256 Census of Production 96 Chan, A. 268 Chatelain, Jean-Bernard 11, 19–20 Chick, V. 22n2 Chief Executive Officers 112, 126 Chirinko, R.S. 8, 8, 31, 51, 116 Cho, M. 112, 119 Chow, G.C. 140 Christofides, L.N. 242

Clark, P. 164 Clarke, Kenneth 136 Clayton, T. 253 Coakley, Jerry 170 Coates, J. 127 Coe, D.T. 213 cointegration tests 40–1, 42–8, 232, 233; fiscal policy/investment 296; manufacturing 41–3, 48; output per capita 182, 188, 196n10; wages 236, 237 collective bargaining 323–4 Collison, D. 127 common stock ownership 111 company law 261, 311 comparative advantage 217 competitiveness 34, 266–7, 268, 279–81, 288n5; see also imperfect competition concentration ratios, distribution 94 Conlisk, J. 17 constant risk premium 78 consumption, productivity 339–40 contracts, long-term 268 control: inside/market 109–10, 112; loss of 131n1; outside 122–3, 124 Conyon, M.J. 107, 113 corporate culture 258, 324–7 corporate governance 106–7, 110–14; alternative structures 107–9; decision-making 257–8; innovation 259; managerial discretion 116–20, 122; national economic performance 128; short termism 324–7; taxonomy 109–10; YUC 326–7 corporate performance 113 corporate policies 117 corporate size 118–19 corporation tax reforms 303, 306 Cosh, A. 253 cost, and price 253, 266–8, 281–2 cost-of-adjustment approach 18 cost of capital measure 97 cost function 225, 232 cost minimisation 43, 293 costs: convex 8, 10–11; and price 253, 266–8, 281–2 coverage ratio 51 Cowling, K. 5, 267 Crafts, N.F.R. 166, 171 Craine 79 Craven, B.M. 127 credit rationing 69–70, 88 cross-border investment, multinationals 199 cross-country analysis, building/equipment investment 187–92, 195 cross-shareholding 261 Crotty, J. 17

INDEX

Cubbin, J. 119 cultural factors, investment 206–7 Cunningham, A. 289n23 currency links 207–8 customer location 303, 313n3 Cuthbertson, K. 36, 38, 39, 43 Datastream 111, 120 Davies, E.W. 127 Davies, S. 136 Davies, S.W. 216 Davis, G. 139 debt, supply price 36 debt-capital ratio 57 debt-equity ratio 51, 53, 71, 114, 126 debt ceiling 51, 53, 61–2 debt repayment 20 decision-making, corporate governance 257–8 deflationary policies 309 delivery lag model 9 DeLong, J.B. 169, 170, 182, 183, 184–5, 187 demand: inflation 259–60; and mark-up 79; and output 56; uncertainty 83, 87, 95 Demetriades, P.O. 168 Demirag, I.S. 127, 128 Dennison, E.F. 170 Denny, K. 31, 37, 38 Denny, M. 225, 227 Department of Trade and Industry 256 Devereux, M. 10, 122 Dickerson, A.P. 123 Diederen, Paul 21 diffusion of technology 136, 151 diminishing returns 164, 165, 251 directors 112, 113, 259, 324 discounted present value, tax savings 35, 37 dividends: investment 10, 34, 256, 257, 258, 324; Lagrangean 58–9; taxation 54, 306 Dixit, A. 12, 15, 31, 79, 81, 136 Dixit-Pindyck model 15, 100–3 Dixon, R. 140 Dorrell, Stephen 301 downsizing 257 Dreze, J.H. 268 Driver, C. 13, 17, 31, 84, 136, 137, 138, 139, 253, 262, 310 Dupont, M. 217 Dyke, Greg 314n7 dynamic cost function 242–4 dynamic factor demands 232 Eberly, J. 84, 138 econometrics 31, 69–73 Economic Outlook, London Business School 320 economic policy, investment 251 Economic Policy and Social Partnership (TUC) 326

The Economist 314n5, n8 education, employability 317 Edwards, J. 114, 115, 116 efficiency, FDI 172–3 Eichenbaum, M. 232 Eisner, R. 12 Elston, J.A. 116, 122 Eltis, W. 210 Emerson, R. 268 employability 317 employment: capital deepening 176; capital stock 173, 174, 175, 240, 309; casual 312–13; FDI 204; investment 173–8, 182, 252, 310, 325; low paid 324; productive 165; supply-side policies 310–11; UK 222; see also job creation EMU 326 endogenous growth theory 172, 182, 195n1, 340 England, Bank of 326 Engle, R.F. 139, 192 equipment investment 332; and building investment 188–9; cross-country analysis 187–92; GDP 331; growth 182, 183, 185, 186, 194; and output 188, 192, 195 equity markets 112, 121 Ergas, H. 171 Ericsson, N.R. 192 Ernst and Young 257 Error Correction Model 40, 296–8 errors in variables model 39 Estrada, A. 73 EU: Amsterdam Treaty 317, 327; Broad Economic Guidelines 317; capital-labour ratio 338; economic recovery 322; employment creation 317; FDI 200, 201–2; investment-GDP ratio 316; investment patterns 206–9, 317; national champions 171; outward investment/exports 211; Single Market Programme 201, 206; unemployment 221, 309, 329, 338 Euler equation 8, 11; financial constraints 51; and flow of funds 70–3; and Lagrange multiplier 61; testing 68–9 Evening Standard 251 exchange rate policy 311, 323, 325–6 expectations formation 31, 37–8 explicit forward looking models 46–7 export performance, FDI 210–11

209

210

INDEX

externalities, lack of 251 extrapolative model, REH 39 factor demand system 238–9 factor prices, exogenous 238–40 Fairness at Work 325 Farinha, L. 216, 217 Fazzari, S.M. 10, 17, 51, 52, 69, 121 FDI 172, 199–200; costs/benefits 199; effiency 172–3; employment 204; EU 200, 201–3; export performance 210–11; flows of 205; growth 207, 216; host economies 211; inward 211–14; manufacturing sector 203–4; R&D 209; technological innovation 172, 209–10, 212–13, 218; US 200, 202 feedback mechanisms, spillovers 165 Feldstein, M. 12, 294 Feldstein-Horioka paradox 22n11 Ferner, A. 206 finance, sources 120–1 finance models 10 financial accelerator 10 financial constraints 9–11, 38, 51, 121–2, 255–6, 304 financial markets, short-termism 256–7 Financial Times 136, 251 firm behavioural model 32–7, 53–9, 130 firm-specific knowledge 130, 172, 217 Fiscal and Economic Strategy report 326 fiscal policy 255, 292, 293–8 Fischer, K. 114, 115, 116 Fischer, S. 78 fixed investment 183, 187 Flemming, J. 294 flexibility of labour 258, 311–12, 316, 323–4 flow of funds equation 70–3, 75n18 food retailing 303, 304 Ford, R. 9 forecasting and investment 3–4 foreign companies 204; competition 288n5; spillovers 216–17; in UK 303–4; see also FDI; multinational firms formalism 22n2 forward markets 251 Foster, J. 289n23 France: capital productivity 337; FDI 200, 202; foreign-owned firms 204; investment-output ratio 333, 334; investment growth 207, 329, 330; wages/profitability 321

France, Bank of 69, 76n19 François, J. 217 Franks, J. 110, 111 free-riders 112 freedom, degrees of 73, 76n23 French firms, debt-equity ratio 51 fuel input 232, 234 full information maximum likelihood estimation 40 Fuss, M. 225, 227 GARCH estimator 85 Gasparro, D. 36, 38, 43 Gavosto, A. 174, 266, 268 GDP: growth rate 245n2; and investment 130, 163, 164, 316, 318, 330, 342n4; UK 223 Generalised Method of Moments 71, 76n22, 89 Georgoutsos, D. 34, 35, 55 Germany: banks 114–15, 116, 129; capital productivity 337; corporate governance 111, 112–13, 114; currency links 207–8; FDI 200, 202; firm growth 130; foreign-owned firms 204; investment-GDP 432n4; investment-output ratio 333, 334; investment growth 329, 330; technical progress 211–12, 214; unemployment 128; wages/profitability 321 Geroski, P. 252, 268, 272 Gertler, M. 52–3, 63 Ghosal, V. 20, 83, 86 Gibson, H.D. 123 Gilchrist, S. 11 goal compliance 117, 131n1 Godfrey, L.G. 215 Gompertz variants 140 goods-producing sector 331–3, 334 Görg, H. 206, 210 Granger, C.W.J. 3 Granger causality tests 274 Greene, W.J. 155 greenfield investment 203, 209–10, 218 Gregory, M. 289n23 Grieve Smith, J. 259–60 Griliches, Z. 140, 171 Grinyer, J. 127, 128 Grossman, G.M. 171, 210 Grossman, S. 113 growth 182, 289n23, 325; building investment 184, 194; endogenous 172, 182, 195n1, 340; equipment investment 182, 183, 185, 186, 194; FDI 207, 216; and fixed investment 183; GDP 245n2; income per capita 222;

INDEX

instability 322–4; managerial culture 261; output 9; and profitability 5; public investment 184, 261–2; and R&D 119; regulation 258; stability 262; UK 256, 322–3; utility function 118 growth orientation 261, 304 growth-valuation frontier 118, 130 Guardian 251 Guiso, L. 20, 23n20 Hall, R.E. 138 Hall, S. 38, 39, 40, 235, 267, 272 Hamermesh, D.S. 8 Hart, O. 56, 108, 112, 113 Hartman, R. 14, 78, 80, 94 Hartman-Abel model 14, 16 Haskel, J. 267 Hay, D.A. 173 Heiner, R. 17 Helpman, E. 167, 171, 210, 213 Hendry, D.F. 192 Henry, Brian 18–19 Henry, S. 38, 40 Henry, S.G.B. 232 high-tech business 307 Himmelberg, C.P. 11 Holland, D. 204, 210, 217 Hope, Paul 15, 20 Hoshi, T. 10, 116, 122, 122 Howenstine, N.G. 204 Howitt, P. 167, 340 Hsiao, C. 71 Hu, X. 10 Hubbard, R.G. 11, 51, 52–3, 63, 69, 121 Hubert, F. 204, 207 human capital: formation of 165, 166; Penrose effect 117–18; underinvestment 316, 323–4, 325 human capital formation 165, 166 Hungary, foreign-owned firms 204 Hunya, G. 204, 206 hurdle rates 255, 257, 304–5 Hurn, A.S. 85 Hutton, Will 107, 221, 251 hybrid model, cointegration 43–5 Im, K.S. 272 IMF, International Financial Statistics Yearbook 143 imperfect competition 34, 79, 82, 83 improvement engineering 184 income per capita 222 income, productivity 339–40 Industrial Finance Initiative 301 industry groups:

dependent variables 276, 278–9, 281; Granger causality tests 274; stationarity tests 273 industry structure 91–2 inflation: demand 259–60; investment 21, 253; manufacturing 283–7; uncertainty 84–5; unemployment 224, 238, 309 information: advantage 123; assets 125; asymmetrical 87, 108, 121–2, 128–9 Ingham, K.P.D. 216 innovation: corporate governance 259; FDI 218; flexible work practices 258, 312–13; incentives 173; and investment 167, 169–70, 183–4; multinationals 206; as public good 83–4; spillover 184; see also technological innovation insider-trading 110 insiders, productivity gains 251 instability, economic 304, 322–4 institutions, short-termism 122–3 intangibles, investment 123 integration 42, 295; see also cointegration interest rates, investment 150, 157n8 International Federation of Robotics 140, 143 investment 3, 96, 97, 98–9, 130, 182, 321, 329–35; capital stock 118, 170, 258; and cash-flow 51, 53, 74; constraints 9–12, 302; cost-of-adjustment approach 18; and debt repayment 20; dividends 10, 34, 256, 257, 258, 324; dynamics 6–9, 37–8, 44; employment 173–8, 182, 252, 310, 325; EU 206–9, 317; firm factors 304–6; fiscal policy 251, 255, 292, 293–8, 307; GDP 130, 163, 164, 316, 318, 330, 342n4; growth 182, 325; incentives 306, 307, 313n3; inflation 21, 253; information asymmetries 121–2; and innovation 167, 169–70, 183–4; intangibles 123; and interest rates 150, 157n8; joint stock companies 108–9; managerial discretion 116–20, 122; neo-classical 51, 53–4, 80; option value of waiting 81–2, 84; and output 94, 187, 260, 333, 334; ownership 119–20; price uncertainty 78–9;

211

212

INDEX

productivity 311, 314n4, 339–41; rates of return 178; real factors 36–7; risk 15; and savings 12, 163–4; short-termism 107, 125–8, 311; sluggishness 23n23, 29; technology 21, 31, 34; uncertainty 12–15, 79, 80–4, 95–6, 136–7; and unemployment 174, 221, 309; US 329; wages 31, 36–7 investment, types: aggregate 182, 330, 331; building 184, 185, 187–92, 194, 195; business sector 318; equipment 182–92, 194, 195, 331, 332; fixed 183, 187; inward 211–12; irreversible 82, 84, 137; manufacturing 29–30, 41–3, 48, 85, 90–3, 319; non-equipment 331, 332; non-manufacturing 209; physical 329; process technology 83–4; public 184, 261–2, 301, 314n11, 320, 326; sectoral 330–5 investment appraisal 3–4, 255–8, 293 investment policy 253–5, 261–2, 306–8 investment regimes 4, 59–61; credit rationing 63–4, 65, 73; econometric test 69–73; financially distressed/consolidation 61–3, 65, 66–7; unconstrained 60–1 investors 78, 307 inward investment 211–12 Ireland: firm-specific knowledge 172; foreign-owned firms 204; high-tech investment 209; inward-investment growth 207; US firms 208 IS-LM equation 239 Italy: capital productivity 337; FDI 202; investment-output ratio 334; investment growth 207, 329, 330 Jackman, R. 223, 310 Jacquemin, A. 260 Jaffee, D. 121 Japan: banks 115, 116, 122, 129; corporate governance 110–11, 113; exports 211; FDI 200; growth of firms 130; growth/investment 119, 128, 129; investment-output ratio 335; investment shares 317;

robotics 147 Jaramillo, F. 51, 53, 54, 64, 65, 63, 68 Jenkinson, N. 12 Jenkinson, T. 256 Jensen, M. 114 Jensen, M.C. 106, 112 Jensen, R. 138 Jensen’s inequality 80, 82 job-creation 310, 317 Johansen procedure 41 joint stock companies 108–9 Jones, C.I. 167 Jones, Charles 338 Jorgenson, D.W. 97 Jost, T. 200, 203 Journal of Economic Literature 51 Junankar, S. 270, 282 Kaldor, N. 222 Kanago, B. 139 Kaplan, R.S. 11 Kaplan, S.N. 51, 52, 53, 68, 69, 122 Karshenas, M. 138, 140 Kashyap, A. 51, 52–3 Kathuria, R. 122 Katz, M. 84 Keller, A. 212 Kelly, C. 36, 100 Keynes, J.M. 259 King, M.A. 293 Kiyotaki, N. 53, 63, 56 knowledge transfer 205, 206–7 Kocherlakota 71 Kokko, A. 210 Krugman, P. 205 labour, and capital 5, 177, 310 labour costs 205, 206–7 labour force: bargaining power 174; casual 312–13; innovation 258; skills 255–6; training 325 Labour Government 312, 313, 325 labour market: Europe 338; flexibility 312, 313, 323–4; long term behaviour 222; shortage 270, 271, 272, 308; stability 325; turnover 312, 313; UK 207 labour productivity 170, 204, 209, 216 Lagrange multiplier 58–9; coverage ratio 51, 53; debt-equity ratio 71; debt ceiling 53, 61–2; Euler equation 61 Lakonishok, J. 124

INDEX

Lamfalussy, A. 130 Lamont, Norman 301 Lansbury, M. 201 Layard, R. 173, 175, 223, 310 Le Chatelier principle 234 Lee, Jaewoo 260–1, 333 Lee, K. 283 Leech, D. 119 lemons premium 10, 57, 72, 74, 121 leverage costs 51, 53–4, 68, 74 Levi, M.D. 87 Levin, R.C. 173 Levine, R. 167, 182, 329 licensing 210 life cycle effect 119 liquidity 36, 41 liquidity constraint 87 Liu, G.S. 173 loans 10, 121 location choices 303, 313n3 locational advantages, multinationals 205 lock in problems 110, 124, 130 Lomax, J.W. 9 London Business School 320 long-run structural model of production 225–7, 241–2 Loungani, P. 20, 83, 86 low dividend companies 51, 52 Lucas, R.E. 59 Lund, P.J. 292 Lyons, B.R. 216 Maccini, L.J. 8 McConnell, J. 126, 127 McKinsey consultants 258, 302, 311, 341 Majluf, N. 121 Malinvaud, E. 174 management: compensation 113; constraints 11; culture of 261; effort 121; goals 117; incentives 306; informational advantage 123; moral hazard 74n9; and ownership 108, 109–10, 118, 123–4, 128–9; shorttermism 123; skills 308; see also principal-agent problem managerial discretion 116–20, 122 Mankiw, N.G. 166 Manning, A. 177, 224, 227, 229, 230 Mansfield, E. 140 manufacturing: capital-labour ratio 341; expectations 45–7; FDI 203–4; inflation rate 283–7; output 41; US 86, 90 manufacturing imports 287n2

manufacturing investment 29, 30, 319; cointegration 41–3, 48; uncertainty 85, 90, 91, 92, 93 marginal revenue product 5, 17, 80, 81 Marglin, S.A. 4, 5 mark-up 79, 288n8 market imperfections 31 market value of firm 118 Markusen, J.R. 205 Marris, R.L. 117, 118, 119 Marsh, P. 123, 124, 126 Marston, C.L. 127 Mata, J. 216, 217 Matthews, R.C.O. 174, 177 Mayer, C. 106, 107, 110, 111, 121, 127, 256, 261 Mayes, D.G. 262 Means, G.C. 106, 117, 129 Meghir, C. 51, 55, 57, 61, 64, 68, 122 Meltzer, A.H. 252 mergers and acquisitions 202–3, 209–10, 218 Merton, R.C. 78 Michie, Jonathan 258, 312 Miles, D. 126, 127, 128 Miller, M. 10 Minford, M. 270 minimum wage 325 mispricing arbitrage 128 modelling cointegration 42–8 Modigliani, F. 10 Modigliani-Miller theorem 75n12, 120 monetary targeting 253 monetary uncertainty 85–6 monitoring 112, 129 monopolistic competition 34 Moore, J.H. 53, 56 moral hazard 74n9, 87 Morck, R. 114 Moreton, D. 13, 84, 136, 137, 138, 139, 262 Morgan, E.J. 256 Morris, D.J. 110, 114, 125, 126, 129 Morris model, short-termism 128 Mowery, D.C. 183 Mueller, D.C. 119, 122 Mullins, M. 125 multinational firms 199, 204–6; see also FDI Muscarella, C. 126, 127 Myers, S.C. 121 NAIRU 174, 224, 241, 259 Nash bargaining solution 36–7, 175, 228 national income, public investment 320 neo-classical economics: capital stock adjustment 126; cointegration 42; cost minimisation 293; investment 51, 53–4, 80; production function 117 net present value 16–17, 257 Netherlands: foreign-owned firms 204;

213

214

INDEX

inward-investment growth 207 New Deal 325 new share issue 55, 74n10 Newsboy inventory model 14, 15, 17 Nickell, S. 31, 37, 38, 124, 127, 223 Nickell, S.J. 8, 12, 83, 112, 141, 310 Nicolitsas, D. 8 Nilsen, O.A. 9 Nixon, James 31, 175, 177, 178, 232, 235, 320 non-equipment investment 331, 332 non-manufacturing investment 209 Odagiri, H. 113, 114, 115, 117, 19 OECD: capital stock/employment 174; investment flows 200, 201; R&D 167 OECD Intersectoral Database 331 opportunity costs of funds 5 optimal factor demands 31 option values 4, 23n16, 81–2, 84 options, real 15–17 Oswald, A.J. 231 Oulton, N. 169, 183 output: building/equipment investment 186, 188, 192, 195; cointegration analysis 182, 188, 196n10; and demand 56; elasticity 295; and growth rate 9; investment 94, 187, 260, 333; labour productivity 170; manufacturing 41; and profitability 18; robot price levels 142; uncertainty 84–5 output market 341 Owen, D. 36, 100 ownership 106; concentrated 109, 110, 120, 129; dispersed 123; investment 119–20; and management 108, 109–10, 118, 123–4, 128–8 PACEC 210 Pain, N. 172, 201, 202, 204, 206, 205, 207, 210, 211, 215, 216, 217 Parigi, G. 20, 23n20 Partners for Progress (TUC) 327 patents 83, 171, 172, 173, 206 Peck, S.I. 107, 113 Peck, Simon 11, 20 Peeters, M. 53, 54, 59, 64, 73 Pencavel, J. 228 Penrose, Edith 8, 11 Penrose effect 117–18 Pesaran, M.H. 38, 272, 283, 289n23 Petersen, B.C. 51, 52, 69 Pfann, G.A. 8 pharmaceuticals 303, 304 Phillips curve analysis 272

physical investment 329 Pindyck, R.S. 9, 12, 15, 16, 31, 79, 81, 82, 84, 136, 138 Poret, P. 9 Porter, M.E. 111 Portugal: foreign/domestic firms 217; inward-investment growth 207 Potters, J. 258 Prais, S.J. 209 Precious, M. 35 price 266–7; capacity utilisation 269, 270, 271; competitiveness 266–7, 268, 279–81; and cost 253, 266–8, 281–2; cyclical behaviour 232, 287n4; elasticity 295; fiscal policy 293; robot stock 142, 150–1, 152; uncertainty 78–9 Price, S. 85, 103–4 price equation 268–70; estimation/results 271–9; OLS results 285–7 principal-agent problem 108, 119–20, 124 probability distribution approach 38 process technology investment 83–4 product market 340–1 production, research-based 206 production function 108, 117 productivity: Europe/UK 256–7, 314n8; government policy 307; income effect 339–40; insider gains 251; investment 311, 3114n4, 339–41; labour shortage 272; output-investment ratio 260; sources 244; substitution effect 340; total factor 213; UK 314n8, 337; unemployment 223; US 165, 184; wages 224–5, 231, 236, 237, 320–1, 338; see also capital productivity; labour productivity Productivity and Social Partnership (TUC) 320 profit 121, 228, 252–3 profit maximisation 10, 108, 117, 171, 257–8 profitability 5, 18, 41, 125, 321 Prowse, S. 111, 115 public investment: concern 301, 314n11; growth 184, 261–2; national income 320, 326; UK 327–8 Q models 121 q-theory approach 8–9 quadratic adjustment costs 34 quality ladder approach, R&D 171

INDEX

quasi-likelihood ratio tests 47 R&D 8; expenditure 171, 206; FDI 209; and growth 119; OECD 167; patents 173; pool of knowledge 171; spillovers 165–6, 252; taxation 313; total factor productivity 213 rates of return 125, 129, 178, 252; see also diminishing returns rational expectations hypothesis 38, 39–40, 45–7 Ravenscraft, D. 114 real business cycle 232 real options models 15–17, 137–8, 150 real wages 41, 229, 252–3 Rebelo, S. 232 regression based approach 38 regulatory influence, growth 258 Reinganum, J.F. 138 Renelt, D. 167, 182, 329 retained earnings 34, 55, 121 returns to scale, decreasing 82–3 Richard, J.-F. 192 risk 15, 78, 130–1, 137, 252 risk premium 82 risk sharing 108 robot stock 140, 143; adjustment speed 147, 149; cross-country statistics 144, 145, 147; diffusion 136; equilibrium 149, 150; investment in 143; prices 142, 150–1, 152; uses 141 Robson, M.H. 292 Romer, P. 206 Rosenberg, N. 183 Rosovsky, H. 184 Rotemberg, J. 267 Rothschild, M. 8, 13, 14, 139 Rowthorn, R. 174, 178, 177, 179n4, 235, 253, 310 Ruane, F. 210 Russell, A. 127 Russell, T. 121 Sachs, Jeffrey 338, 341 Salkever, D.S. 215 Saloner, G. 267 Sargan test 73, 89 Sargent, J.R. 176, 177, 179n5 Savage, D. 268 savings 12, 163, 339 Scherer, F.M. 8, 114, 173, 261–2 Schianterelli, F. 9, 10, 11, 34, 35, 51, 53, 54, 64, 55, 68, 70, 122 Schleifer, A. 124 Schreyrogg, G. 115

Schultze, C.L. 5 Schwalbach, J. 113 SEC, Office of Chief Economist 127 sectoral investment 330–5 Sentance, Andrew 18–19, 21, 38, 43, 44, 268 service-producing sector 331–3, 334, 341 Shannon, D.P. 204 Shapiro, C. 84 share prices, short-termism 126, 127, 128; see also new share issue shareholders 109, 112, 327 Sharp, M. 122 Sheard, P. 116 Sheehan, M. 312 Shepherd, David 253 Shin, Y. 272 Shleifer, A. 107, 114 shocks 137–8 Short, H. 119 short-termism 122–8; corporate culture 324–7; financial constraints 304; financial markets 256–7; fiscal policy 298; investment 107, 125–8, 311; Morris model 128; shares 126, 128; Stein’s model 123 signal jamming 123 Single Market Programme 201, 206 skills, labour 308, 313, 316–17 Small, I. 256 Smith, Adam 165, 173 Smithers, Andrew 251 social partnerships 327 Solimano, A. 84, 136, 138 Solomon, R.F. 216 Solow, R.M. 118, 165 Soteri, Soterios 258, 259 Spain, inward-investment growth 207 spillovers: endogenous growth 182; equipment investment 170–1; foreign-domestic firms 216–17; innovation 184; R&D 165–6, 252; rates of return 178; technology 171 stability 262, 325 Stacey, R.J. 127 stakeholder society 107 Stalk, G. 126 stationarity tests 273 Steinman, H. 115 Stein’s model, short-termism 123 sterling 303 Stiglitz J. 13, 14 Stiglitz, J.E. 36, 87, 121, 139 stock adjustment model of investment 141 stock exchange 125–6 stock market capitalisations 111

215

216

INDEX

Stokey, N.L. 59 Stoneman, P. 21, 138, 140, 252 subsidies for investment 254 Sugden, R. 5 Summers, L. 32, 169, 170, 182, 183, 184–5, 187, 237, 293 Sumner, Michael 9, 254, 256, 262, 292 supply constraints 281–2, 287n2 supply-side policies 238–41, 308, 310–11 Sweden: FDI 202; foreign-owned firms 204 takeover threats 113–14, 324–5 taxation: discounted present value 35, 37; dividends 54, 306; investment 253, 254, 294, 307; Modigliani-Miller theorem 75n12; R&D 313; and retained earnings 121 Taylor, J.B. 9 Taylor, M. 39 Taylor developments, first/second orders 66–7 technical progress: capital-labour ratio 224; capital stock 223; Germany 211–12, 214; labour-augmenting 227 technological change 165, 169, 172, 260–1 technological innovation 172, 183–4, 209–10, 212–13 technology: catch-up 164; diffusion 136, 151; and investment 21, 31, 34; sourcing 204; spillovers 171; uncertainty 138 Temple, Paul 11, 20 Theil, H. 283 Thomas, D.G. 28923 time series analysis 41 Tirole, J. 83 Tobin’s q 18, 31, 41, 67, 72, 126 total factor productivity 213 Tovanen, Otto 21 trade barriers removed 206 trade unions 36–7, 41, 314n9; see also TUC training: employability 317; low paid 324; workforce 325 translog 225–7, 245n6 TUC (Trades Union Congress) 258; corporate governance 326–7; Economic Policy and Social Partnership 326; Partners for Progress 327; Productivity and Social Partnership 320 Turner, Adair 314n8 two-period investment model 16

UK: aggregate production structure 225–32, 241–2; building investment 185, 189; capital stock 302, 319; company law 311; corporate governance 111–12; employment rate 222; equipment investment 186; FDI 200, 202–3; foreign-owned firms 204; GDP 223; growth 256, 322–3; investment-GDP ratio 316, 318; investment growth 207, 329, 330; investment levels 30, 251, 312, 317–20; investment-output ratio 333, 334; labour market changes 207; manufacturing investment 90, 91, 92, 93; overseas operations of companies 22n3; post-war investment 163; price 267; productivity 314n8, 337; public investment 327–8 UN, robot data 140, 143 uncertainty 12–15, 137; aggregate investment thresholds 138; CBI survey 20; demand 83, 87, 95; econometrics 31; indicators 145; industry-led 23n25; inflation 84–5; investment 12–15, 79, 80–4, 95–6, 136–7; manufacturing investment 85, 90, 91, 92, 93; monetary 85–6; output 84–5; price 78–9; and risk 75; technology 138; and volatility 139–40, 145 UNCTAD 199, 203 underemployment 173–4 unemployment: capacity utilisation 259; capital costs 240, 241; equilibrium 244–5; EU 128, 221, 309, 329, 338; and flexibility 311–12; and inflation 224, 238, 309; and investment 174, 221, 309; natural rate hypothesis 221; and productivity 223; social partnership 327; US/Europe 329; see also NAIRU unemployment model, Layard-Nickell-Jackman 223 union model, cointegration 42 Urga, Gioυanni 18–19, 21, 31, 38, 43, 44, 175, 177, 178, 243 US: business in Ireland 208; capital productivity 337;

INDEX

corporate governance 110–11; FDI 200, 202; fiscal policy/investment 293; foreign- owned firms 204; investment-output ratio 333, 334; investment growth 329, 330; investment shares 317; labour market 338; manufacturing industry 86, 90; non-bank affiliates in Europe 208; price 267; productivity 165, 184; unemployment 329 utility function, growth 118 Uzawa, H. 117 Vallès J. 73 Van de Klundert, Th. 258 Variato, A.M. 17 Venables, A.J. 205 Vickery, William 12 Vishny, R. 107, 114, 125 volatility 139–42, 145, 147–8 Wadwhani, S. 125, 127 wage bargaining 174–5, 177, 263n2 wage equation 231, 237 wages 227–32, 258, 308; capital deepening 175–6; cointegrating regression 236, 237; dynamic regression 238; elasticity 230–1; and investment 31, 36–7; low 324; minimum 325; prices 223; and productivity 224–5, 231, 236, 237, 320–1, 338; profitability 321; profits 252–3; real 41, 229, 252–3; restrictions/super-neutrality 244–5; structural 229 Wakelin, K. 211 Wardlow, A. 5, 21 Wass, Douglas 256 Weiss, A. 36, 51, 53, 54, 64, 68, 87, 121 Weitzman, M. 267 Welfare to Work 310, 313 Whited, T.M. 10, 11, 51, 52–3, 63, 64 Whittaker, R. 268 Williams, G.A. 232 Wood, Andrew 170 Woods, R. 38, 43, 44, 88 working conditions, flexibility 312 working hours 325 workplace, social partners 325 World Robot Statistics 152 Wren 253, 254 Wright, R.E. 85

Young, G. 5, 31, 169, 183, 262 Yurko, Allen 251 Zingales, L. 11, 51, 52, 53, 68, 69, 122

217