Macroeconomic Perspectives on the Danish Economy Edited by Torben M. Andersen, Svend E. Hougaard Jensen and Ole Risager
MACROECONOMIC PERSPECTIVES ON THE DANISH ECONOMY
This page intentionally left blank
Macroeconomic Perspectives on the Danish Economy Edited by
Torben M. Andersen Professor of Economics University of Aarhus
Svend E. Hougaard Jensen Head of Research Division Ministry of Business and Industry Copenhagen
and
Ole Risager Professor of Economics Copenhagen Business School
Selection, editorial matter and Chapter 1 © Torben M. Andersen, Svend E. Hougaard Jensen and Ole Risager 1999 Chapter 3 © Stean Nielsen and Ole Risager 1999 Chapter 7 © , Ninette Pilegaard Hasen and Svend E. Hougaard Jensen, Martin Junge 1999 Chapter 9 © Torben M. Andersen and Torben D. Schmidt 1999 Chapters 2, 4–6, 8, 10–11 and comments © Macmillan Press Ltd 1999 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1P 9HE. Any person who does any unauthorised act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The authors have asserted their rights to be identified as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 1999 by MACMILLAN PRESS LTD Houndmills, Basingstoke, Hampshire RG21 6XS and London Companies and representatives throughout the world ISBN 0–333–73331–2 A catalogue record for this book is available from the British Library. This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. 10 08
9 07
8 06
7 05
6 04
5 03
4 02
3 01
2 00
Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham, Wiltshire
1 99
Contents List of Figures
vii
List of Tables
ix
Preface
xi
Acknowledgements
xii
Notes on the Contributors
xiii
1
2
3
4
5
6
Macroeconomic Perspectives on the Danish Economy: Problems, Policies and Prospects Torben M. Andersen, Svend E. Hougaard Jensen and Ole Risager
1
EMU and the Outsider Nations Niels Thygesen
40
Comment by Arne Jon Isachsen Comment by Jesper Jespersen
56 61
Macroeconomic Perspectives on Stock and Bond Investments in Denmark since the First World War Steen Nielsen and Ole Risager
66
Comment by Tom Engsted Comment by Björn Hansson
92 99
Would the Flat Tax Fall Flat in Denmark? Søren Bo Nielsen, Niels Kleis Frederiksen and David Dreyer Lassen
106
Comment by Michael Lundholm
131
The Danish Tax Reform Act of 1993: Effects on the Macroeconomy and on Intergenerational Welfare Kathrine Lange, Lars Haagen Pedersen and Peter Birch Sørensen
134
Comment by Peter Broer
162
Minimum Wage Contracts and Individual Wage Formation: Theory and Evidence from Danish Panel Data Lars Haagen Pedersen, Nina Smith and Peter Stephensen
165
Comment by Donald O. Parsons Comment by Jan Rose Sørensen
188 191
v
vi 7
8
9
10
11
Index
Contents Government Solvency, Social Security and Debt Reduction in Denmark Ninette Pilegaard Hansen, Svend E. Hougaard Jensen and Martin Junge
193
Comment by A. Lans Bovenberg
221
Economic Expansions and Fiscal Contractions: International Evidence and the 1982 Danish Stabilization U. Michael Bergman and Michael M. Hutchison
225
Comment by Frank Barry Comment by Anders Møller Christensen
257 261
The Macroeconomics of the Welfare State Torben M. Andersen and Torben D. Schmidt
264
Comment by Per Callesen Comment by Douglas A. Hibbs, Jr
285 290
The Welfare State and Economic Incentives Jørgen Søndergaard
294
Comment by Torsten Sløk Comment by Thorvaldur Gylfason
325 330
Effort Commitment in Active Labour Market Policy Claus Thustrup Hansen and Torben Tranæs
335
Comment by Niels Blomgren-Hansen Comment by Jørgen Elmeskov
359 363 365
List of Figures 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 1.11 1.12 1.13 1.14 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 C3.1 C3.2 C3.3 4.1 5.1 5.2 5.3 5.4 5.5 5.6 5.7
Real GDP for Denmark, EU and the United States, 1960–97 Real GDP for Denmark, 1973–97 Labour supply and employment, 1960–97 Wage and price inflation, 1972–97 Current account in per cent of GDP, 1960–97 Relative wage and employment, 1948–91 Nominal interest rate and interest rate spread to Germany Nominal and real interest rates Foreign debt and current account Gross investment and saving, 1973–96 General government expenditures, 1960–96 General government revenues, 1960–96 Gross government debt, 1960–96 Projected changes in demographic structures, 1995–2040 Annual nominal stock returns, 1921–95 Real stock price index, 1921–95 Nominal bond yields, 1921–95 5-year nominal equity premium, 1921–90 Bounds on moments of m and sample estimates Perfect foresight stock price decomposition, 1921–95 Dividend yield, 1921–95 Perfect foresight stock prices under the consumption-CAPM and real interest rate discounting, 1921–95 Cumulative real return index for Swedish and Danish stocks from 1922–96 Cumulative real return index for Swedish and Danish long-term government bonds, 1922–96 95 per cent confidence interval for Danish stocks and bonds, constructed from bootstrapping of annual data Tax as percentage of income: single person, no children, Denmark, 1992 Aggregate employment (hours) Aggregate human wealth Aggregate non-human wealth Physical consumption of non-durables Aggregate physical stock of business capital Aggregate physical stock of housing capital Intergenerational welfare effects vii
2 3 4 5 5 10 15 15 17 19 22 23 26 33 69 70 73 78 82 84 85 87 100 101 104 115 148 149 150 151 152 153 154
viii 6.1 7.1 7.2 8.1 8.2
List of Figures
Frequency of wage rates Macroeconomic effects of debt elimination Welfare effects of debt elimination Actual and expected private consumption, 1982:3–1987:4 Historical decomposition of unexpected private consumption attributable to structural shocks, 1982:3–1987:4 9.1 Budget sensitivity and public sector size measured by revenue to GDP ratio, EU countries, 1995 9.2 Public sector size and standard deviation of cyclical measure for total activity, 1971–90 9.3 Growth and public sector size, 1971–90 10.1 Dependence on public support among the 18-66-year-olds, 1960–96 10.2 18-66-year-olds dependent on different types of social security 10.3 Annual working hours (manufacturing) and part-time frequency (all sectors), 1948–94 10.4 Labour market survival rates for active at age 50, 1970 and 1994 10.5 Retirement on early retirement scheme for unemployed aged 50–59, 1994:1–1996:4 10.6 Participation in leave schemes, 1994:1–1996:4 10.7 Index of log GDP per capita, 1884–94 C10.1 Equality and efficiency C10.2 Size of government and growth 11.1 Beveridge curve for Denmark, 1980–96 11.2 The basic incentive problem 11.3 The effort–benefit system 11.4 Productivity below the threshold 11.5 Wage setting
171 209 211 251 252 268 268 275 305 306 309 312 313 314 321 331 333 342 344 345 349 353
List of Tables 1.1 1.2 1.3 1.4 2A.1 C2.1 3.1 3.2 3.3 3.4 3.5 3.6 C3.1 C3.2 C3.3 C3.4 4.1 4.2 4.3 5.1 5.2 5.3 5.4 6.1 6.2 6.3 7.1
Average real GDP growth, 1960–97 Unemployment rates for 25–64-year-olds, by level of educational attainment, 1995 Structure of the tax system in Denmark, EU and USA Working-age population, employment and recipients of public transfer payments, 1960–97 Reserves and monetary base To be or not to be a member of EMU Average annual stock market returns, 1922–95 Average 5- and 10-year bond and stock returns, 1922–95 Return–risk of stock and bond portfolios, 1922–95 Equity premium for the sample Variance–covariance between annual returns, yields and consumption growth rate, 1922–95 What amount, X, would make a person with CRRA utility indifferent? Annual real returns, 75 observations, 1922–96 Real returns per annum for 5-year investment horizons, 71 overlapping observations, 1922–96 Real returns per annum for 10-year investment horizons, 66 overlapping observations, 1922–96 Real returns per annum for 20-year investment horizons, 56 overlapping observations,1922–96 Flat tax rates and personal allowances Changes in disposable income, by flat-tax income groups Changes in disposable income, by gross income groups Statutory marginal tax rates on labour income, before and after the 1993 tax reform Statutory marginal tax rates on capital income, before and after the 1993 tax reform Estimated effective tax and subsidy rates, before and after the 1993 tax reform Calibration of the EPRU model Sample means in 1990 Test of instruments Fixed-effect estimation of wage function Principal budget components, unemployment, interest and growth rates, 1960–95 ix
2 10 25 29 53 56 70 73 74 77 79 84 101 102 103 103 112 121 123 135 136 138 145 180 182 184 195
x 7.2 7.3 7.4 8.1 8.2
List of Tables
Alternative measures of public debt and the deficit, 1995 Magnitude of required fiscal adjustment Generational accounts of current and future generations Fiscal consolidation and macroeconomic performance Structure of fiscal consolidation and macroeconomic conditions prior to the first year of consolidation 8.3 Danish fiscal adjustments following the October 1982 announcement 8.4 Forecasts published by Economic Council, 1982–7 9.1 Government spending and revenues: international comparison, 1993 9.2 Cyclical sensitivity of budget revenues and expenditures in the EC, 1995 9.3 Cyclical properties of tax rates and social security transfers 9.4 Variance decomposition ∆y 9.5 Regressions of volatility in private sector activity on various measures of taxes and social security transfers 9.6 Public debt, transfers and public consumption 10.1 Characteristics of the Danish system of income transfers 10.2 Combined effective marginal tax rates, 1996 10.3 Welfare state expenditures, 1960, 1980 and 1995 10.4 Expected welfare state incentive effects on labour force participation 10.5 Reported attitudes and search activity among unemployed with at least 3 months of unemployment, 1994 C10.1 Denmark, Sweden and the world: investment, trade and growth
200 201 205 234 236 243 247 265 267 270 271 273 280 298 303 304 310 318 332
Preface This volume contains the proceedings of the conference ‘Macroeconomic Perspectives on the Danish Economy’, held in Hornbaek, Denmark (19–20 June 1997). The conference was jointly organized by the Economic Policy Research Unit, University of Copenhagen and the Department of Economics, University of Aarhus. The book contains 11 chapters, including an introductory survey on the Danish economy. The aim of each chapter has been to illuminate an important policy issue confronting the Danish economy, using the insights of up-todate methodology and the most recent evidence. Each chapter is followed by comments from a leading international expert. This should, we hope, make the book of interest also to readers outside Denmark. In any case, many European countries are faced by problems and prospects similar to those addressed in the book. Several chapters have also benefited from the perspectives offered by Danish economists responsible for the practical conduct of macroeconomic policies. In editing the book we have, therefore, had a broad readership in mind: our own colleagues in academia, civil servants involved with day-to-day aspects of the Danish economy, economists in other countries working with related issues and last, but not least, students of economics who feel a need to bridge the gap between topics addressed in economic theory and the so-called real world. To put it more bluntly: it is our ambition that this volume will be an indispensable text to anyone who is seriously interested in the Danish economy. Our principal debt is to the contributors and the conference participants, who have all helped making us feel that this project has been a worthwhile undertaking. For financial support we are grateful to the Danish National Research Foundation who provides the prime funding of EPRU. Thanks are also due to the Danish Social Research Council through its funding of the project ‘The Welfare State: Threats, Problems and Possible Solutions’, which has initiated the research underlying some of the papers in the book. Grethe Mark, EPRU’s administrative secretary, has handled the logistics of the conference with professionalism and she also supervised the conference’s financial arrangements. Finally, competent assistance by Christoffer Kok Sørensen and Ian Valsted has been essential to the book’s swift publication. Copenhagen
TORBEN M. ANDERSEN SVEND E. HOUGAARD JENSEN OLE RISAGER
xi
Acknowledgements The editors and publishers acknowledge with thanks permission from the Journal of Financial Economics for permission to reproduce Table 3.6, from N.G. Mankiw and S.P. Zeldes, ‘The consumption of stockholders and nonstockholders’ (1991).
xii
Notes on the Contributors Authors Torben M. Andersen, Professor, Institute of Economics, University of Aarhus and Research Fellow, Centre for Economic Policy Research, London. U. Michael Bergman, Associate Professor, Department of Economics, University of Lund, Sweden. Niels Kleis Frederiksen, PhD student, Economic Policy Research Unit, University of Copenhagen. Ninette Pilegaard Hansen, Economist, Computable General Equilibrium Modelling, Statistics Denmark. Claus Thustrup Hansen, Assistant Professor, Department of Economics, University of Copenhagen and Research Fellow, Economic Policy Research Unit, University of Copenhagen Michael M. Hutchison, Professor, Department of Economics, University of California at Santa Cruz and Research Fellow, Economic Policy Research Unit, University of Copenhagen. Svend E. Hougaard Jensen, Head, Division of Economic Analysis, Ministry of Business and Industry and Research Fellow, Economic Policy Research Unit, University of Copenhagen. Martin Junge, Research Assistant, Department of Economics, University of Copenhagen. Kathrine Lange, Research Assistant, Economic Policy Research Unit, University of Copenhagen. David Dreyer Lassen, Research Assistant, Economic Policy Research Unit, University of Copenhagen. Søren Bo Nielsen, Research Professor, Economic Policy Research Unit, University of Copenhagen and Research Fellow, Centre for Economic Policy Research, London. Steen Nielsen, PhD student, Economic Policy Research Unit, University of Copenhagen. Lars Haagen Pedersen, Head, Computable General Equilibrium Modelling, Statistics Denmark.
xiii
xiv
Notes on the Contributors
Ole Risager, Professor, Department of Economics, Copenhagen Business School and Research Fellow, Economic Policy Research Unit, University of Copenhagen. Torben D. Schmidt, PhD student, Institute of Economics, University of Aarhus. Nina Smith, Professor, Department of Economics, Aarhus School of Business, Research Fellow, Center for Labour Market and Social Studies, University of Aarhus and Chairman, Council of Economic Advisors, Denmark. Peter Birch Sørensen, Professor, Department of Economics, University of Copenhagen, Research Fellow, Economic Policy Research Unit, University of Copenhagen and Chairman, Council of Economic Advisors, Denmark. Peter Stephensen, Economist, Computable General Equilibrium Modelling, Statistics Denmark. Jørgen Søndergaard, Director, Danish National Institute of Social Research, Copenhagen. Niels Thygesen, Professor, Department of Economics, University of Copenhagen and Research Fellow, Economic Policy Research Unit, University of Copenhagen. Torben Tranæs, Associate Professor, Department of Economics, University of Copenhagen and Research Fellow, Economic Policy Research Unit, University of Copenhagen. Discussants Frank Barry, Professor, Department of Economics, Trinity College, Dublin, Ireland. Niels Blomgren-Hansen, Professor, Department of Economics, Copenhagen Business School. A. Lans Bovenberg, Professor, CentER, Tilburg University, the Netherlands, Research Fellow, OCFEB, Erasmus University, the Netherlands and Research Fellow Centre for Economic Policy Research, London. Peter Broer, Economist, Netherlands Bureau for Economic Policy Analysis and Research Fellow, OCFEB, Erasmus University, the Netherlands. Per Callesen, Deputy Permanent Secretary, Ministry of Finance, Denmark. Anders Møller Christensen, Vice Director, Danmarks Nationalbank. Jørgen Elmeskov, Chief Economist, OECD, Paris. Tom Engsted, Associate Professor, Department of Finance, Aarhus School of Business.
Notes on the Contributors
xv
Thorvaldur Gylfason, Professor, Department of Economics, University of Iceland and Research Fellow, Centre for Economic Policy Research, London. Björn Hansson, Docent, Department of Economics, University of Lund, Sweden. Douglas A. Hibbs, Professor, Department of Economics, University of Gothenburg, Sweden. Arne Jon Isachsen, Professor, Centre for Research in Monetary Economics, Handelshøyskolen BI, Oslo, Norway. Jesper Jespersen, Professor, Department of Economics, Roskilde University Centre. Michael Lundholm, Professor, Department of Economics, University of Uppsala, Sweden. Donald O. Parsons, Professor, Department of Economics, George Washington University, USA. Torsten Sløk, Economist, Ministry of Business and Industry, Denmark. Jan Rose Sørensen, Associate Professor, Institute of Economics, University of Aarhus.
This page intentionally left blank
1 Macroeconomic Perspectives on the Danish Economy: Problems, Policies and Prospects1 Torben M. Andersen, Svend E. Hougaard Jensen and Ole Risager 1.1
INTRODUCTION: TRENDS AND CYCLES
Countries are often ranked according to their GDP per capita, properly adjusted for differences in price levels. Based on the purchasing power parity (PPP) adjusted GDP statistics, Denmark is one of the most affluent societies in the world. Compared to the EU average and the OECD average, Denmark was, respectively, 12 and 9 per cent ahead in 1995, and only a few countries B including Luxembourg, Norway, Switzerland and the United States – had a higher GDP per capita (The Economist, 1997). The level of real GDP was in 1997 about three times as large as in 1950 and, from the perspective of economic welfare, it is also important to note that over the same time period the average annual working hours have fallen by about a third. However, Denmark’s GDP growth rate has been slightly lower than in a number of other countries. Figure 1.1 shows real GDP for Denmark, (the average of) 15 EU countries and the United States; in all three cases real GDP is indexed to 100 in 1960. Over the whole period 1960–97, Denmark enjoyed an average real GDP growth rate of 2.8 per cent. With the corresponding number being 3.0 per cent for both the EU and the United States (see Table 1.1), the difference in growth performance has been fairly small. However, if this difference also persists into the next century, Denmark will of course not be able to retain its high position among the world’s richest countries. It is also remarkable that the two sub-periods, 1960–72 and 1973–97, respectively, are characterized by very different growth performances. Indeed, for Denmark and the EU as a whole the GDP growth rates in the second sub-period were only half as high as in the first. While US growth rates have also fallen, the decline in Europe is much more dramatic, a phenomenon sometimes referred to as ‘eurosclerosis’. 1
Andersen, Jensen and Risager
2 Figure 1.1
Real GDP for Denmark, EU and USA, 1960–97
300
250
200
150
100
50 1960
1963
1966
1969
1972
1975
Denmark
1978 EU15
1981
1984
1987
1990
1993
1996
USA
Note: Index numbers (1960 = 100). Source: OECD.
The slow growth period 1973–97 was characterized by deep recessions but also by impressive upswings. As indicated by Figure 1.2, this period saw three recessions and two expansionary periods with growth significantly below and above the mean, respectively. The first two recessionary periods occurred in the mid-1970s and at the beginning of the 1980s, and were mainly triggered by well known international events such as the two OPEC shocks and the associated economic policies adopted by the leading OECD economies. The two expansionary periods – one from 1983 to 1987, and another from 1993 until today – were, however, mainly driven by domestic factors, as we shall explain in more detail below. In both upturns, economic policy has played an important role in triggering the upswing in the economy. Economic policy also played a role in Table 1.1
Denmark EU USA Source: OECD.
Average real GDP growth 1960–97 per cent 1960–97
1960–72
1973–97
2.8 3.0 3.0
4.4 4.6 3.8
2.1 2.3 2.6
Problems, Policies and Prospects Figure 1.2
3
Real GDP for Denmark, 1973–97
280 260
Contractionary policy: Potato Diet
240 Expansionary policy: Tax cuts
220 200
OPEC II OPEC I New non-accomodative non-accommodative New policy policy regime
180 160 1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
Note: Index numbers (1960 = 100). Source: OECD.
setting off the recession that started in 1987 and continued until 1993, a period in which Denmark’s major trading partners were booming, at least until the late 1980s. Hence, it is interesting to note that domestic factors and Danish economic policy in particular have played an important role in the performance of the economy.2 However, it is also important to stress that owing to the economy’s fairly high degree of openness – as indicated by an export share of GDP around 35 per cent as well as strong financial linkages with the rest of the world – the economy is of course also vulnerable to international shocks. The international recessions associated with the two OPEC shocks are cases in point. As shown by Figure 1.3, these shocks were followed by sharp declines in employment, while labour supply continued to grow, actually at an increasing rate following the first OPEC shock. The expansionary period in the mid1980s and the current boom starting in 1993 were both associated with sharp increases in employment, but it has not so far been possible to bring the unemployment rate down to the level prevailing prior to the adverse shocks of the 1970s. Over the years, a number of measures have therefore been adopted to cope with the unemployment problem. Most of the measures have been passive in nature, aiming at reducing the registered unemployment rate by lowering labour supply through early retirement schemes, paid leave schemes and so forth. More recently, initiatives have been taken which directly aim at bringing the unemployed back to work again (see section 1.3). The persistent high level of unemployment has had strong effects on the government budget, as elaborated in section 1.6. This is to be expected for a welfare state like the Danish one since a key objective is to moderate the
Andersen, Jensen and Risager
4 Figure 1.3
Labour supply and employment in heads, 1960–97
3100 2900 2700 2500 2300 2100 1900 1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996
Labour supply
Employment
Note: Numbers are in thousands. Figures for 1997 are preliminary. Source: Statistics Denmark and Ministry of Economic Affairs.
effects of recessions. There has thus been a rapid increase in public sector transfer payments, while traditional public sector activities have not increased relative to GDP since the 1970s. The number of individuals living on UI benefits and other public transfer payments has thus increased enormously (see section 1.7). Although the tax burden also has been increasing, there has been a persistent gap between revenues and expenditures. This has, of course, led to increasing government debt over the years. While the current government debt-to-GDP ratio is below the average in the EU, it remains a fact that the debt and the burden of debt service are significant. Moreover, it may be argued that the government finances are intergenerationally imbalanced, suggesting that significant bills will be passed onto future generations. The extent to which this is the case is discussed further in section 1.8. The high growth in the 1960s led to a gradual build-up of goods and service price inflation which culminated in the mid-1970s owing to the sharp increase in energy prices and the wage hikes (see Figure 1.4). Following these events, inflation declined but both price and wage inflation remained high until the beginning of the 1980s. In late 1982, there was a shift in economic policy towards a disinflationary policy strategy that helped bringing inflation under control. This is addressed in greater detail in section 1.4. An important element of this non-accommodative policy strategy was the shift towards a fixed exchange rate policy. Earlier, policy makers had quite often resorted to currency devaluations in an attempt to offset the deteriorations in competitiveness and the current account that resulted
Problems, Policies and Prospects Figure 1.4
5
Wage and price inflation, 1972–97
25.0%
20.0%
15.0%
10.0%
5.0%
0.0% 1972
1974
1976
1978
1980
1982
1984
1986
1988
% Increase Increase in in Hourly hourly wage % Wage ininmanufacturing Manufacturing
1990
1992
1994
1996
% Increase in GDP deflator
Source: OECD.
from high wage and energy price increases. Indeed, as we discuss at length in section 1.5, the external (im)balance was for long seen as the Achilles’ heel in the Danish economy. This is witnessed by the fact that there was a deficit on the current account in each of the consecutive years from 1964 to 1989 (see Figure 1.5).
Figure 1.5
Current account in per cent of GDP, 1960–97
4% 3% 2% 1%
-2% –2% -3% –3% -4% –4% -5% –5% -6% –6%
Source: Statistics Denmark.
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
0% –1% -1%
6
Andersen, Jensen and Risager
As a background to the more detailed treatment of selected issues of importance for the macroeconomic performance of the Danish economy, section 1.2 provides a brief overview of the different economic policy strategies that have been pursued in recent decades.
1.2
GENERAL ECONOMIC POLICY STRATEGIES
The growth period in the 1960s led to an optimistic view about the possibilities for fine-tuning the macroeconomy. The subsequent developments showed that macroeconomic management was very difficult and a variety of different policy strategies were pursued in an attempt to improve upon macroeconomic performance. The initial response to the international recession and the rise in unemployment in the mid-1970s was a so-called ‘bridging policy’. This policy was based on the perception that the recession was temporary and could therefore be mitigated by an expansionary demand management policy. The perception that the recession was temporary soon proved wrong, and the overall policy strategy from the mid-1970s up to 1982 was guided by a so-called ‘switching strategy’, according to which the twin problems of unemployment and current account deficits could be solved by switching private demand with a high import ratio into public demand with a low import ratio. As a consequence, numerous policy packages included tax increases to curtail private demand and expansions of public demand and in particular public employment programmes. These changes did not necessarily take place simultaneously as the policy had a ‘stop–go’ character with tax increases in times where the current account was considered out of control and public expansion in periods with focus on the unemployment problem. This was also the prime time for policy packages, appearing frequently in an atmosphere of acute need for intervention. A loose exchange rate policy (crawling peg) was part of this strategy since discrete devaluations were believed to undo the harm caused by excessive domestic wage increases. Underlying this policy view was an accommodative attitude towards the labour market in the sense that the government accepted a responsibility for the employment level. Accordingly, the labour market policy was fairly passive, mainly concentrating on reducing labour supply and mitigating the economic consequences for the unemployed. This policy strategy was not very successful in terms of solving the overall macro problems. The problems of current account deficits and unemployment remained unsolved, while inflation and public debt were growing. In the autumn of 1982, a centre-right minority coalition replaced a Social Democratic minority government. The new government launched a new economic programme based on the idea that the problems underlying the
Problems, Policies and Prospects
7
Danish economy could be solved in a medium-term perspective only by an expansion of private sectors capable of facing international competition. The programme consisted of a disinflationary policy grounded in a fixed exchange rate policy and a tight fiscal policy. The policy shift signalled that the government was attempting to leave more responsibility for the determination of employment with the labour market. This disinflationary policy was initially very successful and a change in the economic mood was observed accompanied by increases in private demand, higher GDP growth and a fall in unemployment from 1983 to 1986 (see Figure 1.7, p. 15). This policy was to some extent killed by its own success (see Andersen, 1994; and Bergman and Hutchison, Chapter 8 in this volume). The tremendous upturn in the mid-1980s was associated with a substantial increase in domestic demand as both private consumption and investment soared, but it turned out that supply capacity could not match this increase. The capacity problems applied both to the production capacity of firms and the availability of labour with the necessary qualifications. As a consequence, it led to a further deterioration of the current account and wage increases (in 1987) much in excess of the norms of the disinflationary policy. The development over this period was interesting since domestic demand was booming. Therefore, shortage of aggregate demand could be ruled out as the primary reason for the unemployment problem. The fact that substantial wage increases were released at an unemployment rate around 8 per cent was taken as strong indication that there were severe adjustment problems in the labour market. A fairly pessimistic view on the unemployment problem developed during the late 1980s and early 1990s, and the view that ‘we have to live with unemployment’ and ‘there is no need for so much labour’ became fairly widespread. A minority coalition government led by the Social Democrats took over in 1993. Among its new initiatives was a paid leave scheme aiming at distributing the burden of unemployment more equally. It also launched a fiscal expansion, partly financing the transition period of a tax reform designed so as to lower marginal tax rates and broaden the tax base. Although a shift in the business cycle was to be expected after a prolonged recession, there is little doubt that the expansionary fiscal policy had a role in triggering the boom in economic activity and the fall in unemployment. The accompanying increase in employment motivated a more active labour market policy focusing much more explicitly on the incentive effects of various labour market policies. The policy shift towards youth – involving a tightening of the incentive structure – is a prime example of this. There is some indication that the wage formation process has been affected, but it is still too early to judge how fundamental these changes are. Nonetheless,
8
Andersen, Jensen and Risager
recent developments have shown that a combination of demand and active supply measures – a two-handed approach – can be successfully used to boost employment.
1.3 WAGES AND STRUCTURAL SHIFTS IN THE LABOUR MARKET Denmark experienced increasing wage inflation during the 1960s. In the mid1970s, wage inflation reached a climax as nominal wages increased by more than 20 per cent, see Figure 1.4. This led to an increase in real (consumption) wages by almost 10 per cent. Since then wage inflation has gradually declined. The major downward adjustment occurred from 1982 to 1983–4, where the annual wage increases fell from around 10 to less than 5 per cent. The considerable decline in wage inflation shows that the shift towards the low-inflation, non-accommodative policy stance in late 1982 gained a fairly high degree of credibility rather quickly among the wage setters in the economy. Since then wage inflation has been fairly low. The only exception to this rule is 1987, where there was a revitalization of wage inflation primarily due to rapid growth and increasing employment, which coincided with a new round of collective wage bargaining. Due to the wage hike in 1987 and the tightening of economic policy, unemployment increased substantially whereas wage inflation declined. It is only very recently that employment has started to rise again; from the beginning of 1993 to late 1997, total employment increased by around 150,000 persons or by almost 8 per cent. The considerable rise in employment that has taken place since 1994 has not so far led to an accelerating wage inflation, which is somewhat surprising given past experience. There are several potential explanations for this. First, it should be noted that the present employment increase is not as big as the one observed in the mid-1980s, where private sector employment grew by almost 200,000 persons. Second, the coincidence of a low and stable inflation and falling unemployment may simply reflect a learning process. Thus, the experience in the 1980s clearly demonstrates that pressing for ‘too’ high wage increases is futile as the only certain outcome is higher unemployment and very little else. Third, fierce competition in goods and service markets associated with the increased international integration process, often noted by the business community, may also have helped to hold back wage and price increases. Indeed, since the coincidence of low inflation and falling unemployment is also a world-wide phenomenon, a truly international explanation has some appeal. Fourth, the labour market reforms outlined below may also have contributed, as argued by Hansen and Tranæs in Chapter 11. These reforms, undertaken in 1994 and 1996,
Problems, Policies and Prospects
9
represent a gradual tightening of previous policies (see also BlomgrenHansen’s comments on Chapter 11). There are several elements in these reforms. First, there is an increased attempt to avoid mismatch and bottlenecks. Thus, regional labour market councils have been endowed with additional instruments in order to avoid the fallacies of the mid-1980s. Second, there has been a tightening of benefit eligibility requirements. Access to UI benefits now requires 52 weeks of work over the last three years as opposed to 26 weeks before. The maximum period on benefits has been reduced to five years. Third, insured unemployed persons can no longer extend their period on UI benefits through job training, and uninsured unemployed persons (outside the UI benefit system) can no longer earn their right to UI benefits through subsidized jobs provided by local authorities. The previously existing possibility of receiving UI benefits ‘for life’ thus no longer exists. Fourth, work availability requirements and enforcement have also been tightened in general and in particular as regards young unemployed workers below 25 years of age. The new rules stipulate that benefits should be activated after an unemployment spell of less than six months, and at a 50 per cent reduction of the benefit level. Finally, each unemployed is, in collaboration with the local authority, supposed to work out a plan for education or job training that eventually will lead to an exit from unemployment. Another interesting feature of the wage formation process is the compression of wages that occurred until the beginning of the 1980s. To illustrate this phenomenon in a simple way, Figure 1.6 plots the relative wage between unskilled and skilled male workers over the period 1948–91. The diagram also shows the relative employment level between the two types of labour measured along the right-hand axis. In the period from 1964 to the beginning of the 1980s, the relative wage between unskilled and skilled labour increased – that is, wage differentials declined. Declining wage differentials were, however, also associated with declining job opportunities for the unskilled workers insofar as the relative employment level declined. As unskilled labour became more expensive in relative terms, firms reduced their demand for unskilled workers and increased their demand for skilled labour (see Risager, 1993). This study therefore lends support to the view that the concentration of high unemployment among unskilled and inexperienced workers is in part due to the compressed wage structure. The unequal unemployment distribution is documented in Table 1.2. There is clear evidence that Denmark suffers in particular from high unemployment among low-skill groups, but it also appears that unequal unemployment distribution is an international phenomenon. It is therefore likely that there are other important explanations to which we return when we have further highlighted the role of wages.
Andersen, Jensen and Risager
10
1.8
0.9
1.7
0.88
1.6
0.86
1.5
0.84 1.4 0.82 1.3
0.8
1.2
0.78
Relative wage
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
1958
1956
1954
1
1952
1.1
0.74
1950
0.76
Unskilled employment/Skilled employment
Relative wage and employment, 1948–1991
0.92
1948
Wage for unskilled/Wage for skilled
Figure 1.6
Relative employment
Note: Employment figures are only available only 1991 due to a new wage– employment survey method. Source: Risager (1993) and Danish Employers’ Federation.
Jensen, Rosenholm and Smith (1994) also show that part of the unemployment problem in Denmark has to do with too high wages, and in particular with too high entrant wages. Their statistical analysis shows that 25 per cent of young unemployed men and 15 per cent of young unemployed female workers are unable to live up to the entrant wage – that is, their productivity is not high enough – and that is why firms are unwilling to employ them. However, they also show that a fall in the entrant wage by 10 per cent Table 1.2
Countries Austria Denmark Germany Netherlands USA
Unemployment rates for 25–64-year-olds, by level of educational attainment, 1995 Below upper secondary education 5.7 14.6 13.3 7.9 10.0
Source: OECD (1997a).
Upper secondary education 2.9 8.3 7.9 4.8 5.0
Non-university tertiary education 1.4 5.3 5.2 na 3.6
Universitylevel education 2.1 4.3 4.7 4.1 2.5
All levels of education 3.5 10.0 8.1 5.6 4.7
Problems, Policies and Prospects
11
will lead to a substantial increase in employment among this group of workers. Proposals of this kind have been ‘vetoed’ by the trade unions. Our interpretation of their attitude is inspired by contemporary economic labour market theory: as a fall in entrant wages is likely also to reduce wage growth for the incumbent workers due to increased outsider competition, a fall in entrant wages is likely to lead to a loss for the majority of the incumbent workers. Against this, we have the gain associated with the increase in employment and income for those who change status from unemployed to employed. However, as the latter group is a minority amongst the members of the unions, the trade union policy stance seems clear. Another concern has been the extent to which a reduction in entrant wages will lead to social imbalances. It is therefore of interest briefly to discuss the income distribution effects of a fall in entry wages. In the short term, income distribution among the employed workers will obviously become more unequal, whereas the distribution within the potential workforce is a more complicated matter as those workers who move from unemployment into employment will experience an increase in income, provided that UI benefits also are reduced in line with entrant wages. In order to get a feeling for the long-term effects, it is important to know something about wage dynamics – that is, the speed by which workers move up the income ladder. Bingley, Bjørn and Westergård-Nielsen (1994) have studied this issue by following a group of workers belonging to the lowest income decile in 1981. After five years, 75 per cent of those workers had climbed up the ladder. Hence, there were only 25 per cent left in the low wage decile. More than 55 per cent had reached the third decile or higher plateaus in the earnings hierarchy. The fairly high degree of wage mobility shows that ‘low pay status’ is not a life-time stigma for the vast majority of the wage earners. The introduction of low entrant wages will therefore provide opportunities to many outsiders who are unable at present to get a job. Needless to say, there will always be a small group who are unable to move up the ladder. Many of these persons are now squeezed out of the market, living on UI benefits or other transfer payments. Finally, as regards the functional income distribution, a stretchning of the income distribution towards a higher proportion of incomes just below the present minimum wage will of course increase the return to capital. Because the UI system sets a maximum benefit level, UI benefits are of greater value to the low-paid than to the high-paid unemployed workers, and it is therefore also likely that the UI system is an important explanation of the narrow wage differentials (see Andersen and Risager, 1990, and Hansen, Pedersen and Sløk, 1996). Pedersen, Smith and Stephensen (Chapter 6 in this volume) estimate wage equations using panel data, and they also find
12
Andersen, Jensen and Risager
that UI benefits affect workers’ wage demands – as indeed any labour market model would suggest. Moreover, the effects are largest for low-income groups. Since probably all political parties in Denmark – at least those likely to take office – seem unwilling to cut the unemployment benefit level in a significant way, as recommended by, for example, the OECD (1997b), the option ‘reduce benefits–lower wages–increase employment’ is not really on the policy agenda. In this context it should, however, be noted that the OECD’s summary measure of benefit generosity shows that Denmark has the most generous system among all 21 countries in the OECD (1997b) study. A recent study by the Danish Economic Council (Det Økonomiske Råd, 1997) also shows that a large proportion of low-skilled/low-paid have no incentives to get employed when account is taken of taxes and transfers. Thus, more than 40 per cent of those who are unemployed for more than 80 per cent of the year, have an effective replacement ratio that exceeds 90 per cent. Moreover, there are also many low-paid employed workers who have no economic incentive to work. The philosophy of the present government is to preserve the high benefit level while at the same time increasing the demands that society puts on those who collect benefits and welfare money in general. In this context, it is interesting to note that following the tightening of the work availabilty and benefit eligibility conditions for young unemployed workers mentioned earlier, the unemployment problem for this group of workers disappeared rather quickly; it is probably important to note that the elimination of the youth unemployment problem occurred simultaneously with an upswing in the economy. Finally, it should be mentioned that the generous UI system has made it easy for firms to lay-off workers not only in recessions but also at other times where firms find it desirable to reduce the workforce. That feature of the system is one among several reasons why the Danish hiring–firing system is highly flexible in an international context (see also Emerson, 1988). The latter aspect may also help to explain why employers’ federations seem to have accepted the high UI benefit level in spite of the fact that it increases firms’ direct wage costs. The concentration of unemployment among low-skill groups is likely to reflect a compressed wage structure, but other factors are probably also at play, as indicated by the simple observation that other countries like the United States, characterized by a very unequal wage distribution, also to some extent suffer from this problem. Technological progress may play a role insofar as investments in new technology may lead to an increase in the demand for skilled and highly educated persons relative to unskilled workers. Unfortunately, it is hard to measure technical progress and how it has changed over time. This clearly complicates testing this hypothesis (see, however, Economic Council, 1997).
Problems, Policies and Prospects
13
Owing to the unwillingness to implement policies that make lowproductivity workers more competitive via the (wage) price mechanism and to the view that part or probably most of the unemployment problem is caused by technical progress and an inadequate skill structure, increased emphasis has been put on making the unemployed more competitive via education and training efforts. Whether the strong emphasis on education and training is enough to bring about a substantial reduction of the structural unemployment problem is an issue that goes beyond the scope of this chapter. However, there is evidence which indicates that some of these initiatives have very little impact. For example, a statistical study of the (unemployed) individuals who participated in the so-called AMU (labour market training) programmes shows that the probability that they get a job afterwards is not higher than non-participants’ probability of finding a job, controlling for a number of other factors (see Jensen and Jensen, 1996). The social return of this targeted but fairly expensive investment in education programmes has therefore been questioned. Another policy option that has been discussed recently is to give low-income individuals better incentives to participate actively in the labour market through a so-called earned income tax credit (see Sørensen, 1997, and Danish Economic Council, 1997). In recent years, the Danish collective wage bargaining system has changed from being a mixture of centralized and decentralized bargaining to a predominantly decentralized system. Thus, since the beginning of the 1990s almost 90 per cent of the workers in the manual labour market have been on the so-called minimum wage system. Under this system, trade unions and their counterparts on the employer side negotiate a base wage level. The minimum wage level varies across occupations. To give some indication, the average Danish base wage level is roughly twice as high as the legislated minimum wage in the United States. On top of that, there is a supplement negotiated at the local level between management and workers. The bottom line is thus that entrant wages in firms may be considerable above the minimum wage level. The movement towards more decentralized wage bargaining has so far not led to an increase in wage differentials across workers. In view of the fact that the public sector is a large employer, it should be noted that wage negotiations in the public sector often follow private negotiations. Moreover, the agreements usually mimic the agreements in the private market – that is, certain wage indexation clauses attempt to link wage increases in the public sector to the wage increases granted in the local private negotiations. For more than a decade, wages in the public sector have, however, declined relative to the private sector. In the public sector, there has also been an attempt to increase the element of decentralized wage negotiations and to allow for more flexibility in the pay schedule.
14 1.4
Andersen, Jensen and Risager EXCHANGE RATES, PRICES AND INTEREST RATES
Denmark has over the years taken part in various schemes for monetary cooperation in Europe, starting with the so-called ‘snake’ arrangement set up in 1972, and at present Denmark participates in the exchange rate mechanism (ERM) within the European Monetary System (EMS). 3 The official position has for long been that Denmark pursues a fixed exchange rate policy, but what that really means has changed over the years. In particular in the late 1970s a crawling peg policy was effectively followed as the Danish government undertook four discrete devaluations in the period 1979–82. The crawling peg era was abandoned in late 1982 when the new government declared a fixed exchange rate policy as an integral part of its disinflationary policy. This policy remains in force and enjoys broad political support. Although the precise meaning of the fixed exchange rate policy has never been explicitly defined it is understood to have the D-mark as its reference point. The shifts in exchange rate policy are clearly mirrored in the development of inflation (Figure 1.5) and in nominal interest rates (Figure 1.7). Indeed, nominal interest rates peaked in 1982 and fell fairly rapidly after the initiation of the disinflationary policy based on a fixed exchange rate policy (see Andersen and Risager, 1988). Since then inflation has remained at a low level below the European average. That the fall in nominal interest rates is tightly related to the exchange rate policy is also seen from Figure 1.7, which shows the interest rate spread between Denmark and Germany. It is apparent that along with the establishment of a credible fixed exchange rate policy and, accordingly, an inflation convergence to the German standard, the interest rate spread has diminished dramatically. The fact that there in periods has been a moderate interest rate spread between Denmark and Germany can be taken to reflect the fact that financial markets do not rule out the possibility of a return to a more inflationary policy, in particular given the Danish opt-out position on Economic and Monetary Union (EMU). Although nominal interest rates have fallen rapidly, this is not the case for real interest rates, as revealed by Figure 1.8. The real after-tax interest rate is still fairly high, but has nevertheless fallen since 1993. The fairly high interest rates that have prevailed since the late 1970s is an important explanation why the excess return on stocks relative to bonds has been modest. However, in the last 30 years, the market portfolio of stocks has outperformed bonds, but not by as much as in many other countries (see Nielsen and Risager, Chapter 3 in this volume, and the comments by Björn Hansson). The shift towards greater ‘fixity’ in the conduct of exchange rate policy has gone hand in hand with initiatives to deregulate capital markets. Thus, capital markets have been gradually liberalized since the late 1950s, both
Problems, Policies and Prospects Figure 1.7
15
Nominal interest rate and interest rate spread to Germany, 1969–97
25
Interest rate/Spread
20
15
10
5
0 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997
Nominal interest rate Nominal Interest Rate
Interest rate spread Denmark-Germany Interest Rate Spread Denmark-Germany
Note: Nominal interest rate: effective rate of return on bonds (iwbz), interest spread: nominal interest rate less effective rate of return on long-term bonds in Germany (iwbdm). Source: ADAM databank, Statistics Denmark.
Figure 1.8
Nominal and real rate rates of interest, 1951–97
25
Interest
rate
20 15 10 5 0 –5 -5 1951 1954 1957 1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996
Nominal interest rate
Real interest rate
Real after tax interest rate
Note: Nominal interest rate: effective rate of return on bonds (iwbz), real rate of interest: nominal interest rate less actual inflation measured by the GDP deflator, real after-tax interest rate: nominal interest rate after tax (tsa0u) less actual inflation. Source: ADAM databank, Statistics Denmark.
16
Andersen, Jensen and Risager
domestically and towards the international capital markets. These reforms were completed in the early 1980s when practically all legal restrictions on international capital mobility (in and out of Denmark) were eliminated. Rules and regulations for financial intermediaries have also been liberalized, implying that binding credit constraints for private households have been eased. These liberalizations were important for not only the reduction in the interest rate spread but also for the fall in interest rates. As the interest rate spread is primarily determined by exchange rate expectations, there would be no spread with fully liberalized capital movements and a credible fixed exchange rate. Financial factors have been important for the expansionary period in both the 1980s and the 1990s. In both periods there was a tremendous increase in loan-financed consumption. In the latter period house-owners had an opportunity to make loan conversions to exploit the falling nominal interest rates, and this clearly contributed to fuelling private consumption. The development in the real estate market is very important in this respect, because the value of owner-occupied houses constitutes by far the most dominant wealth item for most households. Moreover, in both of these periods substantial increases – in the order of 20–30 per cent – in the real price of houses have been observed. This has caused large wealth increases which in turn have boosted private consumption. Capital market liberalizations have clearly contributed to strengthen this channel, and empirical evidence shows now a closer relationship between changes in consumption and changes in wealth than previously (see, for example, Christensen, 1987, and Callesen, 1997). As is well known, the Maastricht Treaty implies three nominal convergence criteria and two fiscal convergence criteria. A broad political spectrum in Denmark remains committed to the underlying principles of nominal convergence and fiscal discipline. Based on 1997 figures Denmark fulfils all the economic criteria for entering EMU. Also, Denmark is at the forefront of implementing the directives behind the EU’s Internal Market programme. Despite these tendencies to play the role as Europe’s ‘top boy’, Denmark has stated that it will not participate in the third stage of EMU in Europe. Indeed, Denmark has already exercised its right to opt out through the so-called Edinburgh Agreement in the European Council in December 1992. Denmark may reverse that position, but this would be subject to a positive outcome of a referendum. Currently, it seems difficult to predict the chances of full Danish participation in EMU. The official Danish position is clearly to seek the closest possible arrangement with EMU participants. Since it now turns out that the euro will become a reality from the beginning of 1999, a likely approach to be followed by Denmark is one of a unilateral peg vis-à-vis the euro. This would be an arrangement similar to that pursued by Austria of pegging to the D-mark for
Problems, Policies and Prospects
17
about two decades before Austria joined the EU and the EMS (see Thygesen, 1996, and his discussion in Chapter 2 of this volume of the relationship between insiders and outsiders in EMU). Indeed, such an arrangement might provide an adequate defence for the krone, as long as macroeconomic policies remain closely aligned with those of the EMU countries.
1.5 SAVING, INVESTMENT, THE CURRENT ACCOUNT AND FOREIGN DEBT The current account of the balance of payments is the difference between national saving and investment. Any gap between national saving and investment is thus filled by foreign saving – that is, by foreign lending/financing of domestic private or public investments. Hence, the current account leads by definition to a change in a country’s net foreign asset position by a figure that is equal to the current account itself. On top of that, the net foreign asset position may change due to re(de)valuations or asset price changes in general, which alter the value of assets and liabilities denominated in foreign currency. Below, we outline both the trend and the cyclical movement of the current account. The Danish current account was in deficit every single year in the period 1963–90. The persistent deficits have, of course, led to a gradual build up of the country’s foreign debt position (see Figure 1.9).
Figure 1.9
Foreign debt and current account, 1960–96
45%
25%
15%
-15% –15%
Foreign debt/GDP
Source: Statistics Denmark.
Current account/GDP
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
–5% -5%
1962
5% 1960
Per cent of GDP
35%
18
Andersen, Jensen and Risager
The magnitude of the debt burden measured in terms of annual GDP culminated in 1988, where the debt–GDP ratio was around 40 per cent. Owing to the persistent current account surpluses since 1990 and to the growth in GDP, the debt–GDP ratio has now fallen to about 20 per cent – that is, there has been a significant reduction of the country’s foreign indebtedness. There are of course many reasons underlying the persistent historical tendency to run deficits, and the turning around of this trend which took place in 1990 (see Nielsen and Søndergaard, 1991). Below, we concentrate mainly on the most important explanations of the recent surplus trend. To this end it is useful to consider the private and public saving– investment diagrams, which show two important things. First, there was a considerable increase in private sector saving after 1986. Secondly, the investment share was low, in particular in the private sector until the mid1990s. Both tendencies contribute to the turning around of the current account balance. The private sector is now a net creditor vis-à-vis the rest of the world. In an accounting sense, the entire foreign debt is therefore a liability of the public sector, and mainly the government. The two most important explanations of the considerable increase in private saving are the so-called ‘potato diet’ (credit market tightenings) introduced in late 1986 and the tax reform in 1987. The latter led to a drastic rise in the after-tax interest rate. As the nominal interest rate was around 10 per cent in 1987, the after-tax interest rate increased from 2.8 per cent to 5 per cent for many private citizens. Because the private sector at that time was a net debtor, not only the substitution effect (it became more costly to borrow) but also the income effect (the debt service burden increased) worked in the direction of an increase in saving. The rise in private saving has more than offset the fall in public saving. Hence, the national saving rate has gone up. Since 1987, the investment share has been low, and it is only very recently that the investment share has reached previous heights. New national account figures have led to a considerable upward revision of investment figures, partly because IT expenditures now are classified as investments and not as intermediate inputs. In 1996, the new estimate for the investment share is 0.21. Likewise, savings figures have also been revised in an upward direction. Due to the appearance of the new national account figures for the period 1988–96,4 earlier worries about too low thrift and business capital investment have faded somewhat. The movement of the current account is closely related to the business cycle. The general picture is that high growth is accompanied by a weaker current account, and vice versa. The last three business cycles clearly illustrate this phenomenon. The boom from 1983–7, driven by high growth in consumption and investment, caused a big deficit in 1986 of nearly 6 per cent of
Problems, Policies and Prospects
19
Figure 1.10 Gross investment and saving, 1973–96 (a) Private
Per cent of GDP
25% 20% 15% 10% 5%
Private gross investment/GDP
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
0%
Private gross saving/GDP
(b) Government 12% 10% 6% 4% 2% 1993
1994
1995
1996
1993
1994
1995
1996
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
-4% –4
1974
0% –2 -2%
1973
Per cent of GDP
8%
-6% –6 -8% –8 Government gross investment/GDP
Government gross saving/GDP
(c) Total 30%
Per cent of GDP
25% 20% 15% 10% 5%
Total gross investment/GDP
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1978
1977
1976
1975
1974
1973
0%
Total gross saving/GDP
Source: Statistics Denmark.
GDP. This motivated a tightening of economic policy in general and a change in the tax treatment of interest payments in particular, and this marked the beginning of a long period from of low growth 1987 to 1993. As a consequence the external account improved. Following the economic recovery in
20
Andersen, Jensen and Risager
1993, the current account has gradually become weaker as anticipated, but the new tendency is that the current account has remained in a surplus. However, as the current account plays a key role for Danish macroeconomic policy making, a considerable weakening is also the signal that may lead the present government to tighten policy. The role of the current account as a discipline device on macroeconomic policy making has been criticized in recent years (see Corden, 1994; Lund, 1989; and Nielsen, 1988). The point of departure of this criticism is the intertemporal approach. According to this line of thinking households and firms make their saving and investment decisions such that they fulfil their intertemporal budget constraints, which is basically saying that over their whole life-time agents do not spend more than they earn. When that is the case they do not pass burdens onto future generations – that is, there is no risk of free-riding on future generations. Hence, households consumption– saving decisions are what they should be and firms’ investment decisions are also optimal. Now since the current account is essentially the difference between saving and investment, the current account is what the current account should be. Thus, if households decide to have very high consumption for a decade, say, and hence supply very little savings to finance firms’ investment projects, then that is perfectly acceptable because firms can then get the finance via international sources. According to this so-called ‘new view’, the associated current account deficit should be of no concern for policy makers. The only snag here is the public sector. Because it is not clear what the public sector is optimizing, the public sector may have a tendency to run too large deficits, reflecting the fact that politicians have an incentive to give presents rather than taxing individuals in order to be popular and get reelected. Hence, according to the ‘new view’, the current account is only a political problem if the public sector is imbalanced. In order briefly to discuss the ‘new view’, it is important first to note that it is certainly true that a deficit may be perfectly sensible insofar as it may simply reflect high domestic profitability (above the foreign level) and investment. In such a situation it will of course be foolish to run surpluses because it is more profitable to channel domestic saving into domestic than into foreign investments. The underlying view that private saving–investment decisions should not be of any concern is, however, problematic for many reasons; only a few will be discussed here. Although frequently assumed in textbooks, international capital markets are by no means ‘perfect’. Country risk is a well known and important practical phenomenon. Thus, if a country runs large deficits and accumulates high foreign debt, it becomes more costly and difficult for otherwise sound investment projects to get finance in the international markets. Hence, there is an externality (negative) from the country’s outstanding debt to individual
Problems, Policies and Prospects
21
loan conditions, and the aggregate debt is thus a figure that policy makers should be concerned with, as they should be concerned with other externalities in the economy. Many examples in history also show that persistent current account deficits tend to make countries more vulnerable to speculative currency attacks, in which case foreign investors pull their money out of the country, which in turn may lead to turmoil and impede a sensible development of the economy. This may also reflect capital market imperfections, but it may also reflect the fact that markets rationally discount that high debt is associated with a positive probability that countries will renege on their payback promises,5 in which case foreign investors may prefer not to park their capital in that particular country or region of the world. Again, policy makers should care about the current account to avoid too large deficits. Besides capital market imperfections, deviations from rationality may also justify some current account targeting. Thus, if agents over several years behave as if it is cheap to borrow just because nominal interest rates are low, whereas real interest rates are actually high, governments and central banks with a highly professional staff may also try to correct such false beliefs, and in practice this may result in, for example, interventions against private consumption because announcements in themselves may have little credibility and hence impact on the private sector. The ‘new view’ has had little effect on policy making in Denmark, but the school’s emphasis on finding the reasons underlying a(ny) particular current account figure is certainly valuable.
1.6
PUBLIC EXPENDITURES AND TAXATION
The government has been increasingly involved in the Danish economy over the last 35 years (see Figure 1.11). The ratio of general government expenditures to GDP has more than doubled, from 25 per cent in 1960 to 62 per cent in 1997. It is mainly within the ‘core’ areas of the welfare state that outlays have increased (see also Andersen and Schmidt, Chapter 9 in this volume). The most expanding area has thus been transfer payments, which have increased fourfold. Government consumption, including health and education costs, grew rapidly in the 1960s, but has grown more or less in line with GDP since the mid-1970s. Government investment, on the other hand, has been in decline as a share of GDP. Reflecting the steady increase in public debt since the early 1970s, interest expenditures have also risen. The rise in transfer payments is to a large extent attributable to the rise in (the structural rate of) unemployment. Not only has the registered rate of unemployment increased substantially, but there has also been a rise in the number of working-age persons depending on some sort of income-
Andersen, Jensen and Risager
22 Figure 1.11
General government expenditure in per cent of GDP, 1960–96
70% 60% 50% 40% 30% 20% 10%
Interest Payments Interest payments Total expenditure Total Expenditures
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
0%
Transfer Payments Transfer payments Government consumption Government Consumption
Source: ADAM databank, Statistical Yearbook (1962, 1968, 1972), Ministry of Finance.
compensating public transfer payment other than unemployment benefits. In particular, two tendencies can be identified: the workforce is retiring earlier, and a larger number of (mainly low-skilled) workers have been expelled from the labour market. These developments have clearly had budgetary effects: the costs of benefits for the unemployed and retirees have increased and tax revenues have fallen. Although a rising share of government spending to GDP is a phenomenon that has been observed in most industrial countries (see Masson and Mussa, 1995), Denmark has, like other Scandinavian countries, undoubtedly carried the welfare state further than most other European countries, both in terms of coverage and generosity (see Hagen et al., 1998). For example, safety nets have not only been provided for the less fortunate in society, but as a result of the so-called ‘universalist’ approach to the welfare state, relatively generous social programmes have been extended to the general population. The generosity of the Danish welfare programmes may also be reflected in the way many transfer payments have been indexed. For example, in the area of public pensions to the elderly, evidence shows that the purchasing power of public pensions increased by more than real wages of blue-collar workers over the period 1970–92 (see Socialkommissionen, 1993). A certain ‘sense of entitlement’ may also have become particularly widespread in Denmark and other Scandinavian welfare states. An example is
Problems, Policies and Prospects
23
early retirement benefits which were introduced in 1979 as a temporary measure to improve job opportunities for young workers by stimulating elderly workers to leave the workforce. However, most people now feel themselves entitled to receive public retirement benefits at the age of 60, regardless of their health status (or other conditions). It has, therefore, become difficult to scale back these benefits. The extent of public sector involvement in the Danish economy may also be assessed from the revenue side of the (general) government budget (see Figure 1.12). While the total revenue share is above 60 per cent, the revenue from taxes constitutes about 50 per cent of GDP. Against an EU (average) and US comparison, these numbers may appear very high (see Table 1.3). Indeed, it is remarkable that the total revenue–GDP ratio in Denmark is almost twice as high as the corresponding US figure, with the EU average being somewhere in between. However, when comparing the Danish tax burden with that of other European countries, it should be kept in mind that most EU countries, unlike Denmark, pay transfers on a net-of-tax basis. Another difference has to do with the way households are supported: in some countries this takes the form of deductions from the tax base, whereas in Denmark this is done through direct transfer payments. These factors have been found to account for almost five percentage points of the difference in the gross tax rate between Denmark and Germany (see Ministry of Finance, 1994; Lassen and Nielsen, 1996).
Figure 1.12
General government revenues in per cent of GDP, 1960–96
70% 60% 50% 40% 30% 20% 10%
Total revenues Revenues Total Indirect Indirect taxes Taxes
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
0%
Personal Personal Income income Taxes taxes Other Revenues revenues Other
Source: ADAM databank, Statistical Yearbook (1962, 1968, 1972), Ministry of Finance.
24
Andersen, Jensen and Risager
From a public finance perspective it is not only the level but also the structure of taxation which is of interest. Table 1.3 compares the relative importance of the major tax sources in Denmark, the EU (average) and the United States. There are several notable differences between Denmark and the rest of the world. First, the earmarked contributions (payroll taxes to social security) of employers and employees count relatively more in the EU and, especially, in the United States. Indeed, while such payroll taxes only constitute about 4 per cent of total taxes in Denmark, they constitute about a fifth in the EU as a whole and even higher in the United States. The tendency in Denmark is, however, to give a higher priority to such taxes in the future. Second, personal income taxes make up a significantly larger share of the tax revenue in Denmark than in Europe in general and in the United States. More precisely, while the revenue from personal income taxes constitutes more than a half of the total tax revenue in Denmark, the corresponding figures for the EU and the United States are only about a quarter and a third, respectively. Third, corporate taxes are used on a relatively smaller scale in Denmark than in the rest of the OECD, notably in the United States. Fourth, the revenue collected from taxes on goods and services has in recent years converged on the European average, now amounting to about a third of the total revenue. In the United States, however, indirect taxes count much less. VAT is by far the most important indirect tax in Denmark, accounting for more than 50 per cent of the total indirect tax revenue. Fifth, since much of private sector interest income benefits from special tax concessions, while interest payments are fully deductible from taxable income, capital income tax in Denmark yields a negative net revenue.6 This fact partly explains why the residual category ‘other taxes’ is substantially lower in Denmark than in other parts of the world. Although the share of government revenues has risen strongly in this period, increases have not been sufficient to balance government budgets. As a result of persistent deficits, primarily since the mid-1970s, there has been a substantial increase in the (gross) government debt–GDP ratio, from 5 per cent in 1970 to almost 80 per cent in 1996 (see Figure 1.13). Apart from the falling debt ratios in the mid-1980s, and again more recently under the current upturn, the ratio has been steadily increasing. Although the government budget thus displays a cyclical pattern, an important asymmetry can be identified: the surplus in ‘good’ years (if a surplus) is significantly smaller than the deficit in ‘bad’ years. For example, while the Danish economy is currently experiencing a strong boom, the government only ran a small budget surplus in absolute amount in 1997. The major contribution to the falling debt–GDP ratio currently observed is thus the high GDP growth rates (‘the denominator’) rather than big surpluses on the government budget (‘the numerator’). However, according to official
Problems, Policies and Prospects Table 1.3
25
Structure of the tax system in Denmark, EU and USA, 1994 Denmark
EU
USA
(% of total tax receipts) Social security contributions Employees’ Employers’ Personal income taxes Corporate taxes Taxes on goods and services Other taxes Total tax receipts (% of GDP)
2.5 0.7 53.8 3.7 32.0 7.3 51.6
10.2 8.1 26.5 6.4 31.4 9.2 42.5
10.8 13.3 35.7 8.9 17.9 13.3 27.6
Source: OECD (1998).
projections the government is expected to run more substantial budget surpluses, thereby supporting the official debt target of the government which is a debt–GDP ratio of 40 per cent by year 2005.7 It should also be noted that since the late 1970s inflation has increasingly been anticipated, so inflation has become a less powerful device for eroding the real value of government debt. The more recent phenomenon of low (expected and actual) inflation, which has led to low nominal interest rates, has clearly diminished the burden of existing government debt in nominal terms. While the public sector has grown larger in the Scandinavian countries than elsewhere in the OECD, this is not the case as regards the size of public debt. When looking for reasons behind the substantial growth in public debt–GDP rates after 1973, Denmark does not, therefore, constitute a ‘special case’. New theories based on various political-economy models of government behaviour have offered some explanations why public debt has increased so much in many countries. For example, it has been found that in countries with a frequent turnover of government power, each government is likely to overspend when in power, thereby leaving a higher public debt to its successor (see Alesina and Tabellini, 1989). In particular, as demonstrated by Roubini and Sachs (1989), coalition governments tend to have the largest deficits, thus suggesting that if the government is formed by parties of different colour it is difficult to agree on painful measures to keep the budget balanced. Since Denmark for most of the period since 1973 has had minority governments made up of more than one party, the record of public debt in Denmark is not inconsistent with the explanation of the deficit bias in fiscal policy as offered here.
Andersen, Jensen and Risager
26 Figure 1.13
Gross government debt in per cent of GDP, 1960–96
90% 80% 70% 60% 50% 40% 30% 20% 10%
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
1970
1968
1966
1964
1962
1960
0%
Source: ADAM databank, Statistical Yearbook (1962, 1968, 1972), Ministry of Finance.
Since the convergence criteria for EMU qualification became guidelines for the conduct of macro policy in Europe, Denmark has like other EU countries – despite its reservations for EMU participation – steered its fiscal policy with a view to the public debt–GDP targets implied by the debt and deficit criteria. It would, however, take us far beyond the scope of this chapter to seriously discuss the role of the Maastricht Treaty as an appropriate device for enhancing fiscal responsibility. Two tax reforms have been enacted within the last decade or so.8 The first reform was introduced in 1987 and had as its prime objective stimulating saving, with the ultimate goal of bringing the ongoing foreign debt accumulation to a halt. In particular, although the reform did not attack the right to deduct all interest expenses from taxable income, the tax value of interest expenses was cut, corresponding to the lowest marginal tax rate of the income tax system (see section 1.5). Tax reform was also a key ingredient in the policy package of the new centre–left government taking office in January 1993. That reform largely followed the recommendations put forward in the report from the Tax Reform Committee (1992). It was designed in the same spirit as most other tax reforms introduced in the OECD area in recent years, namely cuts in marginal income tax rates financed through various measures to broaden the tax base. Also in line with most other tax reforms in the OECD area, the Danish reform has led to a drop in the degree of tax progressivity. Detailed examinations of the 1993 tax reform include Lange, Pedersen and Sørensen (Chapter 5 in this volume) and Knudsen et al. (1997).
Problems, Policies and Prospects
27
The reform has been phased in over a period of five years. By 1998, the reform promised a maximum marginal tax rate on labour income of 58 per cent, to be compared to the pre-reform maximum marginal tax rate of 68 per cent. However, due to the introduction of so-called ‘labour market contributions’ – which are virtually payroll taxes, amounting to 8 per cent of gross labour income – the fall in the effective marginal tax rates is relatively modest. Specifically, since the personal income tax is levied on labour income net of payroll taxes, the effective marginal tax rate for high levels of personal income (above 234,900 kroner) is 61.4 per cent. The revenues from the payroll taxes are earmarked for various labour market initiatives, including unemployment insurance, active labour market programmes and sickness and maternity benefits. It has been estimated that the reform will lead to a fall in the ratio of income tax revenue–GDP from 26 per cent in 1993 to 21 per cent in 1998. From an international perspective a novelty of the reform is the introduction of environmental fees. The so-called ‘green taxes’ are levied on households (most important is the tax on water consumption), but the emission of carbon dioxides from the business sector is also subject to taxation. Since Denmark was among the first countries in the world to impose green taxes on the business sector, there has been a lively debate on what impact it would have on Danish companies exposed to international competition, particularly the most energy-intensive part of industry (see Jensen, 1998). Overall, the 1993 reform is regarded as a revenue neutral broadening of the tax base, and the introduction of payroll taxes has been motivated by the need to harmonize Denmark’s tax structure with the rest of the EU. The reformed tax structure is also relevant to the domestic debate about the need for welfare reforms. Indeed, the tax system may have crucial effects on incentives to work and hire, savings behaviour, the ‘black’ economy, do-ityourself activities and so on. Finally, it should be noted that the introduction of payroll taxes and other earmarked budget items represent a departure from the principle of centrally pooled public finances. Current debates on future tax reforms centre around the question of whether wage taxes can be replaced by green taxes to an even larger extent than has already taken place, thereby reaping a so-called ‘double dividend’ in the sense of both improving labour market structures and helping the environment. Also, the issue of making further cuts in the taxation of capital income (including corporate taxes) and (imputed) returns to owner-occupied housing has been put on the agenda. If implemented, this would be accompanied by initiatives to limit the deductability of interest expenses. While such initiatives can be seen as natural extensions of the tax reforms of the 1980s and the 1990s, a more radical approach to tax reform has also been heard, namely the so-called ‘flat tax’ (FT) (Hall and Rabushka, 1985;
28
Andersen, Jensen and Risager
see Nielsen, Frederiksen, and Lassen, Chapter 4 in this volume). Basically, as suggested by its name, the FT proposal involves giving up the progression built into the tax system currently prevailing in Denmark, and placing instead a single fixed tax on wages and cash flows in domestic firms. Nielsen et al. have calculated that a flat tax rate of 44 per cent would suffice to generate the same revenue as the current tax system generated in Denmark in 1992, even under an ‘no change in behaviour’ assumption. If allowing for efficiency gains likely to accompany a reform of that kind, such as increases in labour supply and saving, the FT rate might be even lower. Compared to the current tax system, the FT system would thus have the advantages of (a) reducing the administrative burden, (b) generating a given revenue at lower marginal tax rates and (c) removing some asymmetries. On the negative side, however, a switch towards a FT system would involve some adverse distributive effects which eventually might render it undesirable. The ‘traditional’ trade-off between efficiency and distribution thus also seems to be present here. Nevertheless, if supplemented by initiatives to preserve the redistributive properties of the current tax and transfer system, a future tax reform along the lines suggested by the FT system might be worth considering.
1.7
THE WELFARE STATE AND MARGINALIZATION
The redistributive capacity of the Danish tax and transfer system is quite significant. Although the distribution of incomes between ‘rich’ and ‘poor’ is not particularly equal, even before taxes and transfers, the distribution of disposable incomes has been found to be much more equal (see Førster, 1994). Unlike what has been found in several other OECD countries, the redistributive effects of taxes and transfers seem to be enhanced by the provision of public services, such as education and subsidized child-care (see OECD, 1996). While this would normally be regarded as attractive properties of the tax and transfer system, the question is whether the extensive welfare system in Denmark has a negative impact on the microeconomic balance of the economy, such as providing disincentives to work. This question appears to be an important one as there has, over the past three decades or so, been a rather sharp increase in the number of working-age persons being dependent on income transfers from the public sector, now amounting to about a quarter of the working-age population. Table 1.4 summarizes some facts about the total number of working-age (15–66) persons, persons in employment, recipients of public transfers (divided into various categories) and, as a residual, the number of ‘nonrecipients’ within that age-group.
Problems, Policies and Prospects Table 1.4
29
Working-age population, employment and recipients of public transfer payments, 1960–97 1960
1970
1975
1980
1990
1995
1997
(000 persons aged 15–66) A Working-age population B Employment C Recipients of public transfers Unemployment benefits Cash benefits – of which unempl. Rehabilitation benefits Early retirement benefits Transitory benefits Leave benefits – of which education Early retirement pension Sickness benefits Maternity benefits D Non-recipients (A–B–C)
2.968 2.137 184 31 39 8
3.232 3.329 2.341 2.338 307 496 24 108 41 77 5 23
3.427 2.442 634 159 107 25 9 54
3.596 2.564 784 217 116 55 22 104
103 205 259 236 10 29 37 54 3 9 16 16 647 584 495 351 (% of persons aged 15–66) E Employment ratio (B/A*100) 72.0 72.4 70.2 71.3 F Recipient ratio (C/A*100) 6.2 9.5 14.9 18.6 G Non-recipient ratio (D/A*100) 21.8 18.1 14.9 10.2
251 43 32 248 71.3 21.8 6.9
3.682 3.721 2.521 2.603 961 908 238 183 138 129 51 37 16 16 120 136 23 43 77 46 28 22 271 272 42 49 36 35 200 210 68.5 26.1 5.4
70.0 24.4 5.6
Source: Ministry of Finance (1997); own calculations.
Between 1960 and 1997, the population of working-age (15–66) has increased by 750,000 persons. While total employment has increased by 460,000 persons, largely concentrated within the public sector, unemployment has over the same period risen by 150,000 persons. In addition to the rise in unemployment, there has been a steep growth in the number of recipients of other public transfer payments – that is, benefits not explicitly related to unemployment; for more details see Søndergaard (Chapter 10 in this volume). This phenomenon is strongly associated with developments in the public pension system. In particular, there has been a sharp rise in the number of recipients of early retirement pension, from 103,000 in 1960 to 272,000 in 1997. This scheme includes persons who for health or social reasons are unable to participate in the labour market. Likewise, the number of recipients of early retirement benefits has increased greatly since the inception of this scheme in 1979. These benefits are available to persons aged 60 and
30
Andersen, Jensen and Risager
above provided they have been members of an unemployment insurance fund for at least 20 years. As a more recent phenomenon, arrangements have been introduced which enable members of the labour force to take paid leave up to one year, either for child-minding or for participation in education and training programmes. On their introduction the leave schemes proved rather popular, as reflected by the fact that about 80,000 persons (more than 2 per cent of the labour force) participated by the end of 1995, but as a result of various tightenings thereafter the number of participants in 1997 had fallen to about 40,000 persons. See Andersen, Appeldorn and Weise (1996) for a detailed study of the leave schemes. A large number of persons also receive cash benefits, available to uninsured unemployed persons from households with an income and wealth below a certain threshold. In principle, cash benefits are granted on the condition that the recipients are available to the labour market. However, the numbers show that less than a third of those receiving cash benefits are actually registered as unemployed. The rest would normally be regarded as permanently excluded from the labour market. The upshot of these developments is that the so-called ‘recipient ratio’ (the number of recipients of public transfers as a percentage of the total number of working-age persons) has risen dramatically over the years, from 6.2 per cent in 1960 to 27.7 per cent in 1995. The good news, however, is that the ratio has actually been falling since 1995, albeit only on a small scale. Against this, it would be relevant to know how many of those recipients actually have the ability to work. Indeed, some critics have interpreted the above numbers as evidence that the ‘true’ rate of unemployment is much higher than the registered rate. For example, Ølgaard (1995) has argued that about 18 per cent of all working-age persons are non-employed but should in principle have the ability to work. While many recipients of public transfers have their status because of disabilities of various kinds, and hence are unable to participate actively in the labour market, it is undoubtedly true that others could, if offered, take a job. From a social viewpoint, it would count as an enormous success if marginalized people could change their status from a passive life on welfare money to a more active and independent life. From a broader economic perspective such a process would not only help in improving government finances, but would also ease the current pressure on the labour market in Denmark where some sectors now appear to be operating at the margin of full capacity utlilization. Less pessimistic conclusions have also been drawn from this evidence. For example, as Table 1.4 shows, it has been noticed that the employment rate (the share of of working-age people in employment) has proved very stable since the early 1960s. Hence, the rise in unemployment is closely related to the rise in the labour force. In particular, the phenomenon must be seen in the light of the significant increase in the labour market participation rate among women.
Problems, Policies and Prospects
31
While this is true, it could be said that a well functioning labour market should be capable of absorbing such an influx to the labour market. It has also been argued that rather than associating the enormous rise in the number of recipients of public transfers with a dramatic increase in de facto unemployment, it simply reflects a change in political priorities in favour of granting public benefits to persons who previously would have been provided for by family or other ‘local’ safety nets, if provided at all (see Callesen, 1995). Hence it is a result of the expansion of the welfare state, as indicated by the dramatic fall in the so-called ‘non-recipient’ ratio in Table 1.4. In other words, a person of working-age who is not working and who is not a recipient of welfare money is today almost a non-existing phenomenon. It is difficult to judge the impact of this development on social welfare. From a social viewpoint, the perhaps most important problem facing today’s society is the number of truly marginalized persons or social outcasts. These individuals are typically concentrated in certain inner-city areas and many of them are second-generation immigrants. The real challenge to the welfare system in the coming years is to bring this marginalization process to a halt. While basically a matter of definition, the question about the ‘true’ number of marginalized individuals has become subject to controversy among different policy-creating communities in Denmark. For example, the so-called Welfare Commission (1995) found that 250,000 persons were truly marginalized in 1995, a number also reported by the Danish National Institute of Social Research. Using an alternative definition, the Ministry of Finance (1995, 1997) reports a much lower number for 1994, namely 104,000, a number projected to fall to 40,000 by 1999. Since it is the rise in public transfer payments which explain why public expenditures have risen over the years, the following question often arises: ‘can we really afford to pay these transfers?’ It is certainly true that the rise in expenditures allocated to public transfers has not only led to higher (marginal) tax rates, but it may also have crowded out expenditures which might otherwise have been allocated to public consumption, including the provision of various services and investments in infrastructure, etc. While high marginal tax rates are well known to have distortionary effects, a continuous fall in the share of expenditures to public service production and infrastructure relative to public transfer payments may in the longer term threaten the legitimacy of the welfare state (see Ingerslev and Ploug, 1996). The point is that many so-called middle-class persons/families may find it increasingly difficult to see their tax money being spent on transfer payments – from which they themselves may derive only little utility – instead of being spent on public services – from which they are likely to derive a lot more utility. Another criticism frequently raised is this: ‘it doesn’t really pay to work any more!’ In general, the incentives offered by the welfare system to
32
Andersen, Jensen and Risager
participate in the labour market are rather mixed. On the one hand, the combination of generous unemployment benefits and relatively long benefit periods should induce people to enter the workforce. Moreover, government subsidization of daycare for children adds to this incentive. On the other hand, the early retirement benefits would typically operate in the opposite direction. Indeed, this scheme has contributed strongly to reducing the effective retirement age from nearly 66 in 1979 (the year when the early retirement benefits were introduced) to about 61 in 1997. A recent study by Pedersen, Smith and Stephensen (1998 – and in Chapter 6 of this volume) shows that 20 per cent of the insured people in work in Denmark have no or only weak economic incentives to remain in employment. Similarly, unemployed persons receiving the maximum UI benefits have only little, if any, economic incentive to accept low-paid jobs, since the benefits typically exceed the minimum wage in the manual labour market (see also section 1.3).
1.8 POPULATION AGEING AND INTERGENERATIONAL REDISTRIBUTION In addition to the pressure on public finances induced by the large number of working-age recipients of public transfers, the projected changes in the age structure of the population are likely to challenge the fiscal system. Indeed, the trends of modest economic growth, high real interest rates, ageing populations and high levels of (existing) public debt have led many economists to question the long term sustainability of fiscal policy as currently practised. A closely related issue is whether, and to what extent, financial burdens are passed onto future generations if current fiscal policies continue without adjustments. Since changing demographics are the major driving force behind the threats to public finances in the future, we start with a brief look at the population projections. Following standard age demarcations we may define the working age population as those between the ages of 20 and 64, the elderly population as those aged 65 and above, and the very old as those aged 80 and above. It is most interesting to look at the development in the ratio of elderly to working-age persons, N65+/N20,64, and the ratio of very old to the general population of elderly, N80+/N65+. While the development in the former ratio provides information about the general ageing process, the latter tells which of either the young or the old group of elderly is growing most rapidly. If both ratios are rising at the same time, we have the phenomenon often referred to as ‘double ageing’. Both ratios are displayed in Figure 1.14, for a projection period ranging from 1997 to 2030.
Problems, Policies and Prospects Figure 1.14
33
Projected changes in demographic structure, 1995–2040
0.40 0.35 0.30 0.25 0.20 0.15 0.10
N65+/N20,64
2039
2037
2035
2033
2031
2029
2027
2025
2023
2021
2019
2017
2015
2013
2011
2009
2007
2005
2003
2001
1999
1997
0.00
1995
0.05
N80+/N65+
Source: Statistics Denmark, own calculations.
The first remarkable observation is the ‘j-shaped’ nature of the N65+/N20,64 trajectory: over the next 10 years or so the overall demographic dependency burden is falling, whereupon it increases rather dramatically. By international comparison, the prospect of a demographic ‘breathing space’ is rather unique. Second, there seems to be a tendency for the two curves to move in different directions, at least over the next 25 years: during the first 10–15 years, the share of the very old is rising, and when general ageing sets in, the share of the very old is falling. From about 2020, however, there will be some ‘double ageing’ in Denmark. From an economic perspective, the apparent absence of double ageing until 2020 may be quite important. Evidence shows, see Socialkommissionen (1993), that there is a strong positive relationship between age of the elderly and public outlays on old-age services. The point is, of course, that very old persons’ need for medical care, home care, etc. is substantially higher than young elderly’s need for those services. To get a first impression of the budgetary significance of population ageing, we may treat government outlays to the elderly as if they constitute a separate government budget. Under a pay-as-you-go financing scheme, the per-period budget constraint says that current government spending on the elderly must be equal to the taxes levied on working-age persons. Recent evidence shows that, given realistic indexation schemes of per capita spending on services and income transfers to the elderly, the taxes needed to cover all old-age expenditures would rise from about 11 per cent of GDP in 1993 to about 15 per cent of GDP in 2030 (see Ministry of Finance, 1995; Jensen and Nielsen, 1995). A more detailed analysis of the relevant issues is offered by Jensen, Hansen and Junge (Chapter 7 in this volume).
34
Andersen, Jensen and Risager
The overall effects of changing demographics may also be studied using the device of generational accounting. Recent work has studied the requirements put on fiscal policy in order to ensure that current and future generations are burdened equally (see Jensen and Raffelhüschen, 1997, 1998). Hardly surprising, the main finding is that in order to equalize the fiscal burdens on newborns and future generations, current generations would have to face tighter fiscal policies. More concretely, the required fiscal tightening is quantifed to a rise of 3.4 per cent in the revenue from all taxes; or a cut of 4.7 per cent in all transfer payments, or a cut in the level of government consumption of almost 10 per cent (see Jensen and Raffelhüschen, 1998). While this might suggest that fiscal policy in Denmark is strongly intergenerationally imbalanced, it has also been demonstrated that if unemployment rates could be brought further down, this could have very positive effects on the generational stance of fiscal policy. A condition for this is of course that the additional taxes paid by current generations are spent in a sensible way – that is, to bring down existing public debt so that future generations will be burdened less. The tax reform act of 1993 also seems to help redressing the generational imbalance, see Lange, Pedersen and Sørensen (Chapter 5 in this volume). The question is what else can be done to temper the effects of population ageing. In view of the breathing space, Denmark seems to have a unique opportunity for improving the fiscal positions over the next decade – that is, to bring down public debt relatively fast. Also in the area of public pensions, some policy options are open. However, since for at least half of today’s pensioners, and for at least a third of pensioners 30 years from now, the public pension benefit will represent the only (or certainly the most important) means of old-age security, it is difficult to think of a significant fall in the replacement rate being possible. Instead, an obvious measure would be to raise the retirement age. In particular, this would involve a reform of the so-called early retirement benefit scheme. At a more general level, it seems that the appropriate policy response to the ageing effect would be to increase the workforce. In that respect, the apparent constancy of the employment rate over the last three decades may have serious implications for the extent to which the ageing problem can be overcome by expanding the workforce. Indeed, if it is a ‘stylized fact’ that the employment rate is virtually decoupled from the growth in the workforce, it would be pointless to induce people to postpone their retirement from the labour market. From the point of view of economic theory, however, it would be very hard to believe in such a peculiar relationship in the medium-to-long run.
Problems, Policies and Prospects 1.9
35
CONCLUDING REMARKS
Over the last 25 years the Danish economy has had difficulties in growing as fast as other EU countries and the United States. While the average growth difference is small, it signals that if this trend persists into the next century, Denmark will not be able to maintain its high position in the world income hierarchy. Moreover, during these years, the number of individuals living on transfer incomes have increased dramatically. Although we interpret both tendencies as signals of structural weaknesses, we are also aware that these developments may reflect that other goals in economic policy have been pursued, such as protecting the environment and/or achieving certain redistributive objectives. The Danish experience shows that economic policy has significant effects on the economy. Thus the upturns in the beginning of the 1980s and 1990s were both intimately linked to policy reforms. Similarly, the recession starting in 1987 was also to a large extent caused by economic policy. Hence, it does matter what governments do. In particular, while the scope for pursuing macroeconomic stabilization policies in traditional sense may be limited, a lot can be done in the area of structural reform. The current upturn in the Danish economy is of great interest since it has to some extent been underpinned by structural reforms, mainly in the labour market. Although the current business cycle situation is favourable, it is not realistic that growth can be maintained without further policy initiatives. Further labour market reforms are essential both for the direct effect on unemployment and the possibility of giving a larger group of the population a genuine chance of entering the labour market, but also for the indirect effects it has on the public budget. It is not realistic to make substantial improvements in public finances without a reduction in the number of individuals on transfers. In a longer-term perspective this will also be an important step in creating leeway to deal with the problems of an ageing population. As part of the reform policy further tax reforms are needed both to strengthen the incentive to work and to save. It is also important to continue the deregulation process to ensure more competition in product markets and to create a more flexible and efficient public sector. For a society aiming at a high and fairly equally distributed living standard, it is essential in a world of intensified international competition that the labour force has a level and distribution of qualifications which makes it possible to attain this target on market terms. An important challenge for the future may thus be to have an efficient and flexible educational system to ensure that qualifications relevant for the labour market are provided.
Andersen, Jensen and Risager
36 Notes 1. 2.
3.
4.
5. 6.
7.
8.
Comments from Per Callesen, Ole Jørgensen, Jan Olesen and Peter Birch Sørensen are gratefully acknowledged. We also thank Christoffer Kok Sørensen for highly effective research assistance. The importance of economic policy for the behaviour of the economy and for welfare in general is an important motivator for this book. Thus, we believe that a better understanding lays the foundation for better policies that may improve the Danish citizens’ economic well-being. Countries participating in the ERM of the EMS fix a parity relative to the ECU. The exchange rate was initially allowed to fluctuate around this parity by ±2.25 per cent, but after the European exchange rate crisis in 1993 this was widened to ±15 per cent, effectively meaning that the bands are now nonbinding. Because the new National Accounts only cover the short period 1988–96, we are forced to use the old figures to get a historical perspective. However, the old figures are also likely to be more reliable the further we move back in time. IT investments were, for instance, of much less importance 10 years ago. Another difference between the two systems is the use of base year; the old system is based on 1980 which is fine in (say) 1986, but probably not in 1993, and the new system therefore uses 1990 as a base year. Finally, there has also been a sectoral redefinition. Notice that this is the case where there is a default risk usually ignored in the intertemporal deterministic approach, but nobody can afford to ignore this in the real world. As to the principles of taxation, Denmark was the first Nordic country to move away from the principle of global income taxation where a single progressive tax schedule is applied to the sum of the taxpayer’s income from all sources toward a system of so-called dual income taxation where the taxation of capital income is separated from the taxation of other sources of income (see Sørensen, 1994). Discussions of public debt are often confused by the fact that several alternative definitions are offered. The debt ratio mentioned here refers to what is called ‘gross government debt’ which includes the government’s outstanding (domestic and foreign) debt in the form of bonds. This is different from the so-called ‘EMU debt’, used by the EU for intercountry comparisons, and which in 1996 constituted 69.9 per cent of GDP. If some specific Danish institutional features are allowed for, including a state-owned pension fund’s holdings of non-government bonds and the state’s deposits in Danmarks Nationalbank, the ratio actually falls below 60 per cent. Further details on recent tax reforms in Denmark are provided by Sørensen (1990, 1994), Matthiessen (1994) and Lange, Pedersen and Sørensen (Chapter 5 in this volume).
References Alesina, A. and G. Tabellini (1989) ‘A positive theory of fiscal deficits and government debt in a democracy’, NBER Working Paper, 2308. Andersen, D., A. Appeldorn and H. Weise (1996) ‘Orlov – evaluering af orlovsordningerne’, Socialforskningsinstituttet, Rapport, 9611. Andersen, T.M. (1990) ‘Stabilization policies towards internal and external balance in the Nordic countries’, Scandinavian Journal of Economics, 92, pp. 177–202.
Problems, Policies and Prospects
37
Andersen, T.M. (1994) ‘Disinflationary stabilization policy: Denmark in the 1980s’, in J. Åkerholm and A. Giovannini (eds), Exchange Rate Policies in the Nordic Countries (Cambridge: Cambridge University Press for the Centre for Economic Policy Research). Andersen, T.M. and O. Risager (1988) ‘Stabilization policies, credibility, and interest rate determination in a small open economy’, European Economic Review, 32, pp. 669–79. Andersen, T.M. and O. Risager (1988) ‘Wage formation in Denmark’, in Chapter 3 L. Calmfors (ed.), Wage Formation and Macroeconomic Policy in the Nordic Countries (Oxford: Oxford University Press). Bergman, M. and M. Hutchison (1996) ‘The “German view”, fiscal consolidation and consumption booms: empirical evidence from Denmark’, EPRU Working Paper Series, 1996–10. Bingley, P., N. Bjørn and N. Westergård-Nielsen (1994) ‘Determination of wage mobility in Denmark in the 1980s’, Working Paper, Centre for Labour Market and Social Research, Aarhus. Callesen, P. (1995) ‘A note on policies to reduce structural unemployment’, Ministry of Finance, mimeo. Callesen, P. (1997) ‘The policy experience of structural reforms in Denmark’, Ministry of Finance, mimeo. Christensen, A.M. (1987) ‘Indkomst, formue og privatforbrug’, Danmarks Nationalbank, mimeo. Cordon, W.M. (1994) Economic Policy, Exchange Rates, and the Intertemporal System, Chapter 6 (Oxford: Oxford University Press). Dalgaard, E. (1996) ‘Kan velfœrdsstaten opretholdes?’, in E. Dalgaard, O. Ingerslev, N. Ploug and B.R. Andersen, Velfœrdsstatens Fremtid (Copenhagen: Handelshøjskolens Forlag). Det Økonomiske Råd (Danish Economic Council) (1994) Danish Economy (June). Det Økonomiske Råd (Danish Economic Council) (1997) Danish Economy (Autumn). Economist, The (1997) Guide to Economic Indicators. Emerson, M. (1988) ‘Regulation and deregulation of the labour market: policy regimes for the recruitment and dismissal of employees in the industrialized countries’, European Economic Review, 32(4), pp. 775–817. Frederiksen, N.K., P.R. Hansen, H. Jacobsen and P.B. Sørensen (1995) ‘Subsidising consumer services: effects on employment, welfare and the informal economy’, Fiscal Studies, 16, pp. 71–93. Førster, M. (1994) ‘Measurement of low incomes and poverty in a perspective of international comparisons’, OECD Labour Market and Social Policy Occasional Papers, 14. Giavazzi, F. and M. Pagano (1990) ‘Can severe fiscal contractions be expansionary? Tales of two small European countries’, in O. Blanchard and S. Fischer (eds), NBER Macroeconomics Annual, pp. 75–123 (Cambridge, Mass: MIT Press). Hagen, K.P., E. Norrman, P.B. Sørensen and G. Teir (1998) ‘The Nordic welfare states and European economic integration’, in P.B. Sørensen (ed.), Tax Policy in the Nordic Countries (London: Macmillan). Hall, R.E. and A. Rabushka (1985) The Flat Tax (Stanford: Hoover Institution Press). Hansen, C.T., L.H. Pedersen and T. Sløk (1996) ‘Danske resultater om sammenhången mellem marginalskat og løn’, Nationaløkonomisk Tidsskrift, 134, pp. 153–74. Ingerslev, O. and N. Ploug (1996) ‘Velfœrdsstatens Udvikling’, in E. Dalgaard, O. Ingerslev, N. Ploug and B.R. Andersen, Velfœsstatens Fremtid (Copenhagen: Handelshøjskolens Forlag).
38
Andersen, Jensen and Risager
Jensen, A.M. and P. Jensen (1996) ‘The impact of labour market training on the duration of unemployment’, Working Paper, Centre for Labour Market and Social Research, Aarhus. Jensen, J. (1998) ‘Carbon abatement policies with assistance to energy intensive industry’, Ministry of Business and Industry, mimeo. Jensen, P., M. Rosenholm and N. Smith (1994) ‘Unemployment and minimum wages – a micro-econometric analysis’, Working Paper, Centre for Labour Market and Social Research, Aarhus. Jensen, S.H. (1996) ‘Recent developments in the Danish economy: problems, policies and prospects’, EPRU Analyses, 13, Economic Policy Research Unit, Copenhagen. Jensen, S.H. and S.B. Nielsen (1995) ‘Population ageing, public debt and sustainable fiscal policy’, Fiscal Studies, 16, pp. 1–20. Jensen, S.H. and B. Raffelhüschen (1997) ‘Generational and gender-specific aspects of the tax and transfer system in Denmark’, Empirical Economics, 22, pp. 615–36. Jensen, S.H. and B. Raffelhüschen (1998) ‘Public debt, welfare reforms, and intergenerational distribution of tax burdens in Denmark’, in A. Auerbach, L. Kotlikoff and F. Leibfritz (eds), Generational Accounting Around the World (Chicago: Chicago University Press). Jensen, S.H., B. Raffelhüschen, P. Jacobsen and M. Junge (1996) ‘Et generationsregnskab for Danmark’, Nationaløkonomisk Tidsskrift, 134, pp. 39–60. Knudsen, M.B., L.H. Pedersen, T.W. Pedersen, P. Stephensen and P. Trier (1997) ‘Modelling structural reform: a dynamic CGE analysis of the Danish Tax Reform Act of 1993’, paper presented at the OECD conference on The Effects and Policy Implications of Structural Adjustments in Small Open Economies (Amsterdam) (23–24 October). Lassen, D.D. and S.B. Nielsen (1996) ‘Skattebyrder i Danmark og andre EU-lande’, Nationaløkonomisk Tidsskrift, 134, pp. 209–22. Lund, L. (1989) ‘Kritik af betalingsbalancemålsæjtningen’, Ekonomisk Debatt, 2, pp. 116–27. Masson, P. and M. Mussa (1995) ‘Long-term tendencies in budget deficits and debt’, in T. Hoenig, (ed.), Budget Deficits and Debt: Issues and Options, Federal Reserve Bank of Kansas City. Matthiessen, L. (1994) ‘Perspektiver på skattereformen’, Nationaløkonomisk Tidsskrift, 132, pp. 36–58. Mendoza, E., A. Razin and L. Tesar (1994) ‘Effective tax rates in macroeconomics. Cross-country estimates of tax rates on factor incomes and consumption’, Journal of Monetary Economics, 34, pp. 297–323. Ministry of Economic Affairs (various issues), Økonomisk Oversigt. Ministry of Finance (1994, 1995), Finansredegørelse. Ministry of Finance (1995) Pensionssystemet og fremtidens forsørgerbyrde (Copenhagen). Ministry of Social Affairs (1994) Den sociale indsats for sindslidende og socialt udstødte (Copenhagen). Nielsen, L.T. (1988) ‘Udlandsgæjldens årsager og konsekvenser’, Økonomi og Politik, 61(2), pp. 32–40. Nielsen, S.B. and J. Søndergaard (1991) ‘Macroeconomic policy and the external constraint: the Danish experience’, in G. Alogoskoufis, L. Papademos and R. Portes (eds), External Constraints on Macroeconomic Policy: The European Experience (Cambridge: Cambridge University Press). OECD (1996) Economic Surveys, Denmark (Paris: OECD). OECD (1997a) ‘Education at a glance’, OECD Indicators, Centre for Educational Research and Innovation (Paris: OECD).
Problems, Policies and Prospects
39
OECD (1997b) ‘Implementing the OECD jobs strategy: lessons from member countries’ experience’ (Paris: OECD). Ølgaard, A. (1995) ‘Borgerløn ad bagdøren’ (Copenhagen: Børsen). Pedersen, P.J. and N. Smith (1996) ‘Indkomstskatter og arbejdsudbud’, Nationalkonomisk Tidsskrift, 134, pp. 1–23. Risager, O. (1993) ‘Labour substitution in Denmark’, Oxford Bulletin of Economics and Statistics, 55, pp. 123–37. Roubini, N. and J. Sachs (1989) ‘Government spending and budget deficits in the industrial countries’, Economic Policy, 11, pp. 100–32. Socialkommissionen (1993) De œldre – en belysning af œldregenerationens forsørgelse (Copenhagen: DBK Bogdistribution). Sørensen, P.B. (1990) Dansk skattepolitik i 1990’erne (Copenhagen: Jurist- og Økonomforbundets Forlag). Sørensen, P.B. (1994) ‘From the global income tax to the dual income tax: recent tax reforms in the Nordic countries’, International Tax and Public Finance, 1 May), pp. 57–79. Sørensen, P.B. (1997) ‘Public finance solutions to the European unemployment problem?’, Economic Policy, 25 (October), pp. 221–65. Tax Reform Committee (1992) (Copenhagen: Danish Government). Thygesen, N. (1996) ‘The prospects for EMU by 1999 and reflections on arrangements for the outsiders’, Swiss Political Science Review, 2, pp. 172–7. Vastrup, C. (1989) ‘Economic policy and adjustment in Denmark’, in M. Monti (ed.), Fiscal Policy, Economic Adjustment and Financial Markets (Washington, DC: IMF). Welfare Commission (1995) (Copenhagen: Danish Government).
2 EMU and the Outsider Nations1 Niels Thygesen 2.1
INTRODUCTION
It has been predictable since the drafting of the Maastricht Treaty in 1991 that not all EU Member States would join the third stage of Economic and Monetary Union (EMU) at the same time. Some might want to preserve a right not to join – an ‘opt-out’ – others were likely to find themselves unable to meet the economic criteria for entry. The Treaty therefore had to deal with the prospect of so-called ‘variable geometry’ in relation to EMU. Section 2.2 reviews how this was achieved. Since all EU Member Stares are, in principle, obliged to regard their macroeconomic policies as a matter of common concern and, in particular, to observe the principles of the so-called Stability and Growth Pact – though with the important proviso that non-participants in EMU cannot be subjected to the sanctions that may ultimately bite (i.e. deposits and fines) – the most obvious question in relation to the outsiders is how exchange rate cooperation will be organized between them and the euro area. Section 2.3 addresses this issue by analyzing critically the framework agreement for an ERM 2 reached in the course of 1996. The agreement in question seems unlikely to be up to the task of preparing for future cooperative relations in the most reassuring way. Section 2.4 therefore discusses a bolder proposal than ERM2 for countries close to convergence – an ‘associate status’ in EMU. A brief section 2.5 offers some conclusions.
2.2
VARIABLE GEOMETRY
What are the institutional, economic and political consequences of the form of variable geometry foreseen in the Treaty? Does it represent a serious hurdle for EMU, especially for those willing, but temporarily unable? The institutional problems of variable geometry are already addressed in the Treaty which explicitly regulates how the ECB (and to some extent ECOFIN) will have to deal with a variable number of participants in the third stage. The initial group of countries will, of course, have more influence than 40
EMU and the Outsider Nations
41
the latecomers on the institutions since the Treaty stipulates that only the countries that participate in the initial group will decide on the composition of the Executive Board (the President, Vice-President and up to four other members) of the ECB.2 Even if it is decided that not all six members are appointed immediately (this possibility is already foreseen in the Treaty), the members of the Executive Board will initially have an important position in the Council of the ECB because experience with Federal central banks such as the Federal Reserve and the Bundesbank shows that the members of the daily management (Board, Direktorium) tend to play a major role within the larger governing bodies (FOMC, Bundesbankrat). The initial group of countries participating in EMU will also take certain decisions concerning the implementation of the common monetary policy and the instruments to be used by the ECB that will have to be accepted by the countries that join later. These are minor issues, however, compared to the economic and political ones. It is tempting to compare the transitional arrangements for EMU to those that have always been made for weaker member countries (and for new members). There is one crucial difference, however. In the legal field, the transitional periods can be defined and limited exclusively in terms of time. In the monetary field, this is not possible, because what counts are the results – i.e. the fulfilment of the convergence criteria – and they cannot be guaranteed by the passage of time. What are the incentives created by the Maastricht provisions? In general, the gains from EMU increase with the number of participants. The old adage ‘the more the merrier’ thus applies to EMU as well. However, once a core group that includes at least France, Germany and some other countries has started EMU, the marginal gains in strictly economic terms from adding ‘peripheral’ countries are modest, because the latter account only for a small share of trade and output of the Union. This is not necessarily true the other way round: for a ‘peripheral’ country, the gain from participating in EMU might be very large indeed. In general, this implies that the nature of EMU will be affected by the composition of the group of countries that takes the first step. The latecomers will have to adapt to what has already been decided (see Pisani-Ferry, 1996). The asymmetry in interests between the core and the rest could create problems. For example, it has been argued (see, among others, de Grauwe, 1997, and Genberg, 1997) that if the initial group perceives that it has a higher preference for price stability than the rest of the EU, it might be tempted to increase the requirements for subsequent participation by the countries that are thought to be weaker. In this view, it would even be possible that the countries that cannot participate in EMU from the start might be excluded indefinitely. It is difficult to see how this could come about in reality, however, since the convergence criteria are defined in
42
Thygesen
objective terms. For the main exception – i.e. the rule on public debt – it is possible to devise a rule of thumb that would eliminate most arbitrariness from this criterion as well (see Gros and Thygesen, 1998, Chapter 8). Furthermore the prudence created by the interpretation of the convergence criteria for the first group will in practice become binding for subsequent entrants as well. A more important reason why variable geometry is likely to create problems is that the countries that are not part of the initial core will fear financial market pressures as EMU goes ahead without them. For example, the reaction of financial markets to the remarks by the German Minister of Finance in October 1995 shows that news about the likelihood of participation in EMU can have important consequences. The most obvious impact was on Italy; the lira depreciated immediately by several percentage points and interest rates rose by more than one percentage point. This and other similar episodes suggest that the uncertainty about the composition of the participants is likely to lead to increased financial market volatility for those countries that would like to participate in EMU and undertake adjustment efforts mainly for that purpose. If markets perceive that the efforts are undertaken not because fiscal adjustment is desirable in its own right, they might conclude that if the interest rate is high the country will not be able to participate in EMU (because the deficit will be high). This attitude might also lead to a slackening of adjustment efforts. However, another equilibrium is also possible: if interest rates are low, the adjustment required to enter EMU is smaller and hence it is more likely that it will actually be undertaken. Italy has appeared at times to exemplify the ‘bad’ equilibrium, whereas Belgium seems to exemplify the ‘good’ equilibrium. All this implies that for some countries financial market uncertainty could persist until the European Council takes its final decision in May 1998. Until then, financial markets could oscillate between the two equilibria causing large swings in exchange rates and interest rates. Countries that are close to fulfilling the criteria but may not be admitted into the first group because they fall a little short will fear a strong reaction from financial markets. Hence, they could plead that a vital national interest is at stake for them and that EMU should therefore be postponed until they can also join. This would require, however, a Treaty revision, which would put the EU in an extremely difficult position. Alternatively they might even refuse to ratify the Amsterdam Treaty agreed in 1997. The main problems created by variable geometry in the monetary area thus come from economic and political mechanisms that are amplified by financial markets. In a nutshell, the problem is that countries that cannot participate in EMU in 1999 might try to delay its start if they are close to satisfying the convergence criteria. In this sense, it is the ‘near periphery’
EMU and the Outsider Nations
43
that creates more of a problem than the ‘far periphery’ – that is, those countries that are clearly some way from satisfying the convergence criteria or the opt-outs.3 Whether or not countries that are excluded, despite nearly fulfilling the criteria, will suffer a confidence crisis depends mainly on two factors: first on their determination to press ahead with convergence efforts so that they can join the near future and, secondly, on the exchange rate regime between the common currency of the core and the countries with a derogation. This issue is taken up in section 2.3.
2.3 WHAT EXCHANGE RATE MECHANISM FOR THE OUTSIDERS? One of the convergence criteria in the Treaty is ‘the observance of the normal fluctuation margins provided for by the exchange-rate mechanism of the European Monetary System, for at least two years, without devaluing against the currency of any other member state’ (Article 109J). Article 109K2 then goes on to say that once EMU has started, there has to be, at least every two years, an examination of the countries with a derogation. A country can join EMU after such an examination only if it fulfils all the convergence criteria. Hence, the Treaty implicitly assumes that the exchange rate mechanism (ERM) of the EMS will continue to exist. This does not necessarily imply, however, that the system has to continue in exactly its present form. Since the circumstances will change radically once Stage III begins, one could even argue that it has to change. This is indeed what has been decided. The 1995 Madrid European Council asked the EMI to provide a blueprint for an ERM 2 which was agreed upon one year later at Dublin. The main elements are: (1)
(2)
(3)
Participation in the new ERM is voluntary (as for the old one). It could hardly be otherwise. But participation in it for two years is likely to remain a condition for EMU membership later, though this provision is challenged by the United Kingdom (and Sweden). The new exchange rate mechanism operates on the hub-and-spoke system: it is to be based on central rates against the euro, not on a grid of parities, as in the old one. There will be ‘relatively wide’ margins of fluctuation around the central euro rates; the Amsterdam European Council has since confirmed that margins will be ± 15 per cent as in the ERM, but individual ad hoc arrangements to limit the fluctuation margins are possible for countries with good convergence performance.
44 (4)
(5)
Thygesen In principle intervention is unlimited at the margins, but both sides (in practice, the ECB) can refrain from intervening if this would threaten price stability. Adjustment of the central rates should be done in a timely fashion. Both sides, including the ECB, will have the right to initiate a (confidential) procedure to realign an outsider currency.
We have already emphasized that countries which cannot participate in the first wave should make clear that they will continue and perhaps even increase their convergence efforts to be able to join EMU at the next possible date. This would help to reduce the potential financial market instability, but it might not be sufficient to rule out speculative attacks. It was difficult to evaluate the blueprint for the ERM 2 when it appeared because what kind of system will be desirable (and feasible) in 1999 depended on the number of countries likely to remain outside initially and whether or not they had a credible perspective of joining soon. Consider the following two extreme scenarios which both appeared possible in 1996 when ERM 2 was drafted: (1)
(2)
By early 1998 11 countries have made enough progress to participate in EMU and in addition the United Kingdom, Sweden and Denmark elect to ‘opt in’. Only Greece still has an excessive deficit, but she too has made so much progress that she will probably be able to join a couple of years later. Only France, Germany and four to six other small countries start the third stage in 1999. The United Kingdom, Sweden and Denmark confirm their intentions to opt out and convergence in the rest of the Union has been slower than planned so that the remaining countries will need some years before they can join EMU.
It is apparent that the nature of the exchange rate system that will be needed will be radically different, depending on which of these two scenarios is closer to reality. Under the first scenario – which became the more realistic one as to initial economic readiness, but not regarding the opt-outs – there is little need for a fully-fledged exchange rate system, since exchange rates are likely to be stable anyway. By contrast, under the second scenario, there would be a real need for some system to limit exchange rate fluctuations. However, a few simple considerations show it is unlikely under any circumstance that it will be possible to recreate the old EMS. The main reason is the difference in size, which has two implications: asymmetry and bilateralism (or, rather, ‘hub and spokes’): (1)
Asymmetry: The core (even assuming it contains only Germany, France and some smaller countries) will be several times as large in terms of GDP and trade as the set of countries that are outside (excluding the
EMU and the Outsider Nations
(2)
45
United Kingdom which anyway would not be interested). 4 The difference would be even larger in terms of the size of financial markets and the reputation for stability. A formally symmetric system like the EMS becomes impossible under these circumstances. In the pre-1992 system Germany played a central role, although in terms of trade and GDP, she never accounted for more than 45 per cent of the area covered by the ERM. ‘Hub and spokes’: The trade of the periphery with the core that will represent the ‘hub’ of the system is several times larger than the trade among the ‘spokes’ (that is, the likely outsiders). Hence the new system will not be multilateral. It will effectively be bilateral in the sense that it regulates the bilateral relationship between the ECB and each national central bank of the outsiders.
This does not imply that each outsider should have a different agreement. On the contrary, the same type of arrangement could be offered to all outsiders. However, even if all outsiders are willing to sign up to the same system it will de facto not work in the same way for each participant. The case of the United Kingdom does present a particular problem since the British government decided to opt out and also refused to enter into an agreement concerning the exchange rate of sterling, an example that might be followed by Sweden. The other opt-out country, Denmark, has already indicated that it will seek a narrow-margins arrangements with the ECB. Hence, Denmark and the remaining non-participant (Greece) are likely to be willing to cooperate with the ECB to stabilize exchange rates. It is sometimes argued that countries that remain outside will be tempted to resort to competitive devaluations. But this fear, which is based on recent experience, seems to be unfounded. First of all, exchange rates are difficult to control since they evolve with market expectations about future policy. Hence, it will be difficult for any government (or national central bank) to ‘engineer’ a competitive depreciation without running a serious risk of starting a cycle of inflationary expectations and high interest rates that is difficult to control. Moreover even large exchange rate changes tend to have only a limited impact on unemployment and the trade balance. Finally, since the criterion on exchange rate stability will continue to apply, any country that is tempted by this policy would know that the price for a ‘beggar-thy-neighbour’ policy of this type would be a further delay in EMU membership. Over most of 1996 the EMI prepared a framework arrangement for ERM 2 which was accepted first by ECOFIN and later by the European Councils in Dublin and Amsterdam. It was a difficult compromise and it will remain incomplete until it can be finalized by the ECB, in the light of the considerations outlined, including the number of non-participants.
46
Thygesen
How should one evaluate the compromise reached on ERM 2? It might seem ungenerous to criticize the proposed solution to an issue that seemed extremely difficult at the outset. When the discussion of an arrangement for the ‘outs’ started in 1995, pessimism prevailed. Given the strongly divergent views of exchange rate constellations at the time, accentuated by allegations of competitive devaluations, aversion to intervention commitments for the ECB and divergence of interests among thc prospective outsiders, the odds on agreeing on a workable ERM 2 seemed highly unfavourable. It is quite an achievement that the EMI was in the end able to come up with an agreed report, even though on closer examination some important provisions remain more open-ended then is desirable. When reading the text issued by the EMI in October 1996 and annexed to the Dublin conclusions one is, however, struck by the impression that it appears to to have been drafted not so much with the more immediate problems of what to do with the non-participants in Stage III as EMU starts on 1 January 1999 as with the longer-term problems in an EU with several new Member States from Central and Eastern Europe some time over the coming decade. Another interpretation is that officials were concerned about winning the last war – that is, they wanted to avoid at all cost a repetition of the problems of 1992–3 and 1995, not recognizing that circumstances will be totally different from 1999 onwards. Not only is participation in ERM 2 to be voluntary – as was already admitted above, that may have been unavoidable – but the whole framework is not tailored for countries that have long been actively involved in convergence efforts under the Maastricht Treaty and have subscribed to the Stability Pact. Some of those initially left outside, maybe all except Greece, might have missed entry narrowly with an economic performance only marginally inferior at the photo finish to those of the entrants. To them, some of the provisions must seem less than reassuring. This will also apply to the two countries which presently express a preference for remaining outside ERM 2 (Sweden and the United Kingdom), providing them with an additional argument for doing so. It certainly applies to Denmark, which is openly seeking a tighter arrangement despite its opt-out from monetary union itself. Two provisions in the framework point toward this interpretation. The ECB commitment to intervention is to be conditional and the ECB reserves – indeed emphasizes – its right to suggest realignments of the central rates for outsider currencies. An important substantive point in any future arrangement between an outsider and the ECB is the extent to which interventions will be, in principle, unlimited and unconditional at the margins. It can be argued that focusing mainly on mandatory interventions at the margins is misleading, as long as the margins are anything like as wide as in the present EMS (see below). A currency that is driven to its wide margins will no doubt need to be defended
EMU and the Outsider Nations
47
strongly by raising domestic short-term interest rates – or a realignment will have to be undertaken. Intramarginal interventions will then become the important ones, but unfortunately little is said to reflect the new situation relative to present arrangements which are to continue in the form agreed at Basle–Nyborg in 1987. This Agreement implies that unilateral intramarginal interventions by an outsider in euro will be subject to prior approval by the ECB above certain thresholds and that access to the Very Short-Term Financing Facility (VSTFF) for intramarginal interventions will continue to be subject to the ceilings agreed upon nearly 12 years ago. However, these ceilings no longer seem adequate in tomorrow’s climate of very high capital mobility, which did not exist in 1987. It is important not to restrain the freedom of action of the outside currencies in defending their central rates; in any case, the short duration of the credit facilities constitutes a deterrent against excessive reliance on them. An overhaul of the Basle–Nyborg Agreement seems overdue. Maintaining the principle of mandatory interventions at the wide margins is more important than it may seem, because margins may be narrowed when EMU starts. Countries missing entry in 1998 may in the end seek narrow margins, as part of an understanding that they will enter soon – even though the prospective ‘pre-ins’ have not so far asked for that. They clearly want to give prime emphasis to their candidacy for the first group rather than to give the impression that they are actively seeking a reinforcement of the safety-net arrangements available to outsiders. So although only Denmark has so far shown an interest in a narrow-margins arrangement of the pre-1993 EMS type, others may still do so in 1998 or later. Fortunately the possibility of setting up stronger links with outsiders than the presently envisaged continuation of the lax EMS, with some limited modification to reflect its future euro-centred nature, has been retained; they could take the form either of publicly announced narrower margins or of informal target ranges inside the normal margins, provided the out-country has achieved a sufficiently high degree of convergence. Another potentially worrisome feature of the framework is that all parties to the agreement, including the ECB, would have the right to initiate a confidential procedure aimed at reconsidering central rates (EMI, 1997, p. 69). In other words, flexibility in the form of wide margins was not considered enough; realignments are openly envisaged. Jointly, these provisions amount to a revival of the so-called ‘Emmingerletter’ of 1978 summarizing the understanding between the German Federal Government and the Bundesbank that otherwise mandatory interventions might be suspended and a realignment initiated if interventions were perceived to threaten price stability in Germany. This understanding, which had almost been allowed to lapse into oblivion, was finally invoked in September 1992 when interventions in support of the then weak currencies
48
Thygesen
of the EMS, notably the lira, seemed to throw Germany’s capacity to stick to its chosen monetary target into doubt. In retrospect, the problem looks less serious than was claimed at the time – and still today – by German officials (see Gros and Thygesen, 1998, Chapter 3). If the problems turned out to be manageable in the situations of extreme tension in 1992–3, they are more likely to be manageable also for the ECB, which will be much larger relative to any outsider currency than was the Bundesbank. A greater readiness to contemplate the constructive role of interventions also in periods of considerable tension would have seemed appropriate, both vis-à-vis the Southern European countries likely to be only in the ERM 2 for a short transitional period (if at all) and for the two reluctant outsiders, Sweden and the United Kingdom, who need to be persuaded that the new system does offer a more reassuring option than individual floating. (That they have doubts on this point is understandable in view of their painful experience of futile individual defence of their currencies in 1992). In short, the ERM 2 framework does not appear to be helpful and confidence-inspiring for countries that for one reason or the other are unable to join the first EMU group. If the participants in the euro area are as keen as they proclaim to be to retain a close grip on the exchange rate relations with the initial outsiders, the ECB will have to come up with some improvements to the outline prepared by the EMI – or, at least, with a bold interpretation of the possible tighter linkages foreseen in the framework. Particularly for a pre-in country which might have only just missed entry into monetary union, a radical approach would be to offer linkage to the euro with narrow, or even zero, margins while still leaving the responsibility for defending the currency to the out country. The latter would give up all monetary autonomy in return for this link, much as happens for a currency board which has linked the value of its currency to a major international currency and the issue of it to fluctuations in its reserve holdings of the latter (the euro). A concrete proposal is developed in section 2.4.
2.4
A CONCRETE PROPOSAL
If one or more countries are to be excluded from the start of EMU, despite a desire to participate, is there anything that can be done to limit the damage? For Italy, in particular, with high public debt, a strong reaction by financial markets could make the fiscal adjustment even more difficult through a higher interest burden. The fear in financial markets (and of many politicians) is that the constitution of the core group is not only a temporary measure and that it will be more difficult to join later. This is certainly not the intention of the Treaty. But what if initial exclusion makes convergence much more difficult, even if there is some exchange rate mechanism for the outsiders?
EMU and the Outsider Nations
49
One solution could be to grant an associate status in EMU. The country concerned could be invited to come under the EMU umbrella to benefit from lower interest rates, but it would not participate in the management of EMU until it had converged in fiscal terms as well. This arrangement could be achieved technically by a unilateral declaration that the country concerned accepted all the obligations arising from membership in EMU,5 but it would be preferable to have a formal agreement between the ECB and the country concerned, with the political support of the ECOFIN Council because this would make it much more credible with the markets. The agreement would specify that the national central bank agreed to follow the monetary policy of the ECB as if it were a full member of the EMU. At the same time, however, it would be clear that for purposes of decision making in the ECB (and ECOFIN), the country would continue to be treated as having a derogation. In essence, the country would give up its national monetary policy and replace it with that of the ECB. More precisely, this means that the exchange rate would be irrevocably fixed, the payments systems would be unified, actions by the ECB would have direct effect in the country concerned and the decisions of the ECB would have to be applied by the national central bank, a portion of whose foreign exchange reserves would be pooled in the ESCB like those of the full participants. Moreover, the country concerned would be subject to the full excessive deficit procedure.6 All this would be officially acknowledged by the Union in conjunction with a convergence programme outlining how the country would, with the help of lower interest payments, satisfy the fiscal criteria by a certain deadline.7 Acknowledgement by both the Union and the ECB would make this arrangement credible and would ensure that market interest rates in the country concerned converge quickly to the level in the core.8 This sort of associate status in EMU will deliver the benefits in terms of lower interest rates, however, only if it is credible. Credibility should come already from the endorsement given by the ECB, but it would be immensely strengthened if markets see that the exchange rate can be defended under any circumstance. This should be the case if one views the proposed arrangement as a ‘Currency Board’. A Currency Board is credible if the national central bank possesses adequate foreign reserves to guarantee conversion of all of its liabilities (the monetary base) into the common currency of the core. Would this be the case? Most of the central banks that might be candidates for this status do indeed have enough reserves to make the currency board approach credible (see the Appendix, p. 53). All the candidates for associate membership in EMU could thus be confident of operating a tight link with the core even under a worst-case scenario. If the markets know that there is no chance that they can force a break in the link to the common currency of the core, they will regard it as
50
Thygesen
credible. Any country that chooses this approach could increase credibility even further by passing a law that obliges the national central bank to defend the exchange rate through unsterilized interventions. Moreover, it is likely that if there were really a totally unjustified speculative attack, the ECB would help the country concerned. If the underlying fundamentals are sound, credibility should thus not be a problem. If the fundamentals are not sound, the ECB would not in any event recommend this approach and no country would (or should) dare to try it against its advice. Technical viability is of course only a necessary, and not a sufficient, condition for the stability of a currency board arrangement. The reason why central banks usually sterilize their interventions is that the large increases in interest rates that might result if they did not are unacceptable, either because of their macroeconomic consequences (the United Kingdom in 1992), or because they could endanger the stability of the banking systems (Sweden also in 1992). Central banks and governments will have to convince markets that they will be willing to accept interest rate increases if the market tests their resolve. The experience of Belgium, which faced a test of its commitment in 1993, shows that it is possible to present this case persuasively. Of course, associate membership in EMU can not constitute a magic wand that eliminates all problems. But it could represent an option for countries that are close to qualifying for full membership in 1998 or on future occasions. Irrevocably fixing the exchange rate with the prospect of full participation in EMU after a couple of years is fundamentally different from fixing exchange rates in the environment of the 1980s (with high inflation and variable exchange rates) or during the early 1990s (when some currencies were clearly overvalued). The argument that experience has shown that fixing the exchange rate is impossible because financial markets could attack any exchange rate should thus not be overrated. There will be little reason for financial markets to attack an exchange rate if inflation is low, deficits are close to 3 per cent of GDP (possibly even below), debt ratios are declining and the external current account indicates a good competitive position. The political viability of this idea depends upon its presentation. If it is characterized as a means of circumventing the convergence criteria, which is not the purpose of this proposal, the core will veto it. The scheme merely aims to help the peripheral countries bridge the gap that separates them from the core without softening the convergence criteria for full participation in EMU. Only countries that have done their basic homework could be encouraged to pursue this approach. In order to qualify, the deficit when recalculated at German interest rates should at a minimum fall below 3 per cent of GDP. This guideline would also ensure that the debt ratio would be declining once
EMU and the Outsider Nations
51
the lower interest rates took effect. It bears reiterating that the country in question would have a derogation in the decision making organs of the ECB, which would underline that the convergence criteria had not been suspended. In a sense, Belgium and Austria have already in the past successfully opted for this course by pegging unilaterally to the D-mark. Could others follow their example? The real test of this approach will come in the case of a large country, for example Italy. Large countries have always experienced more problems in acquiring stability through the exchange rate. But in this case, they would not attempt to use the exchange rate to force adjustment in prices or wages. Their problem is that they are stuck in a low credibility–high interest rate trap out of which it is difficult to escape without outside help. The decision to participate in EMU, even if essentially on a unilateral basis, would constitute one large step away from this trap. Of course, this can only facilitate adjustment. It is in no way a substitute for the resolute fiscal action that has to be taken anyway.
2.5
CONCLUSIONS
This chapter has argued that, given the fact that the original ambition, expressed for example in the Delors Report, that all EU Member States would join EMU from the start, has turned out to be impossible to achieve, the Treaty has dealt in a realistic way with the issue of variable geometry in European economic integration. Furthermore, in view of the diversity of the prospective outsiders, the arrangements for them have to embody some flexibility. But the recent evolution in the prospects for participation, which may in the end confine the group of outsiders to those countries which are likely to opt out (the United Kingdom, Denmark, Sweden) plus Greece and – maybe – Italy suggests that the approach to arrangements with the outsiders should now become more targeted. The framework agreement on ERM2 is a start, but it may turn out to be an empty shell: the United Kingdom and Sweden appear to want to invoke the voluntary nature of participation to remain outside ERM 2, while Denmark has indicated its preference for a much tighter arrangement. If Italy were not to join initially for economic reasons a firmer arrangement for the lira would have to be sought to minimize the costs of exclusion and facilitate subsequent entry. The chapter suggests a form of associate status in this case. On closer examination, the proposed ERM 2 may be suitable only for Greece and for the new Member States from Central and Eastern Europe which are expected to join the EU early in the next century. The ECB, the ECOFIN Council and the European Council will therefore have important matters on the agenda also after the contours of the first group of participants emerges in May
52
Thygesen
1998. In their own interest the EMU participants will have to devote further attention to relations with outsiders within the EU. So will, obviously, the outsiders themselves. This chapter has not looked primarily at the choice facing them from their own point of view. Some who remain outside will have a choice between joining ERM 2 and continuing with a purely domestic stabilization objective – that is, an inflation or a price level target – and they will exercise this choice on the basis of their particular historical experience and their evaluation of the risks of speculative attacks in an ERM 2 caused either by monetary shocks in EMU or divergence in their own performance, as well as of their own likely future exclusion from EMU by remaining outside ERM 2. It is in the interest of the EMU participants to encourage the outsiders to enter ERM 2 by making the latter a more genuinely cooperative arrangement than the present outline and the historical experience with the present system suggest. In this context it is encouraging that several of the prospective outsiders – the United Kingdom, Sweden and Denmark – have taken the main principle of the Stability and Growth Pact more literally to heart than have most of the EMU candidates by all aiming for balance – or a small surplus – in their public finances on average over the business cycle. This will help to assure that their convergence is sustained without the formal underpinning of participation in EMU itself.
Appendix: Technical Conditions for the Viability of a Currency Board Technically a currency board is viable if the central bank has enough reserves to exchange all its liabilities (the monetary base) into foreign currency. This is the case if the foreign exchange reserves are larger than the monetary base. Column (3) in Table 2A.1 below) shows that the ratio of foreign assets to the monetary base is above, or close to, 1 for all countries except Italy. If one takes into account that, even in a worst-case scenario, few people will exchange their holdings of cash, it would be sufficient for a national central bank to have enough foreign assets to cover the remainder of the monetary base (that is, required reserves). Column (4) shows that this is the case by a large margin for all the countries considered below, with the exception of Italy and Finland. Even for these two countries, however, the shortfall is rather limited since the existing foreign exchange reserves already cover 100–110 per cent of the reserves held by banks with the central bank. In the Italian case, this result is due to the unusually large reserve requirements imposed on banks in Italy. The reserve ratios in Italy are at present much higher than in the rest of the EU and have to be lowered anyway if Italy wants to participate in EMU. If reserve requirements were halved in Italy, still leaving them much above the EU average, the
Table 2A.1
Greece (bill. Drachma) Italy (bill. Lira) Portugal (bill. Escudo) Spain (bill. Peseta) Sweden (bill. Kroner)
Reserves and monetary base
(1) Base money
(2) Foreign assets
2500.1
Ratios based on: Monetary base (3) = (2)/(1)
Required reserves (4) = (2)/((1) – cash in circulation)
2975.5
1.2
4.6
150.0 3001.3
86.6 3706.7
0.6 1.2
1.0 1.6
7.800 163.8
6.152 175.7
0.8 1.1
4.8 1.8
Source: International Monetary Fund, International Financial Statistics.
53
54
Thygesen
international reserves of the Banca d’Italia would be much larger than the mobile part of the monetary base. (Some margin is needed since part, say 20–25 per cent, of the foreign reserves will be pooled in the ECB.) Hence, even Banca d’Italia could defend the exchange rate if it previously lowered required reserve ratios towards the EU average – provided, of course, that it did not engage in sterilization, as so often in the past.
Notes 1.
2. 3.
4. 5.
6.
This chapter draws on Gros and Thygesen (1998), and the author wants to record his appreciation of a number of years of cooperation with Daniel Gros. Comments from my two discussants, Arne Jon Isachsen and Jesper Jespersen are acknowledged, even though I have not followed their advice to broaden the chapter with an analysis of the likely stability of EMU. Nor have I systematically updated the chapter to take into account the more recent (March 1998) developments surrounding participation in EMU; the chapter stands much as delivered in June 1997. Research at EPRU has been financed by a grant from the Danish National Research Council. See article 109k5. In principle, all countries signed up to the conclusions of the Madrid Council of December 1995. But important politicians in some countries stated in public that postponement should be considered. It would, however, be dangerous to yield to a demand for postponement because there will be differences in the time required by the ‘near periphery’ to fulfil the convergence criteria. Establishing the principle that everybody should wait for the ones that are close to catching up might set in motion a long chain of countries that were formerly far from meeting them and then come close to being able to participate, as others graduate from being close to actually fulfilling the convergence criteria. In the meantime, all candidates would have to bear the cost of the higher risk premia that will persist until EMU really comes into existence. Moreover, it is important to establish the principle that no single country should impede the others from going ahead. A Protocol to the Maastricht Treaty was designed to achieve this. The combined GDP of France, Germany, the Netherlands and Finland, plus Ireland and Luxembourg is about 3,000 bn ecu, compared to 800 bn ecu for Portugal, Italy, the United Kingdom, Greece, Belgium, Sweden and Denmark. The government would have to declare that it accepted the obligations arising from Articles 104c(9) and (11) (excessive deficits procedure); 105(1), (2), (3) and (5) (monetary policy); 105a (notes and coins); and 108a (empowering the ESCB). The national central bank would also accept the obligations resulting from the ESCB statutes (Articles 3, 9, 12.1, 14.3, 16, 18, 19, 20, 22, 23, 30–34 and 52). However, the restrictions specified in paragraphs 3–6 of article 43 of the ESCB statutes would apply. In addition, the country concerned would not participate in decisions under Articles 109 (exchange rate system with rest of world) and 109a(2)(b) (membership of the executive board of the ECB). Unilateral restricted participation would in effect be a sort of Anschluss much like the period when the D-mark was introduced in the territory of the former GDR. This might not be a very enticing example. In the case of the EU, however,
EMU and the Outsider Nations
7. 8.
55
convergence will take place before, not after, monetary unification. Hence, the economic difficulties that followed German unification should not occur. Acknowledgement by the Union should not necessarily imply that the ECB, when setting its monetary policy, would take into account economic conditions in the country that participated in the same way as those of ‘regular’ participants. The actual debt service burden would decline only gradually, however, until the outstanding high interest debt is retired. Depending on the maturity structure, this might take two years.
References De Grauwe, P. (1997) ‘Exchange-rate arrangements between the ins and the outs’, CEPR Discussion Paper, 1640 (London: Centre for Economic Policy Research) (May). European Monetary Institute (EMI) (1997) The Single Monetary Policy in Stage Three (Frankfurt: EMI) (January). Genberg, H. (1997) ‘Monetary and exchange-rate policy outside a European Monetary Union’, Swedish Economic Policy Review, 4, pp. 155–88. Gros, D. (1996) ‘Towards Economic and Monetary Union: problems and prospects’, CEPS Paper, 65 (Brussels: Centre for European Policy Studies) (January). Gros, D. and N. Thygesen (1998) European Monetary Integration: From the European Monetary System to Economic and Monetary Union (London and New York: Addison Wesley Longman). Pisani-Ferry, J. (1996) ‘Variable Geometry’ in Europe (Paris: CEPII). Thygesen, N. (1996a) ‘The prospects for EMU by 1999 – and reflections on arrangements for the outsiders’, paper presented at the Centre for Economic Policy Research/Banca Nazionale del Lavoro Conference (Rome) (26 February). Thygesen, N. (1996b) ‘Interpreting the Exchange-rate Criterion’, in Peter B. Kenen (ed.), ‘Making EMU happen – Problems and Proposals: A Symposium’, Princeton Essays in International Finance, 199 (Princeton: Princeton University) (August).
Comment Arne Jon Isachsen 1
INTRODUCTION
In the Nordic countries Professor Niels Thygesen is ‘Mr Euro’. Not only has he studied the issue of monetary union in Europe, as a member of the Delors Committee he has also contributed directly to the drafting of the process of creating one. The rest of us, his fellow Nordic economists, can only express gratitude to Niels for his continued interests in and analysis of developments in European monetary affairs. In his Chapter 2, after having argued that Italy may find a Currency Board Arrangement (CBA) within ERM 2 attractive, Professor Thygesen submits that: On closer examination, the proposed ERM 2 may be suitable only for Greece and for the new Member States from Central and Eastern Europe which are expected to join the EU early next century. Professor Robert Mundell, who provided us with the concept of Optimal Currency Area (OCA), goes even further, arguing that Greece may also join EMU in the first wave. Mundell also reminds us that: Before the middle of 1996, few economists would have given Portugal, Spain and Italy much chance for entering the EMU in the first group (Mundell, 1997, p. 215). Clearly, the forecasting of developments in European monetary affairs is a risky business. At present (December 1997) it seems that any remaining uncertainties about the timely launching of the EMU have more to do with France and Germany than with Spain and Italy. In what follows we assume that the European monetary union gets on the air on schedule, and that it turns out to be reasonably successful.1
2
A SUGGESTED TAXONOMY
For a starting point, when studying the process of monetary integration in Europe, each of the 15 EU countries could be allocated to its proper cell in Table C2.1.
Table C2.1
Would like to join Would not like
To be or not to be a member of EMU Is able to join
Is unable to join
1 3
2 4
56
Comment
57
Countries in cell 1 are the ‘ins’; the other three cells are inhabited by ‘outcountries’. As already alluded to, since mid-1996 several countries have moved out of cell 2 and into cell 1 (Italy, Spain and Portugal). The only remaining country in cell 2 seems to be Greece. With close to a blocking minority of countries in cell 2, the launching of the EMU project would have been more difficult. Cell 4 is probably empty, although two years ago, with public finances out of control, Sweden arguably belonged here. Cell 3, those who are able to join but will refrain from doing so, is occupied by Denmark and the United Kingdom, and more recently by Sweden, although it never secured a formal opt-out clause.2 Summing up: Cell 1 contains 11 countries. Of the remaining four, Greece is in cell 2, and Sweden, the United Kingdom and Denmark in cell 3. Of the latter three it seems reasonable to assume that Sweden and the United Kingdom will continue with inflation targeting for the next few years and thus not be interested in the fixing of exchange rates through ERM 2. Thus, in the short term Professor Thygesen’s discussion of how ERM 2 should be organized is relevant only for Greece and Denmark. Prospective new EU members from Central and Eastern Europe, however, may position themselves in light of the rules governing ERM 2. Estonia is a case in point.
3
ERM 2 – TOO FLEXIBLE AN ARRANGEMENT?
Today’s ERM is based on a parity grid, meaning that the value of each currency must not deviate by more than 15 per cent from its central rate against any other currency in the mechanism. This implies that the room for fluctuations between any pair of currencies is much more limited. ERM2 maintains fluctuation margins of 15 per cent. However, as each currency in ERM2 will be pegged to the euro, between any pair of ‘outs’ in ERM2, much larger fluctuations in exchange rates are feasible. To illustrate, assume two ERM2 countries with currencies A and B, respectively, and with central rates on par with the euro. Now, letting currency A in period 1 appreciate to its allowed maximum and currency B likewise depreciate, only to reverse the situation in period 2, the following picture emerges: Central rates: At t = 1: At t = 2:
100 A = 100 euro = 100 B 85 A = 100 euro = 115 B 115 A = 100 euro = 85 B
From period 1 to period 2 the price of 100 B, in terms of A, increases from 74 A to 135 A – that is, a depreciation of currency A against currency B of 82 per cent. A professor in linguistics, I presume, would object to calling such an arrangement a fixed peg system. The scepticism in Sweden and in the United Kingdom to the requirement of participating in ERM2 at least two years before joining EMU is well founded. However, joining an exchange rate mechanism with such wide bands makes the continuation of inflation targeting unproblematic. With the option of entering into narrow-margins arrangements with the ECB, the concept of a fixed peg can still be retained. Denmark may well consider such an arrangement, having successfully adhered to a hard currency policy in the last 15 years. The ultimate peg, however, would be, as suggested by Professor Thygesen, a Currency Board Arrangement (CBA).
58 4
Isachsen A CURRENCY BOARD ARRANGEMENT MAY SUIT GREECE
Professor Thygesen argues that a formal agreement between an out country and the ECB, in which the out country concerned gives up its national monetary policy and replaces it with that of the ECB, is worth considering. Such an arrangement ‘would ensure that market interest rates in the country concerned converged quickly to the level in the core’ (p. 49). However, he adds that the benefit of such a CBA will materialize only if it is credible. Credibility, in turn, means adequacy of foreign reserves – that is, enough to cover the monetary base. In an Appendix Professor Thygesen reveals that in Greece, foreign reserves exceed the monetary base by 20 per cent. However, Frankel (1996) maintains that to deal with potential banking crises, enough reserves to back the entire domestic money supply may be required. Frankel is probably asking for too much. During the banking crisis in Norway in the early 1990s, the government had to spend 25 billion kroner (NOK) to shore up failing banks. At the time base money amounted to 40 billion NOK – that is, public support to the banking system came to more than 60 per cent of the monetary base. With reserves measured in NOK of about 140 billion, rescuing the banks had only minor effects on the exchange rate.3 With a CBA for the drachma against the euro, a similar crisis situation could have been difficult to contain without more ample supply of foreign exchange. Professor Thygesen, however, claims that Greece has enough reserves to make the CBA approach credible. If ‘credibility’ is taken to mean that the interest rate on the drachma ‘would converge quickly to the level of the core’, he might perhaps be a little too optimistic. In a recent paper from the IMF it is argued that a large outstanding stock of shortterm government debt, even if the current fiscal deficit is sustainable, may make CBAs ‘vulnerable to bond-led speculative attacks’ (Balino and Enoch, 1997, p. 15). With government debt of 110 per cent of GDP in Greece one should not completely rule out such an unpleasant possibility. Obviously, the viability of a CBA for outs would improve with increased lines of credit from the ECB. As noted by Professor Thygesen, an overhaul of the Basle–Nyborg Agreement seems overdue. Estonia established a CBA in June 1992, and has been able to hold on to this arrangement in spite of a severe banking crisis the following year.4 After five years of keeping the Estonian kroon steady against the German mark, at 8:1, the interest rate differential is still coming down. During the first three quarters of 1997 the 3-month interbank rate in Estonia fell by one percentage point, from 8 to 7 per cent. In this period, the comparable German interest rate hovered around 3.3 per cent. Now, it seems reasonable to assume that Professor Thygesen’s idea of a CBA for Greece, based on a formal agreement with the ECB, would be more credible than Estonia’s unilaterally fixed peg. By way of illustration, let a Greek CBA bring the interest rate on new government debt down to 6 per cent, that is, a reduction of four percentage points. With outstanding debt of 110 per cent of GDP, totally refinanced in the course of a good year, government interest payments would amount to about 6.6 per cent of GDP. Maintaining a surplus on the primary balance of 4.7 per cent of GDP (as forecast for 1997), Greece would then meet the 3 per cent deficit criterion. Reduced interest payments on government bills and bonds mean less transfer of resources to holders of these assets. For high-interest rate countries this is probably the greatest short-term advantage of joining the EMU. Through a CBA also Greece could probably reap some such benefit. New member countries from Central and Eastern Europe will imply a reduction in net EU transfers to the ‘Club-Med’ countries. However, in terms of lower interest
Comment
59
rates on government debt, this latter group of countries stands to gain the most from the EMU. From a political point of view ‘deepening’ as a prerequiste for ‘widening’ then becomes easier to understand.
5
BUT FOR DENMARK A CBA DOES NOT SEEM APPROPRIATE
Currently Denmark is one of the five EU countries not subject to surveillance under the excessive-deficit procedure applying under the second stage (OECD, 1997, p. 32). Economically fit for the EMU, Denmark is not politically. Joining the EMU requires a referendum, and with the ‘No’ to Maastricht of 2 June 1992 fresh in mind, the government is not likely to put the EMU issue on the ballot in this century. A CBA, which must be considered as a prelude to full membership of the EMU, would not go well down with Danish voters. A more realistic solution is membership of ERM 2. As indicated by Professor Thygesen the option of narrow-margins may suit Denmark well. When the rationale for having a separate currency is to keep the option of later devaluating it, a positive risk premium on the interest rate seems appropriate. However, Denmark does not have economic incentives to abandon her hard currency policy. Thus, the argument that staying outside the EMU may ‘entail a lasting premium to long-term interest rates’ (OECD, 1997, p. 36) may not be valid. If the EMU turns out to be ‘softer’ on inflation than envisaged by most experts, over time the Danish krone may be increasingly appreciated by wealthy owners, thus commanding a negative risk premium versus the euro. Von Hagen (1997), for one, would not be surprised if a large EMU also becomes a soft one. The concept of deepening the EU is not a particularly popular one in Denmark. A primary cause for Denmark to join the EEC in 1972 was the modest one of securing higher prices for agricultural export. However, the option of remaining outside the EMU as, say, Sweden and the United Kingdom join, will be less tempting. More generally, staying put in cell 3 in our Table C2.1 may not be a viable solution in the long run.5
Notes 1.
2. 3. 4. 5.
Professor Martin Feldstein is one of several American economists who is very sceptical about EMU. France, according to Feldstein (1997), wants the EMU in an aspiration for equality, which is incompatible with Germany’s expectation of gaining hegemony through the establishment of a monetary union. He argues further (p. 31) that with a politically determined exchange rate policy, the average level of future inflation ‘will almost certainly’ increase. In formal terms, not being a member of the ERM should ‘qualify’ Sweden for ‘not qualifying’ to join the EMU. The total money stock (M3) amounted to about 540 billion NOK (see Isachsen, 1997). Here also the option of having a CBA for Norway is discussed. In the course of 1993 the number of commercial banks in Estonia was sharply reduced, from 42 to 22. Feldstein (1997, p. 14) maintains that ‘when Danish voters initially rejected EMU and decided against ratifying the Maastricht treaty, Denmark was threatened with losing the free trade benefits of the Single Market’.
60
Isachsen
References Balino, T.J.T. and C. Enoch (1997) ‘Currency Board arrangements. Issues and experiences’, IMF Occasional Paper, 151 (Washington DC: IMF). Feldstein, M. (1997) ‘The political economy of the European Economic and Monetary Union: Political sources of an economic liability’, NBER Working Paper, 6150, Journal of Economic Perspectives, 11(4) (fall). Frankel, J. A. (1996) ‘How well do foreign exchange markets function’?, NBER Working Paper, 5422. Isachsen, A.J. (1997) ‘Norsk penge- og valutapolitikk i støpeskjeen’, Internasjonal Politikk, pp. 505–27. Mundell, R.A. (1997) ‘Currency areas, common currencies, and EMU’, American Economic Review, Papers and Proceedings (May), pp. 214–16. OECD (1997) OECD Economic Surveys 1996–1997. Denmark (Paris: OECD). von Hagen, J. (1997) ‘Monetary policy and institutions in EMU’, Swedish Economic Policy Review, pp. 51–116.
Comment1 Jesper Jespersen 1
INTRODUCTION
Niels Thygesen has presented an analysis of some aspects of the macroeconomic implications for the non-EMU countries of different exchange rate arrangements with the EMU. He explicitly leaves out the discussion of the conditions for intra-EMU stability which I shall argue is a severe limitation of the analysis. Furthermore, his theoretical presumption is a preference for institutional arrangements that imply a high degree of monetary stability within the entire EU. Thygesen does not consider real sector instability as a long-lasting problem when deciding upon currency matters. It is hypothesized that the functioning of the entire EU with regard to macroeconomic stability would be increased if the ‘outsiders’ formally obtained exchange rate links with the EMU. Hence, the chapter is concluded on this point by recommending a bilateral currency arrangement between each ‘outsider nation’ and the European Central Bank which in practice may entail a functional membership of the EMU (‘without voting rights’). Thygesen also considers an alternative case where a so-called ‘ERM 2’ is established with the euro as the core currency. In that case, the chapter favours even more narrow margins than those prevailing before the 1992–3 crisis. The possibility of another breakdown of such an arrangement is not contemplated within the chapter. On the contrary, Thygesen concludes that an ‘ERM 2’ arrangement is a second-best solution because of too much monetary flexibility – better suited for would-be members of EU in Central and Eastern Europe.
2 CAUSES OF MACROECONOMIC INSTABILITIES – AN UNSETTLED QUESTION Beneath the chapter’s arguments there is a vision of a basically stable market economy – especially with regard to the real sector of the economy. Unemployment is determined by structural factors outside the domain of macroeconomic stabilization policies. Hence, monetary instability mainly (if not only) originates from national (short-sighted) policies. At this point the chapter is, of course, in line with the main conclusions from the Committee for the Study of Economic and Monetary Union (‘the Delors Report’),2 expressing a firm belief that the stronger each country has committed itself to ‘German monetary policy’ the more credible the financial market participants will consider the policy and the faster the convergence to overall macroeconomic stability will take place. One of the few dissenting voices within that Committee came, in fact, from the govenor of the German Bundesbank Karl Otto Pöhl. He expressed severe doubts about the possibility of creating macroeconomic stability solely by Member States committing themselves to monetary policy à la Germany:
61
Jespersen
62
A Monetary Union presupposes considerable shifts in the responsability for economic policy to a central authority … Although complete political and institutional union is not absolutely necessary for the establishment of a monetary union, the loss of national sovereignty in economic and monetary policy associated with it, is so serious that it would probably be bearable only in the context of extremely close and irrevocable political integration (p. 144). Isolated steps in the monetary field would overburden monetary policy in political terms and jeopardize the credibility of the process of unification in the longer run (p. 155). The rising unemployment in EU since the signment of the Maastricht Treaty, especially in the major potential EMU Member States has enforced this view point that the stability of the real sector cannot be taken for granted. Europe is, for the time being, experiencing a high level of unemployment. Taking the restrictive macroeconomic policies pursued through the 1990s within the major EMU countries into consideration, it is difficult to avoid the impression that employment and growth have been hampered and that the effects of that policy seems to be long-lasting. If so, the development through the 1990s gives support to the view expressed by the German Bundesbank governor that a precondition for a well functioning monetary union is a highly coordinated and responsive macroeconomic policy within the common currency area. This is not a part of the institutional framework of the Maastricht Treaty (not to speak about the ‘Stability Pact’). If the EMU starts out with a relatively large number of participants there is even an increased risk of generating strong real sector imbalances inside the EMU. The judgement on what kind of potential destability of the EMU area is of crucial importance when giving recommendations on the best suitable exchange rate arrangements for the ‘Outsider Nations’. In fact, a number of economists with a monetarist inclination (Martin Feldstein, Charles Wyplosz and even Milton Friedman) have quite recently expressed doubts about the EMU project and the destabilizing elements inherent in the process. Whether macroeconomies do converge toward general equilibrium by themselves is still a theoretical unsettled question, but the empirical findings are not very supportive.
3
ERM 2?
On the question of an ERM 2 it is necessary to separate the two different categories of ‘outsider nations’. One group of countries is prepared to join the EMU unconditionally for (mainly) political reasons. This group counts of countries from Southern (and Eastern) Europe. Participation in European Institutions is desirable for them for historical and political reasons. Their main interest is a firm commitment from the core EMU countries not to ‘be left behind’ in the ongoing European integration process. For them, the economic consequences are of second-order magnitude and/or they look forward to be compensated through the Cohesion Fund. I will not comment on these cases. The second group consists of countries from Northern Europe (Great Britain, Sweden, Denmark, Norway and Switzerland). They have expressed different degrees of doubts on the usefulness of a membership of the common European currency. These more hesitant attitudes are a mixture of scepticism towards too fast political integration and uncertainty about the macroeconomic arguments in favour of a Monetary Union at the actual stage of economic integration. I shall concentrate on the latter argument.
Comment 3.1
63
Are the Necessary Macroeconomic Conditions Fulfilled for ERM 2?
From a welfare point of view, a stable exchange rate cannot be a goal in its own right. It might be a useful tool promoting more fundamental considerations: (a) low inflation, growth and employment and/or (b) political aspirations. As mentioned already, Thygesen is in favour of a new ERM 2 containing all EU ‘outsider nations’. At this point he is in concordance with the aspiration of the Danish government. It has worked out a proposal for the future Danish relationship with the EMU which entails a fixed exchange rate with a very narrow band. In fact, the Minister of Economic Affairs has expressed a preference for a margin of exchange rate fluctuations similar to that which prevailed under the Bretton Woods system. The Bretton Woods system was set up in a world economy with very limited financial flows. The main aim was to prevent countries from pursuing a narrow minded ‘beggaryour-neighbour’ exchange rate policy. The system – which lasted for 25 years – built upon three important elements: (1) a stable core currency, (2) a relevant exchange rate adjustment mechanism and (3) limited financial capital flows. The Bretton Woods System (BWS) came to an end when all 3 pillars were undermined: (1) The US economy became overheated, (2) exchange rates were not adjusted speedily and appropriately and (3) the size of international financial capital flows had seized power from the monetary authorities (mainly in the form of euro-dollars). In Europe the BWS was replaced by EMS 1 in 1979 which lasted until 1992–3. The break-down story of EMS 1 is quite similar to the BWS story. EMS 1 worked relatively well during the 1980s due to a stable D-mark, a reasonable degree of exchange rates adjustments and still rather restricted policy towards financial flows. The EMS broke down in early 1990s because of challenges to all three pillars: (1) the German economy was destabilized by unification, (2) further adjustments of exchange rates were given up from early 1987 onwards and finally (3) due to the establishment of the ‘internal financial market’, where capital flows were given free. That led to the turbulent period of 1992–3, which ended by some countries leaving EMS, others accepting a much wider adjustment band (+/– 15 per cent) and an acceptance of the necessity of exchange rate adjustments (for example, the Spanish and Portuguese devaluations in 1995 and the 1997 revaluation of the Irish Punt). What are the prospects for and desirability of a revival of the EMS system á la the Bretton Woods System?3
Stable Core Currency As already discussed the euro will be missing the political superstructure which is needed to be considered a core currency by outsider nations.
Suitable Adjustment Mechanism A semi-automatical adjustment mechanism has to be announced. The idea of unchanged exchange rates for decades has in the past demonstrated a number of rather severe and counter productive side-effects. It may help to stabilize short-term exchange rate expectations, but it impedes (or, more correctly, it delays) the much more important adjustment of the exchange rate to different developments in the real parts of the economy. A pretended fixed exchange rate system opens for a one-way speculative bet (which Mr George Soros has so overwhelmingly demonstrated) that has to be compensated for by high(er) rates of interest than otherwise.
Jespersen
64
Free International Capital Mobility It is a matter of fact that the international financial capital flows have grown immensely and have today very little correspondance to real phenomena like foreign trade and direct investments. Furthermore, these uncontrollable financial floods constitute a persistent threat to the internal stability of any minor currency. Hence, it could easily be argued that any pure financial transaction involving more than one currency should be levied a foreign exchange tax – a so-called Tobin tax with the specific aim of reducing the exaggerated turnover in the foreign exchange market. By such an institutional change the central banks in the coming ERM2 countries could regain some of their lost authority in the guidance of the foreign exchange rate. That would improve the credibility of any formal exchange rate arrangements.
4
CONCLUDING REMARKS
Political and economic arguments are often entangled. From a political point of view a common currency might be judged of great importance for EU. In any case, that should not suppress the fact that the macroeconomic consequences of the EMU are uncertain. That makes the recommendable exchange rate relationship between the EMU and the ‘outsider nations’ a theoretically unsettled question: It is not entirely certain that the EMU will be stable … and the stability is of crucial importance for the economic advantages of participation. (Danish Economic Council, 1997, p. 134, translation mine) The historical experiences derived from different exchange rate systems point at a fundamental instability: when the weakest links of any system are stressed by events it breaks down and is replaced by another. This viewpoint of relativity might be helpful when discussing the EMU-project: the establishment of a monetary union might not be an irreversible step for the entire future. (Statens Offentlige Utredninger, 1996, p. 65, translation mine)
Notes 1.
2. 3.
This comment was prepared for the conference on ‘Macroeconomic Perspectives on the Danish Economy’, as a response to Professor Niels Thygesen’s original paper. The future stability of the EMU was debated at the conference, and since then there have been important new developments on this issue which I have partly integrated. Niels Thygesen was a member of the Committee preparing the report in 1989. This question has been intensively debated since the events of 1992–3. One rather recent and extended discussion can be found in Statens Offentliga Utredninger (1996) taking its starting point from the Swedish case. Although time has passed I still find the discussion between the Americans and Keynes (1941–3) on preparing the Bretton Woods Agreements quite illuminating for the key issues involved (cf. Jespersen, 1996).
Comment
65
References Committee for the Study of Economic and Monetary Union (1989) Report on Economic and Monetary Union in the European Community (Luxembourg). Feldstein, M. (1997) ‘The political economy of the European Economic and Monetary Union: political sources of an economic liability’, Journal of Economic Perspectives, 11(4) (Fall). Friedman, M. (1997) ‘Fra monetær enhed til politisk splittelse?’, Berlingske Tidende (21 October 1997). Jespersen, J. (1995) ‘A post-Keynesian perspective on monetary union and the European Central Bank’, Chapter 11 in P. Arestis and V. Chick, Finance, Development and Structural Change – Post-Keynesian Perspectives (London: Edward Elgar). Jespersen, J. (1996) ‘Keynes on the International Monetary System – and the lessons for today’, Working Paper, Institute of Social Sciences, Roskilde University. Økonomiske Raad (1997) Formandskabet: Dansk Økonomi, maj 1997 (Danmark og ØMu’en) (Copenhagen). Statens Offentliga Utredninger, ‘Sverige och EMU’ (EMU-utredningen, Calmforsudvalget), SOU, 1996:158 (Stockholm). Thygesen, N. (1997a) ‘The prospects for EMU by 1999 – and reflections on arrangements for the outsiders’, EPRU Working Paper, 1995–17 (Handelshøjskolen i København). Thygesen, N. (1997b) ‘EMU and the outsider nations’, paper presented to the conference on ‘Macroeconomic Perspectives on the Danish Economy’ (Hornbæk) 19–20 June. Wyplosz, C. (1997) ‘EMU: why and how it might happen’, Journal of Economic Perspectives, 11(4) (Fall).
3 Macroeconomic Perspectives on Stock and Bond Investments in Denmark since the First World War1 Steen Nielsen and Ole Risager
3.1
INTRODUCTION
The purpose of this chapter is first to characterize the return–risk characteristics of Danish stocks and bonds in the period 1920–95. On the basis of the descriptive background we analyze whether the size and development of asset returns can be explained by the Consumption–CAPM, which has become a popular asset pricing model in recent years. We also briefly discuss whether the existing upper limits on Danish pension funds’ stock investments are reasonable in view of our results on the return–risk characteristics of Danish stocks and bonds. Throughout the chapter, we focus on the economics of the stock market rather than on technicalities. Section 3.2 begins by calculating the return on a 1-year, 5-year and 10-year investment in the market portfolio of stocks, and the associated risk as measured by the standard deviation of the portfolios. Subsequently, the chapter outlines the corresponding government bond yields and risks. Next follows a comparison of stock and bond returns – that is, a characterization of the equity premium and its development since the early 1920s. We also examine whether stock investment is more risky than bond investment, which is a commonly held view. In a Danish context, this view has been attacked by Christiansen and Lystbaek (1994). Our results show that the time horizon is the crucial factor in this issue. In simple terms, stock investments are not more risky than bond investments in the short term, whereas stock investments are not more risky than bond investments in the long term. The most important reason why stocks are as safe as bonds in the long term has to do with a strong tendency for real stock returns to meanrevert. Thus, bad years in the stock market are usually followed by good 66
Stock and Bond Investments in Denmark
67
years, whereas bond returns display positive autocorrelation. The results for Denmark are therefore similar to the findings for the United States reported in Siegel (1994). On the basis of stock and bond returns for the last two centuries, Siegel arrives at the conclusion that ‘although stocks are certainly riskier than bonds in the short run, over the long run the returns on stocks are so stable that stocks are actually safer than either government bonds or Treasury bills’. Section 3.3 goes on to examine whether the behaviour of stock returns can be explained by the Consumption–CAPM (see Breeden, 1979; Lucas, 1978). In line with the majority of the papers in the literature, this paper analyzes short-run stock returns, whereas Nielsen and Risager (1997) looks at longrun stock returns. According to the Consumption–CAPM, stocks should yield a higher return than bonds if stock returns are more correlated with consumption than bond yields, because stocks in that case provide a poorer hedge against fluctuations in consumption. The predictions of this model are consistent with the Danish data in the qualitative sense. Whether the model makes sense quantitatively is another issue, to which we return. The Consumption–CAPM (C-CAPM) is also consistent with data for the United States at the qualitative level, but the model is unable quantitatively to explain the magnitude of the United States equity premiums unless it is assumed that agents are much more risk averse than is commonly believed. That was first demonstrated by Mehra and Prescott (1985) (see also the recent survey of the so-called ‘equity premium puzzle’ literature by Kocherlakota, 1996). Tests of the Consumption–CAPM on Danish data are scarce. As far as we know, Lund and Engsted (1996) is the first paper that investigated this issue. They use the VAR technique developed by Campbell and Shiller (1988) to analyze the behaviour of short-run stock returns, and to estimate the underlying parameters (e.g. the degree of risk aversion). Their estimate of the risk aversion parameter ‘turn out to be of the wrong sign, but with large standard errors, so that the hypothesis of risk neutrality cannot be rejected’. In this chapter, we apply the non-parametric approach due to Hansen and Jagannathan (1991). This gives insight into the likely degree of risk aversion. Our results point attention to a degree of risk aversion that seems reasonable in the short end of the market (see p. 83). In spite of that, it would be premature to conclude that the theory can explain market returns. We therefore proceed to examine another aspect of the model. Thus, in the spirit of the influential paper by Grossman and Shiller (1981), we compare the actual stock market index with the index that would apply if agents had perfect forsight and behaved in accordance with the Consumption– CAPM using the information we have on the likely degree of risk aversion. Despite the fact that this test is informal and based on the assumption of
68
Nielsen and Risager
rational expectations, it gives insight into the model’s ability to explain the level and volatility of stock prices. Section 3.4 discusses the upper limits on Danish pension funds’ investments in the stock market, whereas Section 3.5 briefly summarizes the most important conclusions and implications that can be drawn from the chapter.
3.2 DATA, RETURN DEFINITIONS AND VIEW OF THE LANDSCAPE The stock market data are from two sources. Dividend yields are from our own sample of the listed firms on Copenhagen Stock Exchange, which covers about 70 per cent of the total market capitalization, corresponding to a total of about 100 firms. The dividend yield on the market portfolio is estimated as a weighted average of the dividend yield on each share, where the weights equal the value of each stock relative to the total market capitalization of the the firms in the sample. Capital gains are calculated on the basis of the market index published by Danmarks Statistik. The stock returns presented below therefore refer to the market portfolio. Let us now introduce a few return definitions. The 1-year nominal gross return on stocks S1 is defined in (3.1), and equals the dividend yield plus the capital gain (see list of notation below). This calculation disregards the possibility that dividends are reinvested within the year they are paid out.2 The corresponding 1-year real return SR1 seen from an investor’s point of view is given in (3.2), and is only approximately equal to the more common but less exact definition of the real return given as the nominal return less the rate of CPI inflation.3 The 5-year nominal (real) return equals the geometric average of the consecutive annual nominal (real) returns. The formula for the 5-year nominal return S5 is thus given by (3.3). The formula for the 5-year real return SR5 is defined analogously and is therefore omitted. S 1(t) = D(t) / Q(t − 1) + (Q(t) − Q(t − 1)) / Q(t − 1) 1 + SR 1(t) = (1 + S 1(t))C(t − 1) / C(t) (1 + S 5(t))5 = (1 + S 1(t))K(1 + S(t + 4))
(3.1) (3.2) (3.3)
In the above formulas, D(t) denotes dividends from time t – 1 to t, Q(t) and C(t) are the stock market index and the CPI at time t, respectively. Stock returns are compared to 5- and 10-year (annualized) government bond yields. In the absence of any publicly available 5-year bond yield (and 10-year yield) we have had to construct our own series using the available information on payments (amortization) streams. Because the maturity structure on the outstanding Government debt is narrow in particular in the
Stock and Bond Investments in Denmark
69
beginning of the sample, the 5-year horizon is only approximate in the early period of the sample. The 10-year yield to maturity B10 is also approximate in the early years. From 1960 and onwards we link our 10-year bond yield with OECD’s series, see OECD (1996). The 5-year real bond yield BR5 is proxied by, (1 + BR 5(t))5 = (1 + B 5(t))5 (C(t − 1) / C(t + 4))
(3.4)
where B5(t) is the 5-year annualized nominal bond yield. The 10-year real bond yield BR10 is defined analogously. Short-term Stock Returns, Risk and Wealth Effects The movement of the annual nominal stock market return is illustrated in Figure 3.1. The figure shows that the annual nominal return has fluctuated in a relatively stable manner around a constant mean in the period 1922–82 with the high yield in 1972 as a clear outlier. In the period 1983–95 both the mean return and the variance of stock returns have increased sub-stantially. These observations are confirmed by simple summary statistics listed in Table 3.1. The average nominal yield has thus increased from 9.6 per cent in 1922–82 to 17.8 per cent in 1983–95. Figures for 1996 underscore this tendency to higher stock returns. Along with a rise in the average return, Table 3.1 also shows that the standard deviation of the annual return has more than doubled. The stock market has thus become more volatile in the short term. It is, however, interesting to note that the return–risk ratio in
Figure 3.1 1.2 1 0.8 0.6 0.4 0.2 0 – 0.2 – 0.4
Annual nominal stock returns, 1921–95
Nielsen and Risager
70 Table 3.1
1922–95 1922–82 1983–95
Average annual stock market returns, 1922–95, per cent Nominal return
Std dev.
Real return
Std dev.
11.0 9.6 17.8
22.9 17.5 40.0
6.8 5.3 13.4
21.4 16.4 37.1
Source: Own calculations.
real terms – defined as the real mean return divided by the associated standard deviation – has increased in the period 1983–95. The observation that returns were lower in the past is consistent with the sample evidence in Hansen (1974). He finds that the average annual return equals 7.6 per cent in the period 1920–74. The corresponding real returns display roughly the same behaviour as the nominal returns and the figure is therefore omitted to save space. Spectacular returns are recorded in 1972 (83 per cent) and in 1983 (104 per cent), to which we return later. The bad years are 1931 (–11 per cent), 1940 (–15 per cent), 1974 (–28 per cent), 1984 (–25 per cent), and 1992 (–24 per cent). The sharp declines in 1974 and 1984 followed immediately after spectacular bull markets, indicating that the Danish stock market may also overreact to news – that is, display excess volatility. Figure 3.2 presents the corresponding real stock market index, defined as the nominal index deflated by the CPI. The 1920s are characterized by an Figure 3.2
Real stock price index, 1921–95
Stock and Bond Investments in Denmark
71
upward trend but also with considerable declines in 1922 and 1924, which in part reflects a tough monetary policy that aimed at restoring the real value of the exchange rate. Thus, the consumer price index fell by 20 per cent from 1920–4, which was accompanied by several collapses of major banks and industrial companies (see Olsen and Hoffmeyer, 1968). The Wall Street crash in 1929 is associated with a minor fall in the Danish index in 1930. The major adjustment occurs in 1931, where the index went down by 17 per cent. The stock market recovery sets in immediately after the crash, and the stock market reaches a new peak in 1936. Thereafter, the stock market displays a remarkable trendwise decline until the beginning of the 1980s. The only major interruption to this decline is 1972, which is the year when Denmark joined the EEC (now the EU). Along with this, the Copenhagen Stock Exchange was opened up for foreign investors in 1973, which in turn may have been anticipated by the market. The declining index contributed of course negatively to stock returns, but high dividend yields in this period kept them on a positive scale. Following the long period with a declining index, Figure 3.2 shows a very dramatic rise in 1983. In this year, the stock market goes up by exactly 100 per cent (see below). In 1987, stock markets were worldwide characterized by steep declines, and the Danish index is no exception as the real stock price falls by 9 per cent, but that is immediately followed by a 43 per cent increase in 1988. The first five years in the 1990s are characterized by temporary ups and downs along a fairly constant mean. However, since 1994 the index has been on an upward trend and this upward movement persisted through the first three quarters of 1997. The sharp increase in the index in 1983 is the largest jump that has occurred since the First World War. It is common to attribute the 1983 jump to three factors. First, there is a shift in economic policy in late 1982 towards a non-accommodation strategy with tight fiscal policy and fixed exchange rate policy as key elements (see Andersen and Risager, 1988). This change in economic policy was accompanied by a fall in the long interest rate by around 7 percentage points (see p. 73). Second, a new tax on institutional investors’ real bond yields was passed by Parliament in 1983 to become effective in 1984. Because stocks are not subject to this tax, the current and anticipated after tax return to equity increased sharply relative to bond investments. Third, capital market liberalizations meant that all restrictions on Danish investors foreign equity placements were lifted in 1984, whereas the Danish stock market had been open to foreigners since 1973 (see Eskesen et al., 1984). The liberalizations may have enlarged the window to the rest of the World, and because many markets experienced very high returns in these years this may have had spillover effects to the Copenhagen Stock Exchange. Note, however, that there has not been any formal attempt to test and quantify the various explanations.
72
Nielsen and Risager
The stock market jump in 1983 led to a considerable wealth effect despite the fact that the Danish stock market is relatively small by international comparison, and in particular by comparison to countries with an Anglo– Saxon financing mode. As the total capitalization–GDP ratio in 1983 was about 20 per cent, the boom was associated with a wealth increase of about 20 per cent of GDP. Measured in current Danish Kroner that amounts to about 200 billion kr. That is obviously a substantial wealth increase which, however, is partly reversed in 1984. Note also that the windfall gain is ‘gross’. However, as capital gains are tax free in this period for minority shareholders, provided the holding period is at least three years, the net-wealth increase was substantial. It is an important future research topic to find out whether the sharp rise in stock prices fuelled the subsequent rise in consumption and investment (as suggested by Tobin’s q model), for example, or whether it just mirrored the considerable rise in consumption and sales that occurred in the period 1984–6 along with the substantial fall in interest rates in 1983. Whether the market mainly acts as a leading indicator or actually also has important spillover effects to the real economy is discussed in Poterba and Samwick (1995), for example, on the basis of data for the United States. Long-term Stock and Bond Investment: Return–Risk Characteristics Because stock investments often are (or should be) based on long-term considerations, this section focuses mainly on the return–risk characteristics of stocks relative to bonds in a long-term perspective, that is, over a 5- and 10-year horizon. Our 5- and 10-year bond yields are recorded in Figure 3.3. This diagram shows that there has been very little return difference between 5- and 10-year bond investments. Moreover, bond yields appear to be highly positively correlated over time. In contrast to the Stock Market where bad(good) years quickly are followed by good(bad) years (see Figure 3.1), rising (falling) bond returns often persist over many years even through decades. In order to compare bond and stock investment, which indeed is a complex issue, we need to make some assumptions about investor behaviour. Thus, our bond investor is assumed repeatedly to invest in government bonds. Suppose the investor chooses bonds with maturity equal to the investor’s predetermined holding period. The return on a single investment is given by the yield to maturity on the bond or by the so-called ‘effective rate of interest’. The average return over the period 1922–95 is the arithmetic average of all the effective interest rates that have applied in this period. As the investor is uncertain with respect to future interest rates, such an investment strategy is risky even though the investor knows his
Stock and Bond Investments in Denmark Figure 3.3
73
Nominal bond yields, 1921–95
0.25
0.2
Yield
0.15
0.1
0.05
0
nominal return for certain as soon as he has made his investment. If the investment was a one-shot event all nominal uncertainty of course disappears. The sort of uncertainty that stems from inflation will still be there given that the bond is a nominal claim. In sum, our perspective is an investor who repeatedly invests in nominal government bonds and hence is subject to both interest rate and inflation uncertainty. Table 3.2 shows the average bond yield over a 5- and 10-year investment horizon as well as the average 5- and 10-year stock returns (cf. (3.3)). Table 3.2 shows that the average real return on a 5- and 10-year bond investment strategy equals 3.6 per cent and 3.3 per cent, respectively. The corresponding real stock returns are 5.1 per cent and 4.7 per cent, respectively. Hence, stocks yield on average a higher return than bonds. Table 3.2
Average 5- and 10-year bond and stock returns, 1922–95, per cent Nominal return
Std dev.
Real return
Std dev.
8.0 9.5 8.3 9.6
4.2 7.0 4.6 4.5
3.6 5.1 3.3 4.7
4.2 6.3 4.4 3.6
5-year bond 5-year stock 10-year bond 10-year stock Source: Own calculations.
Nielsen and Risager
74
The standard deviations of the 5- and 10-year stock returns are also presented in Table 3.2. The standard deviations decline as the investment horizon increases. As the return on long-term investments is basically an ‘average’ of of the 1-year returns, this result may simply reflect that the variance of an ‘average’ return declines as the number of observations increase. This measure of risk need therefore not be very informative as regards the riskiness of the portfolio – that is, as regards the variance of the Dollar/Kroner value of the investment (see also Bodie, Kane and Marcus, 1993). Let us therefore also look at the riskiness of the portfolio value of different investment strategies. Suppose an investor has a holding period of one year and that he invests 1 kr in stocks in late 1921 which gives him a certain portfolio value ultimo 1922. He repeats this 1 kr investment every year until 1995. By taking the (natural) logarithm of all these portfolio values and by subsequently calculating the average value and the standard deviation of these (logged) portfolio values we arrive at the first row in Table 3.3, which also distinguishes between the nominal and the real portfolio values. Suppose the holding period is five years. The investment in 1921 accumulates therefore to a certain amount in 1926. At the end of 1926, a new investment is undertaken and associated with this a new portfolio value is recorded in 1931 and so forth. Out of this sequence we calculate the average portfolio value in logs as well as the standard deviation, see
Table 3.3
Return–risk of stock and bond portfolios, 1922–95
Average (logarithmic) portfolio value (nominal)
Std dev.
Average (logarithmic) portfolio value (real)
Std dev.
Stocks 1-year 5-year1 10-year2
0.0872 0.4462 0.9052
0.1814 0.3104 0.4249
0.0486 0.2382 0.4481
0.1786 0.3029 0.3565
Bonds 5-year1 10-year2
0.3818 0.7678
0.2013 0.4671
0.1738 0.3107
0.2075 0.4361
Notes: 1. We have calculated average portfolio values and standard deviations for non-overlapping investments starting in 1921, 1922, 1923, 1924 and 1925. Figures in the table are averages of these numbers. 2. Same procedure as for 5-year investments. Starting years are 1921, … , 1930.
Stock and Bond Investments in Denmark
75
the second row in Table 3.3. Following the same procedure, Table 3.3 also records the relevant statistics associated with a 10-year non-overlapping investment strategy. The results show that a 10-year investment is more risky than a 1-year investment in the sense that the standard deviation of the portfolio value is higher. By calculating the variances it is easy to see that the variance of a 10-year investment is not 10 times as high as the variance of a 1-year investment. If the returns had been independently distributed over time, the variance would have been proportional to the investment horizon as shown below. Thus, suppose that the log of 1 plus the return ln(1 + RR1) is independent and normal distributed with mean and variance 2. In this case, the portfolio value of a 1 kr investment over T periods equals PV = (1 + RR1(1)) (1+RR1)(2) ... (1 + RR1(T)), whereas the log value equals ln(PV) = ln(1 + RR1(1)) + ln(1 + RR1)(2)) + ... + ln(1 + RR1(T)). Hence, the mean portfolio value equals E(ln(PV)) = T, and the variance var(ln(PV)) = T2, which is linear in time. The fact that the portfolio variance in the Danish case increases less than proportionally with the investment horizon has profound implications for the optimal portfolio strategy (see below). Consider next a non-overlapping 5-year bond strategy. By assuming that 1 kr is invested in 1921, 1926, 1931, etc. we may calculate the average portfolio value in logs and the standard deviation (see Table 3.3). The same statistics are reported for a 10-year investment strategy. By comparing the stock and bond strategy two results stand out. First, stock investments yield on average a higher return than bond investments. Second, a 10-year stock investment strategy has been less risky than a 10year bond investment strategy.4 The portfolio strategy implication is that stocks are for the long run, whereas bonds are more for the short run. Put differently, a long-term investor should have a higher proportion of stocks in his portfolio than a short-term investor provided these investors are sufficiently risk averse. Thus, stocks should play an important role for young investors who save for retirement purposes (see also section 3.4). Why is it that stocks are more risky than bonds in the short term but not more risky in the long-term? To get a rough answer to this question we briefly examine the time-series characteristics of stock and bond returns. Consider the autoregressive properties of the annual real stock return. The estimated AR(4) process is given below ln(1 + RR 1) t = 0.070 − 0.282 ln(1 + RR 1) t−1 − 0.0901ln(1 + RR 1) t−2 − (0.026) (0.123) (0.128) 0.057 ln(1 + RR 1) t−3 − 0.083 ln(1 + RR 1) t−4 , (0.124) R2 = 0.08(1926 – 95)
(3.5)
76
Nielsen and Risager
where numbers in brackets are standard errors. Our earlier findings indicate strong evidence against the independence assumption insofar as the variance of the portfolios increases less than proportionally with the investment horizon, and this conclusion is further supported by the above regression insofar as the results indicate strong negative correlation in the returns. By gradually deleting those variables that appear to be insignificant5 we arrive at the next regression equation: ln(1 + RR 1) t = 0.064 − 0.286 ln(1 + RR 1) t−1 , (0.021) (0.113)
R2 = 0.08(1923 – 95) (3.6)
which is a simple AR(1) process. The strong negative correlation in the real returns shows that bad years are often followed by good years, and vice versa. The tendency for real returns to mean-revert can also be verified by using the so-called ‘variance ratio test’ applied in Poterba and Summers (1988). This test exploits the fact that if the market follows a random walk such that returns are independent then the return variance should be proportional to the return horizon, as noted also in the earlier discussion. To save space the variance ratio tests are not shown here, but given the earlier findings it is hardly surprising that these tests also lend strong support to the mean reversion hypothesis, and hence provide evidence against the random walk hypothesis. Our test statistics show that mean reversion is a stronger phenomenon in Denmark than in the markets examined in Poterba and Summers (1988). To get further insight into the long-term relationship between risk and return in the stock market, it is instructive to calculate the expected future portfolio value when the return exhibits negative first-order autocorrelation and the variance of the portfolio. Not surprisingly, the mean value of the portfolio grows linearly with time, whereas it can be shown that the variance is growing with a lower speed. Hence, the risk–return ratio declines as the investment horizon increases. It is due to this element of time diversification that stocks can be said to become less risky over time. Bonds behave differently, as indicated also by Figure 3.3. Without going into details, bond yields tend to exhibit positive serial correlation. This can be verified not only by examining the two bond yields presented in Figure 3.3, but also by looking at short interest rates which, unfortunately, are available for only the last 20 years or so. The different time-series properties of the two assets are the most important explanation why stocks are less risky in the long term, notwithstanding that stocks are much more risky in the short term. More on the Equity Premium We have already established that stocks on average yield a higher return than bonds. This return difference shows up in a considerable difference
Stock and Bond Investments in Denmark
77
between bond and stock portfolios, as the following example will show. Suppose an investor had put 100,000 kr into bonds in late 1921 (and subsequently reinvested the coupon and the principal when paid out) whereas another had invested the same amount in shares (and subsequently reinvested the received payments). The real value of these investments in 1995 are 1,660,000 kr and 3,640,000 kr, respectively, using the annual returns and the CPI. Thus, the real value of a stock portfolio is more than twice the value of a bond portfolio. Moreover, it is quite likely that our calculation exaggerates the attractiveness of bond investments insofar as we have used the 5-year interest rate as a proxy for the missing 1-year interest rate. The above calculation shows that stocks clearly have outperformed bonds over long historical periods. That is not the same as to say that stocks always yield higher returns than bonds. Table 3.4 summarizes the equity premium for the entire sample and for the two sub-periods 1922–82 and 1983–93. Over the whole sample the 5- and 10-year equity premium is around 1.5 per cent. In the period 1983–95 the premium is negative. Thus, although the return on stocks is high in this period as noted earlier, the yield on government bonds is even higher. The 1983–95 period is not unique (see Figure 3.4, which displays the 5-year equity premium over a long historical period). Figure 3.4 shows that the premium displays a cyclical behaviour – that is, the premium varies between positive and negative values. There are seven periods in the data, where the equity premium has been negative for more than one year. The equity premium is also negative towards the end of the sample and hence also in 1991. By using the 1996 return data, it can be shown that the equity premium for 1992 is slightly positive. Table 3.4
Equity premium for the sample Nominal
5-year investment 1922–95 5-year investment 1922–82 5-year investment 1983–95 10-year investment 1922–951 10-year investment 1922–82 10-year investment 1983–952
1.5 1.8 –2.7 1.5 1.8 –2.7
Notes: 1. The reason that the 10-year premium in this table is slightly different from the one implied by Table 3.2, p. 73, is that in the present table only bond yields with matching stock returns are included. 2. The premium calculation is based only on three 10-year yields. Source: Own calculations.
Nielsen and Risager
78 Figure 3.4
5-year nominal equity premium, 1921–90
0.2 0.15
Premium
0.1 0.05 0
– 0.05 – 0.1 – 0.15
The Danish premium is lower than the 6 per cent real premium for the United States estimated for the period 1889–1978 by Mehra and Prescott (1985). Note, however, that the premium in Mehra and Prescott (1985) is between 1-year stock returns and T-Bills. In spite of this, there is little doubt that the average historical Danish premium is below the American premium. Thus, a comparison of the 1-year Danish stock return with the 5-year Government bond yield (in the absence of 1-year bond yields) gives rise to a premium of about 3 per cent. The Danish premium is also below the premium in the United Kingdom. Engsted (1996) shows that the UK premium is close to 10 per cent in the period 1919–87. Historical stock returns in Sweden are quite similar to the Danish returns (see Frennberg and Hansson, 1992). In the absence of a well functioning Swedish bond market back in time it is harder to compare equity premia between Copenhagen and Stockholm. The above analysis has been concerned with the return–risk relationship on stocks and bonds in isolation. From a macroeconomic point of view it is of interest to know whether stocks are a poorer hedge against consumption fluctuations than bonds. If that is the case, the Consumption–CAPM asserts that stocks should yield a higher return than bonds. Below, we therefore relate the fluctuations in stock and bond returns to private consumption. Due to the space constraint, the theory testing in this chapter is entirely concerned with the short end of the market; Nielsen and Risager (1997) pits the Consumption–CAPM against the long-run returns. The sample covariance matrix between the 1-year real returns on stocks and bonds and real private consumption growth (per capita) is as shown in Table 3.5.6
Stock and Bond Investments in Denmark Table 3.5
79
Variance–covariance between annual returns, yields and consumption growth rate, 1922–95
Real annual stock return Real annual bond yield Real consumption growth rate
Real annual stock return
Real annual bond yield
0.04558 0.00037 0.00124
0.00285 0.00095
Real consumption growth rate
0.00275
Source: Own calculations.
Table 3.5 shows that stock returns in the short term covary more with consumption growth than do bond returns. At the qualitative level, the Consumption–CAPM is therefore consistent with the data. Hence, it is of interest to test the model in a more rigorous way.
3.3
ASSET RETURNS AND THE CONSUMPTION–CAPM
This section briefly outlines the consumption-based asset pricing model (CAPM), without going into detail with respect to the underpinnings of the model. However, it is important to mention in advance that the fundamental pricing equation derived below can be obtained also in a somewhat less restrictive set-up (see, Kocherlakota, 1996, for example). The most simple version of the model is based on the following three assumptions. First, individuals can be represented by a representative agent with well defined preferences. Second, asset markets are complete. Third, transaction costs are negligible. Under these assumptions, the following equation must always be fulfilled, mu(t)Q(t) = (1 + )–1E [mu (t + 1)(Q(t + 1) + D(t + 1)) | I(t)]
(3.7)
where mu is marginal utility, Q is an asset price or vector of prices, is the constant subjective rate of time preference, D is dividend, and I is the information set. The left-hand side is the increase in utility that occurs if the investor sells her asset and increases consumption at time t, whereas the right-hand side is the discounted expected loss in utility from the fall in consumption due to not having the asset and obtaining the associated payoff at t + 1. In equilibrium, the utility gain must equal the loss. If, for example, the left-hand-side exceeds the right-hand-side, the investor reduces his asset holdings in t, which reduces the price Q(t) and this will continue until there is equilibrium. By assuming mu (t) > 0 we have,
Nielsen and Risager
80
(1 + ) −1 mu( t + 1) Q( t) = E (Q( t + 1) + D( t + 1))| I ( t) mu( t) = E [ m( t + 1)(Q( t + 1) + D( t + 1))| I ( t)]
(3.8 )
where (1 + )–1mu(t + 1)/mu(t) ≡ m(t + 1) is the intertemporal marginal rate of substitution. Equation (3.8) says that the asset price in t equals the discounted expected value of the asset price and payoff in t + 1, where the discount factor is the intertemporal marginal rate of substitution m(t + 1). From the fundamental pricing equation (3.8) it is straightforward to get the following implication, Q(t + 1) + D(t + 1) | I (t) 1 = E m(t + 1) Q(t) = E[m(t + 1)(1 + R(t + 1))| I (t)]
(3.9)
where R is the return. Because E(m(t + 1)(1 + R(t + 1)) = Em(t + 1) E(1 + R(t + 1)) + cov(m(t + 1)(1 + R(t + 1)) we have,7 E(1 + R(t + 1)) =
1 − cov(m(t + 1)(1 + R(t + 1)) E(m(t + 1))
If m(t + 1) and 1 + R(t + 1) are negatively correlated (when high consumption growth goes hand in hand with a high return R, and low consumption growth goes hand in hand with a low return R), then the asset is risky in the sense that it provides a poor hedge against consumption fluctuations. The mean return ER(t + 1) must thus be relatively high, as noted earlier when we discussed stylized facts for Denmark. There are several ways to test (3.8). Below, we outline the visual Grossman and Shiller (1981) test and the more recent Hansen and Jagannathan (1991) method. Informal Tests of Stock Price Volatility à la Grossman and Shiller Do stock prices vary too much relative to the prediction of the Consumption–CAPM-model? In Grossman and Shiller (1981), this question is addressed (in an informal way) by comparing actual stock price volatility with the volatility implied by a perfect foresight version of (3.8) solved forward in time. Besides this, this method of course also allows one to compare the level of the actual index with the index under perfect foresight.
Stock and Bond Investments in Denmark
81
The starting point is to solve (3.8) forward in time by recursive substitution (see (3.10)). The perfect foresight price Q* is obtained by discarding the expectations operator (see (3.11)). Thus, we calculate the price that obtains if agents had perfect foresight with respect to the sequence of dividends D(t + 1), … , D(T) and the terminal price Q(T), using an appropriate sequence of intertemporal marginal rate of substitution (IMRS) parameters m(t + 1), … , m(T) (see below). Q(t ) =
Q * (t ) =
T − t mu(T )Q(T ) mu(t + j) D(t + j) + (1 + )−(T − t ) | I(t ) (3.10) E ∑ (1 + )− j mu(t ) mu(t ) j=1 T −t − j mu(t + j) D(t + j) + (1 + )−(T − t ) mu(T )Q(T ) ∑ (1 + ) j=1 mu(t ) mu(t )
(3.11)
The IMRS (1 + )–1mu(t + 1)/mu(t) ≡ m(t + 1) is obtained by using a specific parametrization of the utility function (cf. p. 83). Thus, with a specific utility function we can estimtate the marginal utilities using our consumption data. In the expression for IMRS, the parameter also enters. The latter is estimated by substituting the (sample) mean values of mu(t + 1)/mu(t) and 1 + R(t + 1) into (3.9). In the calculation of mu(t + 1)/mu(t), using a specific utility function, we will usually also have to take a stand on the degree of risk aversion since that affects the marginal utilities. This piece of information is obtained from the Hansen–Jagannathan method. Before we turn to that, we note that Q*= Q + U, where the (expectations) error term U must be uncorrelated with the current price, which implies that var(Q*) > var(Q). Thus, the perfect foresight price should be more volatile than the actual price. Hansen–Jagannathan Bounds for the Intertemporal Marginal Rate of Substitution According to the pricing formula (3.8), the asset price equals the expectation of the product of the payoff, Q(t + 1) + D(t + 1), and the IMRS, m(t + 1). The Hansen–Jagannathan method determines first the IMRS (mean) and its standard deviation that is consistent with asset market data – that is, consistent with (3.8). That is sometimes referred to as the ‘admissible mean’ and ‘standard deviation’ of the IMRS. The second step is to calculate the mean and standard deviation of the IMRS implied by a particular utility function for a representative agent. Next, we compare the mean and standard deviation of the IMRS implied by asset returns with the mean and standard deviation implied by the utility function in order to check whether
Nielsen and Risager
82
the particular utility/consumption-based asset pricing model falls within the admissible region. By taking unconditional expectations of the pricing formula (3.8) we get, EQ(t) = E[m(t + 1)(Q(t + 1) + D(t + 1))]
(3.12)
On the basis of (3.12) one can then derive the admissible region for the mean and standard deviation of the IMRS (see Hansen and Jagannathan, 1991). The standard deviation m is given as, −1
m ≥ (( EQ − E (Q + D) Em) ∑Q + D ( EQ − E (Q + D) Em))
1 2
(3.13 )
where ΣQ + D is the covariance matrix of asset returns. Under risk neutrality the discount factor is constant – that is, the IMRS is constant, implying that m = 0. Hence, the mean of all asset prices is strictly proportional to the mean of asset payoffs, where the factor of proportionality equals the constant Em. The standard deviation of the IMRS can therefore be thought of as the quadratic form in the deviations of the average prices from their risk neutral prices. Moreover, large deviations from risk neutrality imply large volatility of the IMRS. From (3.13) we then derive the relationship between Em and m – that is, the admissible region consistent with asset market data. Figure 3.5 illustrates such an admissible region. Figure 3.5
Bounds on moments of m and sample estimates
3.5
3 2.5
Bound/Admissible region
2
Sample estimates
1.5
1
0.5
0 0.8
0.9
1
1.1
Mean m
1.2
1.3
Stock and Bond Investments in Denmark
83
A common utility function in this area of research is the one with constant relative risk aversion (see (3.14)), which produces the IMRS given by (3.15). u(Z) = (Z 1− − 1) / (1 − ) −1
(3.14)
m(t + 1) = (1 + ) ((Zt ) / Zt−1 ))
(3.15)
where Z denotes consumption. Given data for consumption and guess values for the risk aversion coefficient and (1 + )–1, we obtain a timeseries for m(t + 1).8 We then calculate the mean and standard deviation of the IMRS consistent with the utility function. This procedure is repeated for different values of . By plotting the pair of mean and standard deviations associated with different risk aversion parameters into Figure 3.5 we may check whether the particular consumption-based asset pricing falls within the admissible region. Notice, that this gives information about the admissible values of the risk aversion parameter. Estimated Risk Aversion and Intertemporal Marginal Rate of Substitution Figure 3.5 shows the estimated bounds on combinations of the mean IMRS and its standard deviation using our asset return data (cf. (3.13)). The mean IMRS and its standard deviation consistent with the CRRA utility function is shown as the dotted line, assuming that the rate of time preference is 1 per cent (i.e. (1 + )–1 = 0.99). As the relative risk aversion parameter increases, Em first declines but increases at a later stage. The standard deviation is always increasing. For a risk aversion parameter = 5, the dotted line is in the admissible region – that is, the consumption-based asset pricing model appears to be consistent with asset returns. Figure 3.5 also shows that Em is in the interval (0.9–1.0), which seems to be the interval where most would expect to find the mean value of the IMRS. The standard deviation is high but below 0.5, suggesting that the IMRS shows a great deal of variability. In case (1 + )–1 = 0.97, the risk aversion parameter will have to equal 3.7 in order for the model to be consistent with market data. A risk aversion parameter of that size may be too high to be credible (see Table 3.6). So far, we have assumed that the utility function is time-separable, implying that individuals are not subject to any form of habit formation. In reality, individuals with a high consumption level today may have strong preferences for maintaining that level in the future. The idea that consumers tend to get spoiled is modelled by the habit persistence utility function. Formally, a high level of consumption today increases the marginal utility of consumption in the future, which therefore pulls in the direction of a high consumption level also in the future. Similarly, when income falls consumers
Nielsen and Risager
84
Table 3.6 What amount, X, would make a person with CRRA utility indifferent between participating in a gamble with equal probabilities of receiving 50 and 100, and receiving X with probability 1? X
70.711 63.246 58.566 53.991 51.858 51.209
1 3 5 10 20 30
Source: Mankiw and Zeldes (1991).
are reluctant in adjusting consumption first – that is, there is a ratchet effect. Habit persistence tends in general to increase the IMRS and its standard deviation; derivations can be obtained on request. Hence, a lower value of the risk aversion parameter is required in order for the CAPM model to be consistent with market prices. Our results show that the risk aversion parameter may be as low as 2 both in the case where (1 + )–1 = 0.99, and where (1 + )–1 = 0.97. As the degree of risk aversion seems to be at a level that does not seem to be too unrealistic, it is of interest to explore further aspects of the consumption-based asset pricing theory. Figure 3.6
Perfect foresight stock price decomposition, 1921–95
Risk averse (real) stock market price under perfect foresight Risk neutral (real) stock market price under perfect foresight Real stock market price
Stock and Bond Investments in Denmark
85
On the Level and Variability of Stock Prices since 1920 What would stock prices be if agents have perfect foresight and behave according to (3.11)? Figure 3.6 plots the actual index Q and the real perfect foresight index Q*, under the assumptions of constant relative risk aversion, (1 + )–1 = 0.99, and = 3.5, where the relative risk aversion parameter is in the neighborhood of our estimated value. The perfect foresight index traces the market index well until 1933. Thereafter, Q* starts to fall whereas Q rises until 1936. The biggest decline in Q* occurs in 1940. Thus, Q* falls from 121 in 1939 to 47 in 1940. Q only declines from 131 in 1939 to 106 in 1940. Hence, Q is more than twice as high as Q* in 1940. It is of interest to find out why Q* drops by so much in 1940. Recall that Q* equals the discounted dividends, where the discount rates are the IMRS. Because consumption falls very significantly in 1940, owing to the outbreak of the Second World War, marginal utility increases substantially. That leads to a significant reduction in the IMRS, which in part accounts for the large fall in Q*. Put differently, the outbreak of war and the sharp fall in consumption leads to a large increase in agents’ subjective real interest rate. Our calculations show that the latter is high though slightly below 20 per cent during the war, where the subjective interest rate is the internal interest rate implied by the IMRS from 1940 and onwards.9 Besides the increase in the real interest rate, there is also a fall in the dividend yield (see Figure 3.7).
Figure 3.7 0.12 0.1 0.08 0.06 0.04 0.02 0
Dividend yield, per cent, 1921–95
86
Nielsen and Risager
It is possible to show that it is the rise in the discount rate that is the most important explanation of the enormous fall in Q*. Thus, by calculating the perfect foresight path under the assumption of a constant discount rate, we get the risk neutral price line also shown in Figure 3.6. The risk neutral price line displays only a tiny fall in 1940. Hence, it is mainly the rise in the discount rate that explains the jump in Q* in 1940. Insofar as Q falls only relatively little, it is plausible that part of the divergence between Q and Q* reflects a discrepancy between the discount rate used in the market and the discount rate implied by the Consumption– CAPM – that is, a failure of the model to produce a realistic discount rate. This need not be entirely due to the theory; it may in part be due to a relatively imprecise measure of private agents’ consumption. Moreover, the discrepancy between Q and Q* may of course also reflect uncertainty and expectations errors. It is important to stress that the divergence between Q and Q* does not allow us automatically to conclude that the theory is invalid. Simulations by Kleidon (1986) show that if stock prices are non-stationary (follow a geometric random walk) and by construction are consistent with rational valuation models like ours, then it is still possible to have deviations between Q and Q* of the magnitude shown in Figure 3.6, even though the underlying data generating process is the theory model. In this context it is, however, important to recall that our previous results reject the random walk hypothesis, as does the earlier Danish study by Jennergren and Toft-Nielsen (1977). Moreover, the large divergence in 1940, which is mainly due to a large discrepancy between the market and the model discount rate, does indicate that in order to get a better understanding of the stock market we need to get a better understanding of how the market discounts the future dividends. It is interesting to note that the problem with tracing the level of the stock market does not carry over to its variability. Thus, the variances of the rate of change of Q and Q* are almost identical. This is surprising in view of the findings for the United States (see Grossman and Shiller, 1981; Kleidon, 1986). Moreover, it indicates that the market over the entire period 1922–95 has not been excessively volatile. This conclusion is underscored by the fact that the sample variance of the perfect foresight index may be severely downward biased (see Flavin, 1983). It is natural to investigate the implications of replacing the theory-based IMRS with the actual real interest rates as discount factors. Figure 3.8 shows the index when the actual real rate is used in the discounting as well as the actual index and the Consumption–CAPM index also shown in Figure 3.6. It appears that the index based on the actual real interest rate as the discount rate does a slightly better job in tracing the level of the stock market, whereas the Consumption–CAPM index is more in line with reality when it comes to explaining stock market volatility.
Stock and Bond Investments in Denmark
87
Figure 3.8 Perfect foresight stock prices under the consumption–CAPM and real interest rate discounting, 1921–95
Risk averse (real) stock market price under perfect foresight Risk stock market price under real interest rate discounting Real stock market price
3.4
POLICY IMPLICATIONS
As mentioned earlier, Danish pension funds and life insurance companies are subject to an upper limit on their stock investments. The current regulations state that only 40 per cent of these institutional investors’ pension liabilities can be held in stocks. The Wage Earners’ Fund (Lønmodtagernes Dyrtidsfond) and the Supplementary Pension Fund (ATP) are subject to even tougher regulations, as the upper limit for them is 35 per cent. The main reason for the quantitative regulation is the perception that stocks are more risky than other assets. This view is, for instance, expressed very clearly in 1981 when the Government proposed to raise the limit that applied in 1981. In the text of the law, stocks are characterized as ‘risky assets’ as opposed to bonds, and in comments attached to the proposal it is argued that: from the savers’ point of view variation in rates of return due to changes in business profitability implies a risk of receiving less return than could have been obtained elsewhere, even if the investment is viewed over a long time span. (Our translation) Another reason for the quantitative regulation may be various governments’ need to finance their budget deficits – that is, to be able to issue government bonds at a favourable price. This argument applies, of course, only under imperfect international capital mobility, and may therefore be of less
Nielsen and Risager
88
importance nowadays. A third concern is political and has to do with the fear of Fund’s socialism. As documented earlier, stocks are not more risky than bonds provided the portfolio is sufficiently diversified and provided the investment horizon is sufficiently long. Hence, the ‘risk argument’ is based on false premises. There is therefore a need to reconsider the present regulatory framework even though it is only the wage earners’ pension fund that effectively has been constrained by the upper limit. One reason why other investors (for example, insurance companies) have not hit the roof so far is that they are also constrained by certain minimum yield requirements, which distort their portfolios towards assets with low short-term risk (e.g. bonds). There are several potential advantages of liberalizing the investment regime. First, by encouraging pension funds and other investors to allocate a larger proportion of their funds to stocks, savers can obtain a higher return on their investments without incurring a larger risk. Of course, this assumes that the equity premium will apply also in the future and that pension funds are not able perfectly to match bond maturity with liabilities. Simple calculations suggest that the gains of increasing the proportion of stocks in the portfolios are substantial. Second, firms will get easier access to capital, which may encourage investment and growth. Finally, as bond markets are nowadays highly integrated internationally, a further softening of the regulatory framework may not affect governments’ ability to finance deficits. Besides the 40 per cent regulation, there is also a 20 per cent upper limit on the holdings of assets denominated in other currencies than liabilities. Given that the liabilities are in Danish Kroner, this means that these investors can only allocate 20 per cent of their funds to foreign assets. The possibility of having a sufficiently internationally diversified portfolio may therefore be severely restricted. This may lead to portfolios that are inefficiently biased towards home assets. The home-bias issue is outside the scope of this chapter, but we plan to return to this issue in future work.
3.5
CONCLUSIONS
The main achievement of this chapter has been to calculate and report the return–risk properties of Danish stocks and bonds over the historical period 1922–95 using our own database. Thus, we have reported the return–risk characteristics of 1-, 5- and 10-year stock investments, and the yield–risk characteristics of 5- and 10-year Government bond investments. In the subsequent theoretical sections, we have tested whether the behaviour of short-run stock returns can be explained by the Consumption–CAPM using
Stock and Bond Investments in Denmark
89
the non-parametric test by Hansen and Jagannathan (1991) and the informal test by Grossman and Shiller (1981). Our results show that stocks yield a higher average return than bonds and that short-run stock investments are much more risky than bond investments, whereas long-run stock investments are less risky. This chapter’s explanation of this apparently provocative and paradoxical finding, reported also in Christiansen and Lystbæk (1994) and Siegel (1994), is the mean-reversion property of real stock returns. Thus, bad years in the market are usually more than offset by the good years that follow later, whereas bond yields display positive autocorrelation. Because the existing regulatory framework for institutional investors is based on the premise that stocks are more risky than bonds, we have argued that these return–risk results call for a further softening of the quantitative regulations, if not a complete abandonment, simply because pension funds and other institutional investors are able to pursue long-run strategies with highly diversified portfolios. We have also argued that such a policy change will be close to a Pareto improvement insofar as both pensioners and firms listed on the stock market will benefit from such a liberalization. Our tests of the consumption-based asset pricing model yield mixed results. The estimate of the risk aversion parameter under constant relative risk aversion and in particular under habit formation is in the plausible range. However, as the level of the implied perfect foresight stock market index is very far from the actual index from the late 1930s and until the beginning of the 1970s, we have doubts about the validity of the underlying theory even though Kleidon (1988) has shown that this type of information may be too soft to reject the theory. However, Nielsen and Risager (1997) also reject the consumption-based view of stock prices and this is indeed also the typical finding in the international literature. It seems to us that the key issue is to get a better understanding of the discounting process insofar as dividends explain only a tiny fraction of stock market volatility. We plan to return to this issue in future work. Notes 1.
The paper on which this chapter is based has been presented at seminars at the MBA Programme and the Institute of Economics, Copenhagen Business School, and at the conference ‘Macroeconomic Perspectives on the Danish Economy’ (Hornbæk) 19–20 June 1997. We wish to thank Syed M. Ahsan, Copenhagen Business School; Tom Engsted, Aarhus Business School; Lars Lund, Copenhagen Business School; Jan Olesen, Copenhagen Business School; Bjorn Hansson, University of Lund and Paolo Pesenti, Princeton University for useful comments and suggestions. We have also benefited from discussions with Henrik W. Mogensen, Tryg-Baltica. Finally, thanks to Ian Valsted and Michael Wieman for efficient research assistance.
Nielsen and Risager
90 2. 3. 4. 5. 6. 7. 8. 9.
Frennberg and Hansson (1992) show that this is of very little importance for the calculation of Swedish stock returns over the period 1919–89. The difference between the exact and the approximate definition can be large when returns are fairly large, which is not uncommon in stock markets. Despite the importance of this result, it should be mentioned that the statistical significance is unclear (and will remain unclear for many years) as there are obviously too few 10-year periods. Real returns are only normal distributed when we omit observations for 1972 and 1983, but in this case we also get strong negative autocorrelation. We assume that the 5-year bond yield can be used as a proxy for the 1-year yield. We have used the law of iterated expectations. Data is real consumption per capita obtained from Danmarks Statistik and Hansen (1974). The internal interest rate for 1940, for example, is defined as the interest rate that discounts a future stream of income to the same amount as our IMRS from 1940–95 does.
References Andersen, T.M. and O. Risager (1988) ‘Stabilization policies, credibility, and interest rate determination in a small open economy’, European Economic Review, 32, pp. 669–79. Breeden, D. (1979) ‘An intertemporal asset pricing model with stochastic consumption and investment opportunities’, Journal of Financial Economics, 7(3), pp. 265–96. Bodie, Z., A. Kane and A.J. Marcus (1993) Investments (Boston: Irwin). Campbell, J.Y. and R.J. Shiller (1988) ‘The dividend–price ratio and expectations of future dividends and discount factors’, Review of Financial Studies, 1, pp. 195–228. Christiansen, J. and B. Lystbæk (1994) ‘Afkast og Risiko På Aktier og Obligationer 1915–1993’, Finans Invest, pp. 10–13. Deaton, A. (1992) Understanding Consumption (Oxford: Clarendon Press). Engsted, T. (1996) ‘Evaluating the consumption–capital Asset Pricing Model using Hansen–Jagannathan bounds: evidence from the UK’, Working Paper, Aarhus Business School. Eskesen, L., F.D. Jensen, E. Rygner and O. Zacchi (1984) Finansielle Institutioner og Markeder (Jurist- og Økonomforbundets Forlag). Flavin, M.A. (1983) ‘Excess volatility in the financial markets: a reassessment of the empirical evidence’, Journal of Political Economy, 91, pp. 929–56. Frennberg, P. and B. Hansson (1992) ‘Computation of a monthly index for Swedish stock returns 1919–1989’, Scandinavian Economic History Review, 1, pp. 3–27. Grossman, S.J. and R. Shiller (1981) ‘The determinants of the variability of stock market prices’, American Economic Review, 71(2), pp. 222–7. Hansen, K. (1974) Om Afkastet af Danske Aktier i tiden 1900–1974, Nordisk Fjerfabriks Jubilæimsskrift, Nordisk Fjerfabrik, pp. 191–220. Hansen, L.P. and R. Jagannathan (1991) ‘Implications of security market data for models of dynamic economies’, Journal of Political Economy, 99, pp. 225–62. Hansen, S.Å. (1974) ‘Økonomisk vækst i Danmark (Economic Growth in Denmark)’ (Copenhagen: Akademisk Forlag). Jennergren, L.P. and P. Toft-Nielsen (1977) ‘An investigation of random walks in the Danish stock market’, Nationaløkonomisk Tidsskrift, pp. 254–69.
Stock and Bond Investments in Denmark
91
Kleidon, A.W. (1988) ‘Variance bounds tests and stock price valuation models’, Journal of Political Economy, 94, pp. 953–1001. Kocherlakota, N.R. (1996) ‘The equity premium: it’s still a puzzle’, Journal of Economic Literature, pp. 42–71. Lucas, R.E., Jr. (1978) ‘Asset prices in an exchange economy’, Econometrica, 46(6), pp. 1429–45. Lund, J. and T. Engsted (1996) ‘GMM and present value tests of the C–CAPM: evidence from the Danish, German, Swedish and UK stock markets’, Journal of International Money and Finance, 15, pp. 497–521. Mankiw, N.G. and S.P. Zeldes (1991) ‘The consumption of stockholders and nonstockholders’, Journal of Financial Economics, 29, pp. 97–112. Mehra, R. and E.C. Prescott (1985) ‘The equity premium: a puzzle’, Journal of Monetary Economics, 15, pp. 145–61. Nielsen, S. and O. Risager (1997) ‘On the long run stock return puzzle in Denmark and the C–CAPM’, Working Paper, Copenhagen Business School (preliminary). OECD (1996) Statistical Compendium, CD-Rom. Olsen, E. and E. Hoffmeyer (1968) ‘Dansk Pengehistorie 1914–1960’ (Danmarks Nationalbank). Pagan, A. (1996) ‘The econometrics of financial markets’, Journal of Empirical Finance, 3, pp. 15–102. Poterba, J.M. and A.A. Samwick (1995) ‘Stock ownership patterns, stock market fluctuations, and consumption’, Brookings Papers on Economic Activity, 2, pp. 295–372. Poterba, J.M. and L.H. Summers (1988) ‘Mean reversion in stock prices – evidence and implications’, Journal of Financial Economics, 22, pp. 27–59. Siegel, J.J. (1994) Stocks for the Long Run (New York: Irwin).
Comment1 Tom Engsted
1
INTRODUCTION
The equity premium – that is, the difference between expected returns on stocks and bonds – has been the subject of a large number of empirical studies. One of the main findings in this literature is that in most countries the premium is too large to be explained by standard intertemporal equilibrium asset pricing models (the ‘equity premium puzzle’) (see Campbell, 1996 for a recent survey with multi-country empirical results). Chapter 3 by Nielsen and Risager adds to this literature by presenting new results for Denmark, and finds that the Danish equity premium appears to pose a much smaller puzzle compared to other countries. The chapter contains three main sections. (1) A description of Danish equity premia with 5- and 10-year bond and stock returns; (2) Evaluation of the consumption-based capital asset pricing model (C–CAPM) with 1-year returns; and (3) Policy implications. In my discussion, I will make some comments on each of these. The main part of my discussion will focus on (2). In particular, I will discuss the way Nielsen and Risager test and evaluate the C–CAPM, and I will compare their results with my own previous results using similar data (see Lund and Engsted, 1996). I will also express some reservations about some of the analyses and conclusions carried out and reached by Nielsen and Risager, and in particular I will argue that the chapter does not support a lifting of the current regulations on Danish pension funds.
2
THE DANISH EQUITY PREMIUM
It has long been recognized (at least since Lawrence Smith in 1924, see Siegel, 1994) that on average stocks earn a larger return than bills and bonds. The usual explanation given is that stock returns are more volatile than bond returns so the higher return is just a compensation for higher risk. However, recently Siegel (1992, 1994) showed that on US data, long-horizon (10–20 years) stock returns are not only higher than equivalent bond returns, but apparently are also less variable – that is, stocks are safer than bonds at long investment horizons. One of the main results of Section 3.2 of Nielsen and Risager’s chapter is that this also seems to hold for Denmark. Siegel (1992, 1994) and Blanchard (1993) give the following explanation for the large equity premium. The stock market crash in 1929 and subsequent depression in the 1930s caused investors to move from stocks to bonds, thus raising the equity premium on stocks. This went on till around 1950. Then the memories of the 1930s faded, investors returned to stocks, and the premium went down steadily until the beginning of the 1970s when high and unanticipated inflation caused the expected real return on bonds to be much higher than the realized return. Thus, since real stock
92
Comment
93
returns did not change much (as opposed to bonds, stocks are a claim to real payments), the observed equity premium increased again until the beginning of the 1980s where inflation rates decreased, causing the observed premium to decline again. So, the large average equity premium since the 1920s is due to the crash in 1929 in combination with inflation misperceptions in the post-Second World War period. Both Siegel and Blanchard find that the equity premium has decreased quite a lot since the beginning of the 1980s, and they both predict that in the future the premium will be much lower than in the 1930–80 period. Nielsen and Risager also find that on the Danish data the equity premium has been much lower (in fact negative) in the 1983–95 period compared to the earlier one. Thus any clearcut statements about the attractiveness of stocks compared to bonds at long investment horizons must be looked at with major reservations, in particular since the uncertainty associated with estimating long-horizon equity premia is very large. As the authors themselves note, the standard deviations reported in Table 3.2 are severely downward-biased due to the inherent smoothness (serial correlation) of long-horizon returns. In general, n-period overlapping returns will exhibit moving average (MA) properties of order n – 1. Thus, on annual data 5- and 10-year returns will be MA(4) and MA(9), respectively. The method most often used for correcting for such effects is the Newey and West (1987) estimator. By using this correction the standard deviations of 5-year real bond and stock returns change from 4.2 per cent to 10.6 per cent, and from 6.3 per cent to 11.3 per cent, respectively. The changes for 10-year returns are: bonds, from 4.4 per cent to 13.3 per cent; stocks, from 3.6 per cent to 8.7 per cent. This reflects the fact that the effective number of observations is smaller due to overlapping returns. Unfortunately, these new, much larger, estimates must also be interpreted with great care. As shown by, for example, Richardson and Stock (1989), correcting for moving-average terms in time-overlapping returns using the Newey–West estimator (or similar methods) can be very unreliable when the degree of overlap is large, which is surely the case for the 10-year returns. Instead, Nielsen and Risager suggest looking at non-overlapping 5- and 10-year returns (Table 3.3). The problem here again, of course, is that this diminishes the number of observations dramatically. In fact, there are only seven independent 10-year returns over the period 1922–95. The bottom line is that the uncertainty associated with estimating the Danish equity premium is so large that, in my opinion, one should abstain from making any recommendations about investment behaviour based on it. (Note that these concerns may not be raised to the same extent towards the analyses in Siegel, 1994, because his sample is much larger (nearly 200 years of annual US data) and his estimated equity premia are much larger than the Danish premia). Finally, as we shall see shortly, a standard deviation may not be the relevant measure of risk from the outset. In the C–CAPM, for instance, the proper measure of risk is not the standard deviation of returns per se, but rather the asset’s covariation with the intertemporal marginal rate of substitution, IMRS (that is, some function of consumption. Thus, a full understanding of risk cannot be achieved without reference to an underlying economic model. In moving to section 3.3 of the chapter, let me add that I find it peculiar that, whereas section 3.2 deals with 5- and 10-year returns, section 3.3 only deals with 1-year returns. Thereby the connections between the two sections become less transparent. It would have been obvious to include in section 3.2 a comparison of 1-year stock and bond returns, and in section 3.3 to evaluate the C–CAPM at 5and 10-year horizons also. As I have access to Christiansen and Lystbæk’s (1994) 1-year holding period return on long bonds (updated to 1995), a comparison with Nielsen and Risager’s 1-year stock return is easily made. The average 1-year equity
Engsted
94
premium is 2.8 per cent with a standard deviation of 2.0 per cent, hence the premium is not significant.
3
THE C–CAPM
The claim that stocks dominate bonds with respect to both risk and return is counterintuitive from a finance theory point of view. How can an asset on average yield both higher return and less risk than another asset? It seems to pose a major challenge to conventional finance theory. Modern finance theory, however, is much more complex than just comparing the average returns and volatilities of different assets. In the C–CAPM the returns on different assets reflect the amount of risk (measured by the asset’s covariance with the IMRS (typically proportional to consumption growth) and the price of risk (measured by the degree of risk aversion)). Thus, according to the C–CAPM the equity premium is large if stocks covary much more with consumption than do bonds, and/or people are highly risk averse. In modern finance literature, this model is surely the most prominent intertemporal equilibrium model of asset returns. One of the reasons is that it provides the link from the macro-economy to asset pricing that financial economists have always sought – that is, risk premia are explained by assets’ covariation with consumption. The famous equity premium puzzle (cf. Mehra and Prescott, 1985) is built directly on the C–CAPM. The puzzle says that over a 100-year period in the United States, the equity premium cannot be explained by the constant relative risk aversion (CRRA) – and complete markets/no transactions costs – specification of the C–CAPM unless people are extremely risk averse. The reason is that 1-year stock returns only covary slightly more with consumption growth than do T-bill returns or short-term money market returns. In the literature various responses to the puzzle have been brought forward. (1) The C–CAPM framework is fundamentally wrong; (2) The C–CAPM framework is correct, but the CRRA specification is wrong; (3) The C–CAPM/CRRA framework is correct, but markets are incomplete and there are transaction costs. (4) The C–CAPM/CRRA framework is correct and people are extremely risk averse. The problem with (1) of course is to provide a better alternative. The response in (4) is simply to deny that there is a puzzle. The problem then is that extreme risk aversion is both highly counter-intuitive and implies a very large risk-free real interest rate (that is the ‘risk-free rate puzzle’, see Weil, 1989). In my opinion, responses (2) and (3) are the most fruitful, and important new insights have been gained along these lines (see, for example, Constantinides, 1990; Campbell and Cochrane, 1995; Luttmer, 1996) for analyses of time-non-separabilities and market frictions within the C–CAPM framework. The C–CAPM is the topic for section 3.3 of the chapter. The first thing to ask in the present context is whether the equity premium also poses a puzzle in the Danish data? It appears from the results in Nielsen and Risager’s chapter that the puzzle is not nearly as severe in Denmark as in the United States (or other countries, see Campbell, 1996; Engsted, 1997), since the degree of risk aversion has only to be 5 in order for the IMRS to satisfy the Hansen–Jagannathan volatility bound (the chapter does not allow for sampling error in construction of the IMRSs and volatility bound; doing that would decrease the required value of even further (see Burnside, 1994). From Table 3.5 in the chapter it is clear that the reason is not that stocks covary much more with consumption than bonds (the two covariances are both close to 0.001). The reason is simply that the Danish equity premium is much smaller compared to the
Comment
95
United States and the United Kingdom, and that Danish consumption is quite volatile (the variance of Danish consumption growth is more than double the variance of US consumption growth (see Table 3.5 in the chapter and Table 1 in Kocherlakota, 1996). One can decrease the level of even further by allowing for habit persistence. The chapter claims that in the Danish data, and under habit persistence, can be as low as 2 without the IMRS violating the volatility bound. The overall conclusion from the Hansen–Jagannathan analysis is that one does not need extreme degrees of risk aversion in order for the mean and standard deviation of the IMRS to be inside the admissible region implied by observed Danish asset returns. Thus, apparently there is no equity premium puzzle in the Danish data. This stands in dramatic contrast to the results observed for other countries. However, there is an important difference between Chapter 3 and the related research on the equity premium puzzle – namely, that in the chapter the premium is not defined as the 1-year stock return in excess of a risk-less annualized short-term interest rate, but rather as the 1-year stock return in excess of a risky annualized 5year bond return. This choice is probably partly dictated by the lack of a proper shortterm interest rate in Denmark back to the 1920s. Although fully legitimate in evaluating the C–CAPM per se by means of the Hansen–Jagannathan bounds, it implies that the results in the chapter cannot be directly compared to the results in the related international equity premium puzzle literature. Further, regarding habit persistence, the chapter does not report the value of the habit persistence parameter, which is unfortunate since for some values of this parameter utility is not well-defined. In addition, one should note that allowing for habit persistence is in some sense an artificial way of reducing the required degree of risk aversion, since risk aversion now varies with the level of consumption relative to habit. Thus a value of equal to 2 does not necessarily imply low risk aversion (for details on these issues, see, for example, Chapter 8 in Campbell, Lo and MacKinley, 1997). Nielsen and Risager also evaluate the model by means of the Grossman–Shiller variance bounds methodology. However, it has long been known that if stock prices and dividends are non-stationary I(1), or nearly non-stationary, then the Grossman and Shiller variance bounds methodology (which builds on Shiller, 1981, is simply invalid, see, for example, Kleidon, 1986). Second, when the discount factor is close to unity, the variance of the perfect foresight price is severely downward-biased in a finite sample (see Flavin, 1983). Nielsen and Risager seem to argue that since Danish stock returns are negatively autocorrelated, stock prices must be stationary – and, hence, Kleidon’s critisism does not apply. However, negatively autocorrelated returns do not in general imply stationary prices. In fact, if dividends are non-stationary, then prices must also be non-stationary according to the C–CAPM. Lund (1992) finds Danish annual real dividends and stock prices to be I(1) over the period 1923–90, and a standard augmented Dickey–Fuller test on Nielsen and Risager’s real stock price series also clearly indicates that it is I(1). Hence, the stationarity assumption required in order to compute the Grossman–Shiller variance bounds appears to be violated in the authors’ data, thereby casting doubt on the claims concerning the variability of prices and perfect foresight prices on pp. 85–7.2 Furthermore, even if we assume stationarity, it is important to note that in general Q* will not be identical to Q when the C–CAPM is true. All one can say is that the variance of Q* should be larger than or equal to the variance of Q. This complicates the interpretation of what happens in specific subperiods during the sample (as, for example, the dramatic jump in Q* in 1940). In my opinion, if one wants to pick out and interpret such specific events, one has to use a method which, in contrast to the Grossman–Shiller method, implies equality between actual and model generated stock prices under the C–CAPM. Such methods have been developed (for example, the Campbell–Shiller method, see p. 96).
96
Engsted
For the reasons stated above, the original variance bounds methodology of Shiller and Grossman and Shiller has been dismissed and replaced by superior methodologies in the recent literature. One such methodology is the vector-autoregressive (VAR) methodology developed by Campbell and Shiller (1988). The idea here is first to obtain a linearized present value model expressed in terms of stationary variables – namely the log dividend-price ratio and first-differences of log dividends and log consumption. Next, a VAR model is estimated for these variables, and a VAR forecast of the present value of future changes in log dividends and log consumption is generated and compared with the actual log dividend–price ratio. By using this methodology on annual Danish data (1923–90) very similar to the data in Chapter 3, Lund and Engsted (1996) find that a substantial part of the variation in stock prices can be explained by rationally expected future dividends, but also that the unexplained part of the variation cannot be accounted for by consumption growth in the manner predicted by the C–CAPM with CRRA utility. Thus, our results are in some sense directly in contrast to the conclusions reached by Nielsen and Risager. Unfortunately, Nielsen and Risager make no attempt to try and reconcile these conflicting results.
3
POLICY IMPLICATIONS AND CONCLUDING REMARKS
In my discussion I have expressed some reservations about parts of the analyses carried out by Nielsen and Risager. In particular I have emphasized the ambiguity and huge uncertainty involved in estimating risks of long-horizon returns, and the problems with the Grossman–Shiller methodology. I therefore suggest that the results obtained by Nielsen and Risager are best interpreted with great care. In particular, the authors’ conclusion that long-horizon Danish stock returns are both higher and less risky than long-horizon bond returns needs to be looked at very critically. Firstly, the estimated mean long-horizon Danish equity premium is simply too small, and its associated standard deviation too large, to conclude with sufficient statistical certainty that Danish stocks yield higher returns, and are less variable, than bonds over the long term. These points are reinforced by noting that, although Danish stock returns have been extremely high in 1996 and 1997, the equity premium in the period 1983–95 has on average been negative, and there are clear indications of lower future premia internationally (see Blanchard, 1993; Siegel, 1994). Secondly, the riskiness of stocks and bonds may not be measured only by their respective standard deviations. Modern finance theory instead suggests that an asset’s risk is intimately related to the asset’s covariation with the underlying economic fundamentals. Thus, in my opinion, there is no support for the authors’ recommendation of a lifting of the current regulations on Danish pension funds (at least not from the results in the chapter). Aside from the reservations stated above, one may argue, as does Finansrådet (1997), that these regulations have actually not been particularly binding or seriously restricted the pension funds’ behaviour. In fact, in 1995 only 15.8 per cent of Danish pensions funds’ wealth was put into stocks, and although this is an increase from 7.4 per cent in 1986, it indicates that for the sector as a whole Danish pension funds actually do not place wealth in stocks up to the limit. As noted by Nielsen and Risager, one possible reason for the small amount of stocks in the Danish pension funds’ portfolios is the specific minimum yield requirements for those funds. However, one should also remember that in Denmark there is a substantial issuance of mortgage bonds to finance real estate. So, from an immunization point of view, it seems reasonable to have a substantial amount of bonds among the private sector’s
Comment
97
assets, such that in case of interest rate changes, which increase the value of private sector’s liabilities, this is offset by a similar increase in the value of its assets.
Notes 1. 2.
I thank my colleagues at the Department of Finance, The Aarhus School of Business, for useful discussions in relation to the Danish equity premium. The usual disclaimer applies. Incidentally, Nielsen and Risager compare the variances of the rate of change of prices (Q) and perfect foresight prices (Q*), although, whether or not prices and dividends are stationary, the model has no implication for such rate of changes. The model has implications only for the variance of the levels of Q and Q*, and then only under stationarity.
References Blanchard, O. (1993) ‘Movements in the equity premium’, Brookings Papers on Economic Activity, 2, pp. 75–118. Burnside, C. (1994) ‘Hansen–Jagannathan bounds as classical tests of asset pricing models’, Journal of Business and Economic Statistics, 12, pp. 57–79. Campbell, J.Y. (1996) ‘Consumption and the stock market: interpreting international evidence’, NBER Working Paper 5610 (Cambridge, Mass.: NBER). Campbell, J.Y. and J. Cochrane (1995), ‘By force of habit: a consumption based explanation of aggregate stock market behavior’, Unpublished working paper, Harvard University and University of Chicago. Campbell, J.Y. and R.J. Shiller (1988), ‘The dividend–price ratio and expectations of future dividends and discount factors’, Review of Financial Studies, 1, pp. 195–227. Campbell, J.Y., A.W. Lo and A.C. MacKinlay (1997) The Econometrics of Financial Markets (Princeton: Princeton University Press). Christiansen, J. and B. Lystbæk (1994) ‘Afkast og risiko på aktier og obligationer 1915–1993’, Finans/Invest, pp. 10–13. Constantinides, G. (1990), ‘Habit formation: a resolution of the equity premium puzzle’, Journal of Political Economy, 98, pp. 519–43. Engsted, T. (1997) ‘Evaluating the Consumption–Capital Asset Pricing Model using Hansen–Jagannathan bounds: evidence from the UK’, Working paper, Aarhus School of Business. Finansrådet (1997) Privates opsparing i Danmark, Danske Pengeinstitutters Forening. Flavin, M. (1983) ‘Excess volatility in the financial markets: a reassessment of the empirical evidence’, Journal of Political Economy, 91, pp. 929–56. Kleidon, A. (1986) ‘Variance bounds tests and stock price valuation models’, Journal of Political Economy, 94, pp. 953–1001. Kocherlakota, N. (1996) ‘The equity premium: it’s still a puzzle’, Journal of Economic Literature, 34, pp. 42–71. Lund, J. (1992) ‘Rationelle bobler i de danske aktiekurser 1923–1991: en empirisk analyse’, Nationaløkonomisk Tidsskrift, 130, pp. 483–97. Lund, J. and T. Engsted (1996) ‘GMM and present value tests of the C–CAPM: evidence from Danish, German, Swedish and UK stock markets’, Journal of International Money and Finance, 15, pp. 497–521.
98
Engsted
Luttmer, E. (1996) ‘Asset pricing in economies with frictions’, Econometrica, 64, pp. 1439–68. Mehra, R. and E. Prescott (1985) ‘The equity premium: a puzzle?’, Journal of Monetary Economics, 15, pp. 145–61. Newey, W. and K.D. West (1987) ‘A simple, positive semi-definite, heteroscedasticity and autocorrelation consistent covariance matrix’, Econometrica, 55, pp. 703–8. Richardson, M. and J. Stock (1989) ‘Drawing inferences from statistics based on multi-year asset returns’, Journal of Financial Economics, 25, pp. 323–48. Shiller, J.R. (1981) ‘Do stock prices move too much to be justified by subsequent changes in dividends?’, American Economic Review, 71, pp. 421–36. Siegel, J. (1992) ‘The equity premium: stock and bond returns since 1802’, Financial Analysts’ Journal (January–February), pp. 28–46. Siegel, J. (1994) Stocks For the Long Run (New York: Irwin). Weil, P. (1989) ‘The equity premium puzzle and the risk-free rate puzzle’, Journal of Monetary Economics, 24, pp. 401–21.
Comment Björn Hansson
1
INTRODUCTION
Nielsen and Risager have in Chapter 3 written a very interesting study of long-run relations on the Danish stock market and fixed income market. My comments will concentrate on their results on the return and risk characteristics of the Danish assets and I will compare their results for the Danish markets with results from my own studies of the Swedish markets. I will also discuss the contentious result that investing in a well diversified portfolio of stocks is less risky than investing in bonds. Thus, an investor with a long time horizon – for example, saving for old age – should tilt her portfolio weights towards stocks. This recommendation is not unusual among practitioners. In fact, it is almost the conventional wisdom that investors with a long horizon should invest more heavily in stocks than investors with a short horizon, since the riskiness of stocks diminishes with the length of the horizon (see Lee, 1990; Thorley, 1995). However, it is also much criticized by well known economists like Paul Samuelson and Robert Merton (see Samuelson, 1994; Bodie, Merton and Samuelson, 1992; Jagannathan and Kocherlakota, 1996). The tentative conclusions from this debate are as follows: the recommendation is correct if stock returns show meanreversion, which means that the returns tend to revert to the unconditional mean, or the relative risk aversion is decreasing – that is to say, with increasing wealth a greater weight should be given to the risk-bearing asset. In this debate an old issue is missing – namely, the so-called ‘estimation risk’, which implies that the estimates of the distribution of asset returns, which are used in practical applications, deviate from the true distribution, which is supposed to be known in the theoretically derived decision rules (see Bawa, Brown and Klein, 1979). This problem can be handled by simulation (see Liang, Myer and Webb, 1996).
2
A DISCUSSION OF THE RETURN CHARACTERISTICS
Figure C3.1 exhibits the development in real terms of 1 SEK and 1 DKK invested in a well diversified stock portfolio from 1922 until 1996 and where all dividends are reinvested.1 The Swedish portfolio significantly outperforms the Danish portfolio, the terminal values in December 1996 are 239 SEK and 36 DKK respectively. A closer look shows that the evolution was very similar up to the Second World War but then the Swedish stock portfolio grows much faster until the beginning of the 1970s.2 Probably, this difference just reflects the enormous Swedish advantage of being lucky enough to stay out of the war; in fact, production went on during the war and the industrial infrastructure came out of the war unscathed to take advantage of the post-war boom. The story of the oil crises in the 1970s is similar despite some differences in economic
99
100 Figure C3.1 1992–96
Hansson Cumulative real return index for Swedish and Danish stocks
policies, Sweden tried to overcome the slump of the mid 1970s by aggressive demand management, while Denmark had more unplanned budget deficits. The driving factor in both countries was probably the great unexpected inflation. The extremely expansionary path of the 1980s, partly dented in Sweden by the collapse of the property markets and some banks in 1990, are also very similar for the two stock markets. Notice that this similarity is despite opposite economic policies: Sweden devalued by almost 20 per cent in 1982 while Denmark followed a non-accommodation strategy. Table C3.1 shows that the annual real return has been higher in Sweden – 9.5 per cent and 6.8 per cent, respectively. However, the risk as represented by the standard deviation has been slightly higher in Denmark; this difference is mainly because of the last period even if the risk on the Swedish stock market has been particularly high during this period. Another way to look at risk is the number of calendar years with negative returns – in Sweden less than 30 per cent of the years have negative returns while there are as many as 43 per cent in Denmark. Looking at the markets for long bonds in Figure C3.2 the development in the two countries is very similar: the cumulative return index for Denmark is slightly better – the bond portfolio has grown by a factor of 10 and while Sweden has grown by a factor of 8.3 The development differs notably only in two periods – the 1930s and the inflationary period of the 1970s. Notice that the Danish economic policy of the early 1980s gave an enormous boost to the bond market, and thus, this policy made a difference to the bond market. A comparison with the figures from the stock market shows that the stock portfolios have done very much better than the bond portfolios. In fact, in Figure C3.2 it is possible to see that for the 30 years following 1945 the bond portfolios hardly had a constant real value.
Comment Table C3.1
101
Annual real returns, 75 observations, 1922–96 Sweden
Arith. mean (%) Std dex. (%) Median (%) Max. (%) Min. (%) P(neg.) (%) P(S
Denmark
Stocks
Bonds
Stocks
Bonds
9.5 20.4 9.2 55.7 –36.3 29.3 37.3
3.4 11.1 1.5 38.3 –37.0 38.7
6.8 21.6 4.3 103.1 –29.6 42.7 45.3
4.0 14.5 1.4 71.8 –19.3 41.3
Notes: P(neg.) shows percentage of negative returns. P(S
3
PORTFOLIO STRATEGIES
In the analysis of the portfolio composition for a long-term investor Nielsen and Risager based their conclusions on a comparison between the real return and standard deviation for investment horizons of five and 10 years, where the return for the bond portfolio is computed from the arithmetic average of all yields to maturity.
Figure C3.2 Cumulative real return index for Swedish and Danish long-term government bonds 1922–96
Hansson
102
For the same type of problem we have constructed the return for the bond portfolio as a strategy where the whole portfolio is rolled over every year and reinvested in bonds with longest possible maturity.4 Tables C3.2–C3.4 show the result for investment horizons of five, 10 and 20 years5 – the mean return is always substantially higher for the stock portfolio in both markets. However, for the 5-year horizon there was still a possibility that bonds were better for 20 per cent of the Swedish observations and 44 per cent of the Danish observations; for the 10-year horizon the probability has fallen to 17 per cent and 28 per cent, respectively. This result is mainly due to a few good years for bonds which is evident from the difference between the mean and median. The overall abysmal performance of the bond portfolio for all horizons is best illustrated by the fact that for approximately 50 per cent of the observations bonds would not even give back the real value of the invested money. To take into account the fact that our estimates are drawn from a random variable, Darenhill (1997) has simulated the returns via bootstrapping and constructed 95 per cent confidence intervals for the returns.5 The results for Denmark are presented in Figure C3.3. There is never the case that the stock portfolio completely dominates the bond portfolio, in the sense that the confidence intervals do not overlap. But for the 20 year-horizon, and almost for the 10-year horizon, the 95 per cent confidence interval for stocks is above the median of the bond strategy. Thus, there was just a probability of 5 per cent that the stock strategy would do worse than the median bond portfolio. Figure C3.3 gives also a nice illustration of the meaning of saying that risk diminishes with the length of the investment horizon: notice the shrinking of the confidence interval for longer horizons. This phenomenon might be explained by mean-reversion. Nielsen and Risager find some evidence for this in an autocorrelation study. On the Swedish market Frennberg and Hansson (1993) have obtained similar results using more powerful statistical methods.
Table C3.2
Real returns per annum for 5-year investment horizons, 71 overlapping observations, 1922–96 Sweden
Arith. mean (%) Std dev. (%) Median (%) Max. (%) Min. (%) P(neg.) (%) P(S
Denmark
Stocks
Bonds
Stocks
Bonds
8.7 19.1 6.4 26.4 –9.6 20.0 20.0
2.8 12.5 0.6 14.1 –10.0 47.1
5.9 16.4 4.5 25.7 –5.6 25.7 44.3
4.1 17.1 1.2 22.7 –8.6 40.0
Notes: P(neg.) shows percentage of negative returns. P(S
Comment Table C3.3
103
Real returns per annum for 10-year investment horizons, 66 overlapping observations, 1922–96 Sweden
Arith. mean (%) Std dev. (%) Median (%) Max. (%) Min. (%) P(neg.) (%) P(S
Denmark
Stocks
Bonds
Stocks
Bonds
8.0 18.4 6.3 23.3 –1.8 7.7 16.9
2.5 14.4 –0.04 11.4 –4.1 53.8
5.5 14.5 4.0 17.3 –0.7 13.8 27.7
3.9 19.3 0.2 15.7 –4.1 49.2
Notes: P(neg.) shows percentage of negative returns. P(S
4
CONCLUSIONS
Looking at the history of stock and bond returns in Denmark and Sweden, the crucial difference is the positive development of the Swedish stock market for almost 25 years after the end of the Second World War. This difference needs to be analyzed further.
Table C3.4
Real returns per annum for 20-year investment horizons, 56 overlapping observations, 1922–96 Sweden
Arith. mean (%) Std dev. (%) Median (%) Max. (%) Min. (%) P(neg.) (%) P(S
Denmark
Stocks
Bonds
Stocks
Bonds
7.0 11.3 6.1 13.0 1.2 0 7.3
1.1 10.6 –0.03 6.1 –2.7 50.9
4.4 9.7 3.8 8.9 0.4 0 20.0
2.2 16.1 –0.2 9.5 –2.8 58.2
Notes: P(neg.) shows percentage of negative returns. P(S
104
Hansson
Figure C3.3 95 per cent confidence interval for Danish stocks and bonds, constructed from bootstrapping of annual data 25% 20% 15% 10% 5% 0% –5% –10%
Is it just a reflection of the fact that Sweden did not take part in the war? On the other hand, although Denmark was occupied during the war there was no substantial damage done to its industrial infrastructure. Or had Sweden, by chance or prescience, an industrial composition which was better positioned to take part in the post-war boom? From a macroeconomic perspective, it would be interesting to do an in-depth study of the development during the last decades. Have the devaluation policies followed by Sweden vs. the different accommodation strategy pursued by Denmark led to any significant differences between the markets? Finally, although I would support the recommendations that the pension funds and other institutional investors should increase their share of stocks, I am certain that the problem of estimation risk or parameter uncertainty has to be taken into consideration. Furthermore, my prior beliefs are such that I find it quite unlikely that future bond portfolios would behave in such an unfavourable way as they did for a long time after 1945.
Notes 1. 2. 3.
4.
The Danish index is taken from Darenhill (1997), which it is quite similar to the figures from Nielsen and Risager; Darenhill has taken the Swedish data from Frennberg and Hansson. We have not taken into account the exchange rate. The inflationary process seems to have been quite similar in the two countries – from 1922 to 1996 the Swedish price level increased by a factor of 19.5 and the Danish by a factor of 20.5. For the Swedish case, we use a consol up until 1949 and thereafter a bond with a duration of approximately 10 years. In the Danish case, the consol is used until 1960, from 1961 until 1983 the bond with maturity in 2007 and finally a 10-year bond for the last period. In the calculation of the mean and the standard deviation we have not taken into account the fact that the data is overlapping.
Comment 5.
105
Darenhill (1997) has used 1-year data as his base for bootstrapping. Ideally one should use moving blocks in order to capture the dependency in the data (see Hansson and Persson, 1998).
References Bawa, V.S., S.J. Brown and R.W. Klein (1979) Estimation Risk and Optimal Portfolio Choice (Amsterdam: North-Holland, 1979). Bodie, Z., R.C. Merton and W.F. Samuelson (1992) ‘Labour supply flexibility and portfolio choice in a life cycle model’, Journal of Economic Dynamics and Control, 16, pp. 427–49. Darenhill, C. (1997) ‘Tidsdiversificering – en empirisk studie av investeringsförhållanden i Sverige och Danmark’, MA thesis, Department of Economics, Lund University. Frennberg, P. and B. Hansson (1993) ‘Testing the random walk hypothesis on Swedish stock prices: 1919–1990’, Journal of Banking and Finance, 17, pp. 175–91. Hansson, B. and M. Persson (1998) ‘Portfolio choice and uncertain parameters’, Department of Economics, Lund University, manuscript. Jagannathan, R. and N.R. Kocherlakota (1996) ‘Why should older people invest less in stocks than younger people’?, Federal Reserve Bank of Minneapolis Quarterly Review, 20, pp. 11–23. Lee, W.Y. (1990) ‘Diversification and time: do investment horizons matter?’, Journal of Portfolio Management, 12, pp. 21–6. Liang, Y., F.C.N. Myer and James R. Webb (1996) ‘The bootstrap efficient frontier for mixed-asset portfolios’, Real Estate Economics, 2, pp. 247–56. Samuelson, P.A. (1994) ‘The long-term case for equities and how it can be oversold’, Journal of Portfolio Management, 16, pp. 15–24. Thorley, S.R. (1995) ‘The time-diversification controversy’, Financial Analysts’ Journal, pp. 68–76.
4 Would the Flat Tax Fall Flat in Denmark?1 Søren Bo Nielsen, Niels Kleis Frederiksen and David Dreyer Lassen
4.1
INTRODUCTION
In recent years there have been recurrent debates in Denmark on both details in the Danish tax system and the structure of the system itself. The discussion in the mid-1980s led to the introduction, in 1987, of a new system with elements of so-called dual income taxation2 (see Sørensen, 1994). A new round of discussions at the beginning of the 1990s caused the Danish government to establish a committee to scrutinize the tax system, after which a tax reform was implemented in 1994. These reforms have not stemmed debate on tax policy in Denmark. The high marginal tax rates on personal labour income as well as a sense of serious inefficiency in the current system, continue to fuel calls for more radical reform of the system of taxation. It is argued that high tax rates discourage labour supply, encourage underground activities and reduce international competitiveness. Further, labour earnings are taxed differently, depending on the form in which remuneration is registered. When received as wages, remuneration is taxed in the progressive personal income tax system. However, when received in non-wage form (stocks, stock options, capital gains, dividends, fringe benefits, etc.), income is taxed at lower rates, or perhaps not at all. Related problems apply to capital income. Different components of capital income are taxed in widely differing ways. For instance, interest on domestic bank deposits is fully registered by tax authorities and taxed at the going personal capital income tax rate. Similarly, mortgage interest can be deducted at the same rate. Imputed rent on housing, however, is more lightly taxed due to a combination of a low imputed rental rate and low official values of houses.3 Positive capital income through (indirect) pension fund and insurance company ownership is not attributed to individuals, and therefore not subject to personal capital income taxation. In fact, these institutions do not pay ordinary capital income taxes, but are subject to a 106
The Flat Tax in Denmark
107
so-called ‘real interest tax’, the effective rate of which lies below that of the personal capital income tax. Furthermore, some implicit capital income is registered as part of surpluses of firms, in which case the effective capital income tax becomes a complicated function of the corporate income tax (with its depreciation and other allowances taken into account), the personal dividend tax, and the personal tax on capital gains. Finally, some capital income which emerges as the return to investment in foreign financial assets may not be taxed at all because of tax evasion. The seriousness of the problems of the Danish capital income tax system is suggested by the fact that the total revenue from capital income taxation (except revenue from the corporate tax) is actually negative. 4 Moreover, there is reason to believe that part of corporate income is not strictly capital income, but rather remuneration of labour effort, rendering the capital income tax situation even bleaker. For these reasons a fundamental reworking of the Danish income tax system may be in order. In this chapter, we consider a hypothetical and rather radical reform of the Danish tax system entailing the abolition of taxation of the normal return to capital and the introduction of a comprehensive labour income tax combined with a business cash-flow tax, at a simple flat rate. Despite the flat rate, it is possible to introduce redistribution by leaving personal allowances intact. It is difficult to assess whether this so-called ‘flat tax’ will lead to as much redistribution as the present income tax system. What is clear, though, is that tax burdens will be less random than under the current system, and that the flat tax rate will be significantly smaller than statutory rates in the existing personal income tax system. Such rate reduction may imply considerable efficiency gains,5 to be shared by all residents. Finally, the flat tax should be much easier and less costly to administer6 and to comply with, and it also represents a recognition of the fact that growth in international capital mobility is making the taxation of capital income increasingly difficult. The chapter is organized the following way: section 4.2 provides a general description of alternative ways of organizing direct taxation. Besides the conventional comprehensive or ‘global’ income tax, found in most countries, we briefly characterize the expenditure tax, the dual income tax, the flat tax, the Comprehensive Business Income Tax and the ACE proposal. Section 4.3 deals with the measurement of wage and cash-flow tax bases for Denmark, while section 4.4 describes the computation of flat tax rates and personal allowances under various assumptions. Section 4.5 discusses some model simulations illustrating the possible dynamic effects of implementing a flat tax in Denmark. Section 4.6 considers distributional issues, the main component being some computations using the Danish ‘Law Model’. Section 4.7 briefly touches on further issues associated with transition to a
108
Nielsen, Frederiksen and Lassen
flat tax and with the openness of the economy. Finally, section 4.8 contains concluding remarks.
4.2
ALTERNATIVE DIRECT TAX SYSTEMS
This section briefly reviews some alternative direct tax systems. We describe, in turn, the comprehensive income tax system, the expenditure tax, the dual income tax, the flat tax, the Comprehensive Business Income Tax and the ACE proposal. In most of the OECD area, the direct tax system approximates a ‘global’ or comprehensive income tax (GIT, for short). Under the GIT, the tax base will be all components of income earned, labour as well as capital income. The base will be net of individuals’ expenditures necessary for earning that income. Generally, the GIT can be made progressive, so that the higher the comprehensive income received by an individual, the higher will be his or her average tax rate. In order to secure progression, as large a part of income as possible will be taxed at the individual level. In reality, it will be difficult to attribute business income to individuals. It is not difficult to incorporate dividends paid out, but retained earnings in businesses are less easy to assign to individual owners of the firm in question. For this reason, the corporate income tax is used in all countries to secure taxation of all business income, whether paid out in the form of dividends or retained in the company. Usually, but not everywhere (exceptions in the OECD are New Zealand and the United States), the corporate income tax rate is lower than personal income tax rates, and the corporate tax is therefore complemented by a tax on distributions and occasionally taxes on capital gains on shares. The GIT is designed to tax not only the remuneration of labour, but also the return to saving. For many years there has been a discussion of the appropriateness of subjecting saving to taxation. Furthermore, the income tax is perceived as a tax on what people ‘put into’ the economy; many economists and lay people instead feel that people should be taxed on what they ‘take out’ of the economy – i.e. on their consumption. This has led to proposals for an expenditure tax (the English terminology for a broad-based direct consumption tax). Under an expenditure tax, the tax base for any individual in a given period would be found as the sum of assets entering the period plus the income earned in the period, minus the sum of assets at the end of the period. This difference by definition corresponds to that person’s expenditures, or consumption outlays, during the period. In the first half of the 1980s, substantial tax reforms were carried out in the United States and in the United Kingdom (and elsewhere), and later on the Nordic countries followed suit. Rather than revising the existing global– comprehensive tax system as in the two former countries, the Nordic
The Flat Tax in Denmark
109
countries opted for a switch towards a so-called Dual Income Tax (DIT). The DIT, in its purest form, combines a progressive tax on labour income with a flat tax on personal capital income, with the lowest tax rate on labour income coinciding with the personal capital income tax rate. The main reforms occurred first in Denmark (taking effect in 1987) and a few years later in the remaining three large Nordic countries. None of these has introduced the DIT in its purest form, since either there still remains some element of progression of capital income taxation, or the low labour income tax rate does not quite coincide with the personal capital income tax rate, or both. But a significant move away from a comprehensive income tax has certainly been taken in this part of the world. (For further discussion of the DIT and the Nordic Tax Systems, see Cnossen, 1997; Sørensen, 1997.) At about the same time as reform of the Danish tax system was extensively discussed in Denmark – that is, in the middle of the 1980s – Robert E. Hall and Alvin Rabushka put forth their proposal for a flat tax (FT). The FT would combine a tax on wages and other labour income with a cash-flow tax on businesses. The rate of tax would be the same for these two components, and a personal allowance graduated according to the size of the family would be conceded.7 The FT has since the first edition of their book (see Hall and Rabushka, 1985), been quite influential in the tax policy discussion in the United States. As pointed out by its proponents, one of the main attractions of the FT is that it would be exceedingly simple to collect – it would be 19 per cent (in the United States) tax on taxable income, which for both individuals and enterprises could be calculated using a tax form of only envelope-size. Obviously, tax compliance and administration would be rendered much simpler.8 Although it is difficult to find any description in the Hall–Rabushka book (in its various editions) of what exactly the tax base for the FT is, it in effect constitutes a tax on comprehensive labour income,9 with the ordinary return to saving/capital being exempted from tax. It will tax risk premia associated with capital investment, pure profits associated with inframarginal investments or natural resources and the random component in the return to capital (see Bradford, 1995). These three components of the return to capital would also be taxed under the Comprehensive Business Income Tax (CBIT) which has been suggested by a working group at the Treasury in the United States (see US Department of the Treasury, 1992). In addition to this, the CBIT would also subject the ordinary return to capital to tax (see Gentry and Hubbard, 1997). The CBIT tax is composed of a tax on wage and other labour income paid out to individuals plus a tax on business, and the tax rate could be the same for all components of the tax base. The tax on business would not allow any deduction for interest on debt, but it would allow deduction for (economic) depreciation. In this way, both equity capital and debt capital will be taxed at the level of the business, whence there will be
110
Nielsen, Frederiksen and Lassen
little need for taxing capital income at the individual level. In accordance with this, only wage income is taxed with individuals. Finally, the English Allowance for Corporate Equity (ACE) proposal (see Gammie, 1992), could be mentioned here. In a sense, the ACE is very similar to a flat tax. The ACE taxes personal income as well as a certain definition of business income. The latter is found as business revenue minus the usual expenses, including interest on debt and economic depreciation of the capital stock. On top of this, businesses will also be allowed to deduct the product of a (risk-free) nominal interest rate and the ‘shareholders’ funds’ – that is the company’s total equity capital. In this way, the ACE tax will exempt the ordinary return to capital from tax at the business level.10 These are some of the radical and consistent tax systems which have been proposed during the last couple of decades.11 There are of course variants of these; for instance, it is perfectly possible to introduce progression of labour income tax in both the FT and the CBIT. Some economists even prefer a version of the FT in which the wage tax component is indeed progressive,12 and the tax rate on cash flow in businesses corresponds to the highest tax rate on personal wage income. Below we have, however, decided to follow the basic FT rather closely, although minor deviations from the original FT proposal – necessitated by the Danish context – can be found in our procedure.
4.3
MEASURING WAGE AND CASH-FLOW BASES
In this section we measure the base for a flat tax for Denmark in 1992.13 Basically, this base will be given as the sum of wage and salary income, cash flows in firms (incorporated as well as personal), and – in accordance with the current income tax system – taxable government transfers. We first measure the total of wage and salary incomes plus taxable transfers on the basis of data from the Ministry of Taxation on personal incomes in 1992. This data stems from individuals’ tax returns. The sum of net wages, taxable transfers, and other labour income, corrected for employee pension contributions is 575.4 billion Danish Kroner. In this total, however, that part of wages which represents contributions to private pension schemes has been subtracted, and conversely, pension benefits have been added in. This we undo and hence add total pension contributions (by employers as well as employees), whereas total pension benefits are subtracted, yielding 590.3. Second, we measure the real annual cash flow of Danish firms, following the procedure described in Gordon and Slemrod (1988), see also Gordon and Nielsen (1997). The cash-flow tax we aim for is the so-called R-base tax.14 Hence, cash flow is defined as total revenue in non-financial firms
The Flat Tax in Denmark
111
minus the sum of expenditures on labour and other inputs, net investment in inventories, and expenditures on investment goods. In principle, therefore, reported taxable profits of firms have to be corrected for all deductions and all net income from financial assets. Depreciation deductions and the cost of goods taken out of inventory are replaced by deductions for new purchases of investment goods and goods added to inventories. The central source of data for this is the so-called ‘company form’ filed by most Danish firms.15 For those sectors that are covered we gross up on the basis of the forms from participating firms (applying figures on depreciation to do so). For remaining sectors, we use supplementary statistics on gross operating surpluses of firms and altogether arrive at an estimate of cash flow plus investment expenditures of 192.1. Financial sector cash flows are not included in our measure of total cash flows. On the other hand, financial sector wages are included in the wage data above (this probably represents a – sensible – deviation from the pure R-base cash-flow tax). Finally, to arrive at a measure of cash flow alone, inventory net investment of –2.2 and new investment (net of investment subsidies) of 78.9 are subtracted to yield a final measure of cash flow in Danish firms of 115.4. Adding wage income of 590.3 and firm cash flow of 115.4 results in a total base for the comprehensive labour income tax of 705.7 billion DKK. We should mention that in the existing Danish corporate tax code net production and investment subsidies are included when the company tax liability is derived. In line with this, we have decided to include these figures in our measure of cash flow and thereby in the flat tax base.16
4.4
COMPUTATION OF CASH-FLOW TAX RATES
When computing the rate of a flat tax there are a number of possible assumptions concerning (1) which taxes supplement the flat tax, and (2) the treatment of personal allowances and itemized deductions.17 In Denmark, the income tax currently consists of three parts: (a) personal income taxes, primarily on personal labour and capital income;18 (b) corporation tax; and (c) other income taxes, which are mainly levied on yields of certain pension scheme assets. The previously mentioned ‘real interest tax’ on pension funds and life insurance companies is the main component in this latter category. In the following we shall calculate the rate of a hypothetical flat tax for Denmark under the restriction that the flat tax should generate the same revenue as the existing income tax code. In doing so, we distinguish between the case where the category (c) ‘other income taxes’ is abolished (so that the revenue generated from this set of taxes is included in the revenue requirement of the flat tax), and the case where the taxes in (c) are maintained.
Nielsen, Frederiksen and Lassen
112
Having fixed the revenue requirement of the flat tax, we can now follow either of two avenues. We may calculate the flat tax rate necessary to raise the same revenue as the existing tax code under specific assumptions concerning the personal allowances and itemized deductions, or we may fix the flat tax rate at, say, 40 per cent and then calculate the level of the personal allowance necessary for the flat tax to meet the revenue requirement, given assumptions about itemized deductions. As an example consider our benchmark case (A) in Table 4.1. In this case we remove all itemized deductions
Table 4.1
Flat tax rates and personal allowances A
B
C D billion 1992 DKK
E
F
Labour income tax base Personal allowance Income deductions Taxable income Income taxes Personal income taxes Corporation taxes Other income taxes
705.7 137.8 0 567.9 250.1 225.3 13.8 11.0
705.7 137.8 39.9 528 250.1 225.3 13.8 11.0
705.7 137.8 0 567.9 239.1 225.3 13.8 0.0
705.7 137.8 26.7 541.2 250.1 225.3 13.8 11.0
705.7 137.8 26.7 541.2 239.1 225.3 13.8 0.0
Flat tax rate
0.440
0.474
0.421 0.453 billion 1992 D–K
0.462
0.442
Personal allowance with a 40% flat tax
80.5
40.6
108.1 68.2 1992 D–K
53.8
81.4
Personal allowance per capita with a 40% flat tax
17 878
9 013
11 945
18 081
24 013
705.7 137.8 39.9 528 239.1 225.3 13.8 0.0
15 148
Note: Per capita measures are per person liable to income tax. In 1992 the number of persons liable to income tax was 4,500,688. Sources on revenue statistics: Statistics Denmark, Taxes and Duties; Ministry of Taxation, Report on Tax Policy. A Excl. income deductions, abolishing ‘other income taxes’. B Incl. income deductions, abolishing ‘other income taxes’. C Excl. income deductions, keeping ‘other income taxes’. D Incl. income deductions, keeping ‘other income taxes’. E Incl. work-related deductions and alimonies, abolishing ‘other income taxes’. F Incl. work-related deductions and alimonies, keeping ‘other income taxes’. ‘Abolishing other income taxes’ means that the revenue is collected through the labour income taxation.‘Keeping other income taxes’ means that the revenue is not included in the ‘target’ revenue.
The Flat Tax in Denmark
113
and abolish the ‘other income taxes’ along with the general capital income taxes. The revenue from the current system of income taxation in 1992 was19 DKK (billion) Personal income taxes Corporation taxes Other income taxes Total income tax revenue
225.3 13.8 11.0 250.1
Maintaining the personal allowances, taxable income becomes Labour income tax base Personal allowances Taxable labour income
705.7 137.8 567.9
The necessary flat tax rate thus becomes (250.1/567.9) = 44.0 per cent. The 1992 personal allowance per capita (on state tax) was 31,500 DKK.20 If this allowance were to be raised to 40,000 DKK, say, then the amount of taxable income would fall to 705.7 – 180.0 = 525.7. The necessary tax rate now becomes 47.6 per cent.21 Instead of varying the tax rate, keeping the personal allowances fixed at a given level, one may alternatively settle for a flat tax rate of 40 per cent, say, and derive the level of the personal allowance necessary to meet the revenue requirement. To obtain the desired 40 per cent tax rate, aggregate personal allowances must equal the difference between the total flat tax base and the total income tax revenue divided by the 40 per cent tax rate. Inserting the relevant magnitudes, we obtain personal allowances of 80.5 which in per (tax liable) capita amounts to 17,878 DKK. This figure is significantly smaller than the 1992 per capita personal allowance. Flat tax rates and personal allowances computed under a number of assumptions are reported in Table 4.1. The flat tax rates are seen to lie in the interval from 42 per cent to 47.5 per cent. It is, of course, interesting to compare the computed flat tax rates with various statutory tax rates that were applied in 1992. The (marginal) tax rate on ‘low’ personal income, defined as the state income tax rate plus the average local government tax rate, was in 1992 51.3 per cent, and the corresponding tax rate on ‘high’ personal income was the ‘tax ceiling’ rate of 68 per cent. Further, the tax rates on low and high capital income were 51.3 and 57.3 per cent, respectively.22 The Corporation tax rate was at 34 per cent in 1992. It is evident that the implementation of a flat tax rate on a broadened tax base would lead to substantial cuts in the marginal tax rates on personal income – even for those that earned a ‘low’ personal income, the marginal tax would on average be reduced by some 7 percentage points. In addition to this, ordinary capital income taxation would be abolished. A direct comparison with the corporate income tax rate is more problematic, since the
114
Nielsen, Frederiksen and Lassen
corporate tax in practice is a tax on both capital income and labour income. Whereas the labour income component of corporate income (as defined by the tax code in 1992) would be more heavily taxed under the flat tax, the component corresponding to the ordinary return to capital would not be taxed at all.
4.5
DYNAMIC AND EFFICIENCY EFFECTS
In the calculations described in section 4.4 we have computed the tax rate necessary for a wage-cum-cash-flow flat tax to generate the same revenue as the existing income tax under various assumptions. However, this is a purely static calculation. Conceivably, there would be dynamic efficiency gains from abolishing capital taxation and from reducing the tax rate on labour income substantially. Distortions to labour supply, saving and capital investment would become smaller or disappear, and the resulting dynamic gains might allow a further lowering of the tax rate while still meeting the revenue requirement. Using a dynamic general equilibrium model of the Danish economy, 23 Frederiksen (1997) examines the potential efficiency gains as well as the macroeconomic implications of implementing a flat tax as described above.24 Frederiksen finds, on the basis of data for 1992, that a flat tax rate of 41.5 per cent combined with a personal allowance of 30,618 DKK could raise the same revenue as the existing tax code provided that itemized deductions are eliminated and ‘other income taxes’ are abolished. That is, the stimulus to labour supply, savings and investment generates additional revenue that in turn allows the flat tax rate to be cut by 2.5 percentage points below the rate implied by the ‘no change in behaviour’ case reported in section 4.4. This yields substantial efficiency gains, through a sharp reduction of the marginal tax rate on wage income, the elimination of capital income taxation and by securing investment neutrality.25 The gain in aggregate efficiency is found to equal the benefits from a permanent increase in private consumption of 5.8 per cent. A transition to a wage-cum-cash-flow tax would call for a number of policy decisions concerning issues such as tax compensation of transfer recipients and old capital, and the size of the individual personal allowance. 26 A thorough examination of these issues would take us far beyond the scope of the present chapter. For example, it is implied by the cash-flow tax on corporations that there is full investment expensing. This tends to generate large impact wealth losses on equities. Retaining tax depreciation on existing assets would in Frederiksen’s model require a flat tax rate of 42.5 per cent, yielding an aggregate efficiency gain equal to an increase in private consumption of 5.4 per cent. That is, some of the gain in aggregate efficiency is
The Flat Tax in Denmark
115
sacrificed in order to shield equity holders from the impact windfall loss they suffer, as the flat tax system is introduced. For a further inquiry into these matters see Frederiksen (1997).
4.6
DISTRIBUTIONAL CONSIDERATIONS
As we saw in section 4.5, a flat tax reform could lead to substantial efficiency gains. Although the efficiency gains will be available to every tax payer in the form of lower tax rates, the flat tax may limit the amount of redistribution from rich to poor in the economy. The (‘upper’) progression in marginal tax rates on wage income is eliminated; however, by preserving personal allowances, there will remain some progression in the average tax rate – although probably less than under the existing tax code. A thorough investigation of the distributional consequences of a switch from the current income tax to a flat tax is best done with the help of microdata simulation methods. A first step in this direction is the series of computations using the Danish ‘Law Model’ (LM, for short) of the Ministry of Economic Affairs reported below. The advantage of using a sample of actual tax returns to assess the distributional impact is that such data incorporates the actual extent of tax arbitrage. Before that, however, we put forth some initial observations. In Figure 4.1 we have drawn average tax rates, defined as tax paid in relation to income earned, for a single person with no children at different levels of
Figure 4.1 1992
Tax as percentage of income: single person, no children, Denmark
60.0 55.0 50.0
Existing tax code Flat tax code (1) Flat tax code (2)
45.0 40.0 35.0 30.0 66
100
133
200
% of APW income
Flat tax code (1): Personal allowance 31 500 DKK, tax rate 44.0 per cent. Flat tax code (1): Personal allowance 40 000 DKK, tax rate 47.6 per cent. Source: OECD (1995) and own calculations.
116
Nielsen, Frederiksen and Lassen
Average Production Worker (APW) income (see OECD, 1995). We compare the average tax paid on wage income under the existing tax system with average tax paid in the two flat tax schedules considered in section 4.3. It is evident that the level of average taxation has been lowered, but it is also worth noting that the progression of average taxes has been considerably reduced. The distributional consequences of less progressive labour taxation depend on the instruments of distribution – it is common to distinguish between the redistributional effects of taxes on one hand and of transfers on the other. As noted in OECD (1996) and Ministry of Finance (1996), in Denmark it is particularly transfers that affect the income distribution, while direct taxes have a comparatively limited effect due to the rather compressed pre-tax income distribution when transfers are taken into account. It is also worthwhile to note that progressive labour taxation raises relatively little revenue, primarily because of high exemptions; in 1992, the revenue from progressive wage income taxation made up only 5.8 per cent of the total income tax revenue.27 Concerning capital taxation, it is often claimed that there must be a role for capital taxation in the tax system because of its alleged positive contribution to redistribution in society. As noted above, however, the effect of direct (capital) taxation on distribution is modest,28 and the recent abolition of the wealth tax in Denmark seems to suggest that Danish politicians do not value the contribution of capital income taxes (direct or indirect) to redistribution all that highly.29 We now turn to a computation of the distributional impact of a switch to a flat tax, using LM. The computation aims at identifying the pattern of absolute and relative changes in individuals’ disposable income following the tax switch, assuming unchanged behaviour. Thus we do not take into account long-run effects, such as those reported in section 4.5, nor the effects on the economy in the short run, primarly through the change in asset prices implied by the tax switch.30 In particular, we are interested in identifying in which income categories there is a disproportionate number of winners (losers). LM basically is a huge database which comprises information about a sample of individuals drawn from various data registers. Since the inception of LM, one of its major applications has been the analysis of tax policy changes as they affect types of individuals. The pivot in the model is the Central Personal Register (CPR) number, which acts as a key to information about income components, taxes paid, government transfers received, etc. LM for 1992 contains a sample from the entire population of about 14,000 people or some 0.3 per cent of the population. The information on income in LM mainly stems from individuals’ tax returns. Supplementary data is available as to various government transfers
The Flat Tax in Denmark
117
received, and from the Department of Duties and Taxes (Told og Shat) the taxes paid by individuals are retrieved. The basic tax unit in our computations is an individual; couples thus count as two individuals, with half the total income of the couple and half the total taxes paid by the couple. To produce an easily understandable measure of how individuals will be affected by a switch to a flat tax we need (1) an income concept for purposes of grouping individuals; (2) information on the taxes paid by the individuals, directly to tax authorities as well as indirectly (that is via the companies in which they hold shares), under the current income tax system; and (3) the taxes paid, directly and indirectly, under the rival flat tax system. The dispersion of the effects on disposable income will depend significantly on which income concept to group individuals by – as discussed below we focus on two income concepts: The flat tax income – that is, the individualized taxable income under the flat tax system – and a gross income concept. The definition of flat tax income will become clear when we compute the individualized tax under the flat tax below. Our gross income concept is found as the sum of (I) pre-tax non-business income; (II) pre-tax income from personal/non-incorporated firms; and (III) pre-tax income from individuals’ ownership of incorporated firms. Pre-tax non-business income is derived from individuals’ tax returns as follows.31 It equals (A) Personal income (wages, pensions, B-income, fees, alimony received, etc.), minus (B) Deductions in personal income (contributions to private pension plans, etc.), plus (C) Capital income (interest income, distributions from unit trusts/mutual funds, surplus from Danish property, etc.), minus (D) Deductions in capital income (interest expenses), minus (E) Other deductions (only alimony paid), plus (F) Foreign income (foreign wages, capital income, etc.), plus (G) Other income (a few special income components). Pre-tax income from personal/unincorporated firms is likewise computed from information in tax returns as, Surplus of firm/rental property + Interest income in firm + Other capital income in firm + Distribution from unit trusts, etc. – Deficit in firm/rental property – Membership fees for firm
118 – –
Nielsen, Frederiksen and Lassen Interest expenses in firm Deduction for spouse working in firm.
Pre-tax income from individuals’ ownership in incorporated firms is much more difficult to compute. ‘Pre-tax’ here means business income before corporate income tax and before dividend (and, perhaps, capital gains) tax. There is no immediately satisfactory way of deriving these figures. For the majority of tax payers in 1992 – that is, all those earning dividend income below 30,600 Danish Kroner – there even is no direct way to infer their after-tax dividend income from incorporated businesses. This is because, starting in 1991, with dividend taxes for these individuals withheld in firms, no further taxation of dividends occurred at the individual level. Hence, for these people there was no recording of dividends in their personal tax returns. A small minority – some 0.2 per cent – of tax payers, with dividend income in excess of 30,600 Danish Kroner, were liable to taxation of dividends according to a higher tax rate (45 per cent in lieu of 30 per cent) on the part exceeding 30,600 Danish Kroner; these individuals were required to submit their total dividend income in their tax returns. To compensate for the absence of direct information on individuals’ pretax income from incorporated firms we do the following: First, we identify total (pre-tax) earnings in incorporated firms. Second, we estimate that share of assets in these firms which are directly held by the household sector. Third, we retrieve that share of total household sector assets which is held by each individual. The individual’s pre-tax income from incorporated firms then is found as the product of these figures – that is, as total earnings times the household sector share times the individual share. Total earnings are taken to be the income subject to corporate income tax. The individual’s share in household sector assets is derived as the individual’s holdings of shares in (a) main shareholder companies; (b) companies traded on the stock exchange; and (c) other incorporated firms, divided by the total household holdings of these same types of companies. The LM contains information on these asset holdings, as submitted by individuals themselves. It is quite possible that there is a tendency to seriously underreport the value of especially holdings of type (a) and (c) shares. However, if this is a general tendency on the part of households, there is less need to worry, as we are interested only in a given individual’s share divided by the total. This leaves the problem of finding a measure of the ratio of shares in company types (a)–(c) which is directly held by households. We have little clue as to this number, but we estimate that some 35 per cent of the holdings of shares in Danish companies are represented quite directly by individuals.32 Performing the multiplication operation described above, we now have the sum of pre-tax non-business income, pre-tax income from personal
The Flat Tax in Denmark
119
business and pre-tax income from incorporated business, that is our gross income concept. Taxes paid under the current system are the sum of personal taxes, directly available in the LM, and a calculation of corporate income and dividend taxes applicable to incorporated firms, as assigned to individuals. Given total corporate income and dividend tax revenue, this is allocated to the household sector and other investors according to the key just given (35:65), and within the household sector in accordance with the relative ownership of each individual. Taxes on income attributable to individuals under a flat tax are computed next. The computation is in three steps: first, taxation of wage income is described, then the cash-flow tax on personal/unincorporated business follows, and finally we consider the cash-flow tax on incorporated business. We apply the benchmark rate of 44 per cent as the average flat tax rate, but maintain the existing variation in local tax rates. The tax base for the personal labour (wage) income tax is computed as: Wages, salaries + pensions, unemployment benefits, etc. + B-income + fees + alimony + certain windfall income + special income – alimony paid – contributions to private pension plans.33 From this – or rather the sum of this base and the base for cash-flow tax on personal firms – we deduct the personal allowance (in 1992 31,500 Danish Kroner) before applying the flat tax rate to the sum. The cash-flow tax base for personal firms is found as Surplus in these firms (as computed in conventional manner) + Interest expenses – Interest earnings + Inventory depreciation – Inventory investment + Depreciation – Investments. For personal firms, there is no separate information on inventory investment or inventory depreciation. We therefore simply allocate a guesstimated fraction of total inventory investment and investment depreciation to
120
Nielsen, Frederiksen and Lassen
personal firms. On the basis of data in Gordon and Nielsen (1996) we can estimate total depreciation in personal firms, while National Account statistics for institutional sectors can be used to derive an estimate of total investments undertaken by personal firms. The difference between depreciation (including inventory depreciation) and investments (including net investments in inventories) for all personal firms now has to be allocated to individual firms (tax payers); the only available key to do so is net interest expenditure in the various personal firms. To find the difference for any given tax payer, we thus multiply the total difference with this individual’s share of total net interest expenses in personal firms. This takes care of cash flow in personal firms. For incorporated firms, we proceed in the following manner. Total cash flow in these firms is computed in exactly the same way as for personal firms. From tax statistics we have the surplus in corporate firms. To this we add net interest payments, depreciation and inventory depreciation, while we deduct inventory net investment and investment (net of investment subsidies). The end result is an estimate of total cash flows in corporate firms. It must be emphasized that whereas we regard the computation in section 4.3 of total cash flows in all firms in Denmark of 115.4 billion Danish Kroner as reasonably correct, the exact division of this figure between personal firms (as reported on individuals’ tax returns) and corporate firms (with their own tax returns) is less well determined. In particular, we do not have much information on the division of net interest payments, depreciation, investment and investment subsidies between personal and corporate firms. However, we can check the above estimate for cash flows in corporate firms by aggregating cash flows for personal firms, via LM, and comparing the sum of the resulting figure and cash flows in corporate firms, etc. with our earlier figure for total cash flows of 115.4 billion Danish Kroner. By now, individuals’ gross income, current taxes, and taxes under the rival flat tax have been computed. In assessing the distributional consequences of switching to a flat tax system, we focus on the absolute change in an (adult) individual’s disposable income (as previously noted, couples count as two individuals). These changes are then first categorized by the flat tax income of the individuals (that is the sum of wages, etc., cash flow in their personal firms and cash flow in corporate firms to the extent it can be assigned to them). The results are presented in Table 4.2. The general impression one gets from looking at the table is the very wide dispersion of the changes in disposable income. Even in the lower income categories there are many examples of gains or losses in excess of 20,000 or 50,000 Danish Kroner. Although over half of the tax payers will see their disposable income rise by more than 1000 Danish Kroner, quite a few low-income individuals will experience a reduction of their disposable income. Especially at the flat tax income levels from 100,000–250,000 Danish Kroner many individuals will
Table 4.2
Changes in disposable income, by flat-tax income groups Flat-tax income (1000 kr.)
Change in disposable income
< 50
50–100 100–150 150–200 200–250 250–300 300–400 400–500 500–750 750–1 m
(1000 kr.) + > 50 1 510 3 322 + 20–50 5 134 20 536 + 10–20 3 624 45 300 + 1–10 78 822 493 770 +/– 1 274 820 375 688 – 1–10 28 388 100 868 – 10–20 302 28 388 – 20–50 302 6 946 – > 50 0 0 Total 392 902 1 074 818 Average change in disposable income (kr.) 1 306 1 985 Average tax rate before tax reform (%) 13.7 19.1 after tax reform (%) 7.9 16.2
3 322 15 704 34 730 419 176 52 246 123 820 49 528 51 038 906 750 470
3 624 18 120 93 016 323 744 39 260 151 000 85 768 64 930 13 892 793 354
2 416 25 066 128 048 134 692 28 690 97 848 56 172 44 696 14 798 532 426
3 926 59 192 56 172 47 716 7 852 27 784 19 026 14 798 11 778 248 244
7 248 74 594 31 408 16 912 3 926 12 080 9 966 9 060 13 590 178 784
14 798 18 120 4 832 2 114 906 2 718 2 114 4 228 11 174 61 004
17 516 6 946 1 812 1 510 302 1 510 1 208 2 114 10 872 43 790
201
–982
–488
5 775
10 328
5 053
2 797
31.6
34.9
37.0
40.3
42.2
40.9
40.6
44.9
32.6
31.5
35.4
37.3
38.2
39.2
39.8
40.2
41.3
44.7
Total
3 322 66 440 302 244 620 0 398 942 302 1 518 758 0 783 690 0 546 318 0 252 774 604 199 018 5 436 87 278 9 966 4 097 838
30 675 –273 493
760
121
Source: Computations from the Danish Law Model.
5 436 906 0 0 0 302 302 302 4 832 12 080
>1 m
122
Nielsen, Frederiksen and Lassen
experience quite substantial losses, primarily due to the abolishment of deductibility of interest on mortgages and study loans, etc. On average, the general picture is that the disposable income will increase for those with low and high incomes while it will fall for those with middle incomes.34 From Table 4.2 we also see that the average tax rates of the very low income groups will decrease as compared to before the tax reform. However, it is worth noting that the average tax rates for most of the remaining income groups are not affected all that much, while their marginal tax rates are lowered substantially. The highest income group, though, will experience a significant increase in their average tax rate as a result of the increased (imputed) taxation of corporate income. The fact that the average tax rate of the top income group exceeds 44 per cent is partly due to the inclusion of wealth taxes in the distributional analysis and partly due to the variation in local government tax rates. It should be noted that, when grouping individuals by their flat tax income, the flat tax system in fact implies monotonically increasing average tax rates, whereas this is not guaranteed in the existing tax system. It can be debated which income concept to group individuals by. On one hand, the flat tax income is probably the income concept most correlated with the individual’s earnings capacity and as such likely to remain reasonably constant through life. On the other hand, the gross income is a measure of the consumption possibilities of the individual at a given point in time; however, this income concept is very sensitive to where the individual is in his life-cycle – for example, young home owners may have large negative capital income which will reduce their gross income, while older tax payers may receive interest on their accumulated wealth, thereby increasing their gross income – these differences are as such not reflecting differences in actual earnings, but rather their different positions in their respective life-cycles. In Table 4.3 we have alternatively grouped tax payers by their gross income. According to this income definition, there are more low-income tax payers, reflecting, as noted above, that people who have recently finished their education or recently acquired a house typically have large negative capital incomes. In turn, this implies that the reductions in disposable income of the low-income tax payers, and the increases for high-income tax payers, become numerically larger, and that the distributional consequences appear more adverse, when the gross income concept is used in lieu of flat tax income. Concerning average tax rates out of gross income, we see that these rates increase for the lowest gross income groups while they decrease for most other income groups. The rather dramatic increase of 52.3 percentage points for the lowest income group arises from individuals with large gross incomes and large interest deductions, as noted above. Thus, when categorizing individuals by their gross income, the introduction of a flat tax seemingly has very adverse distributional effects.
Table 4.3
Changes in disposable income, by gross income groups Gross income (1000 kr.)
Change in disposable income
< 50
50–100 100–150 150–200 200–250 250–300 300–400 400–500 500–750
(1000 kr.) + > 50 0 0 + 20–50 0 3 624 + 10–20 302 34 730 + 1–10 70 366 493 770 +/– 1 286 296 367 534 – 1–10 44 998 108 116 – 10–20 16 308 41 374 – 20–50 23 858 43 186 – >50 23 254 15 402 Total 465 382 1 107 736 Average change in disposable income (kr.) –8 681 –1 212 Average tax rate before tax reform (%) 11.6 19.0 after tax reform (%) 63.9 20.7
750–1m
0 20 234 35 334 437 900 55 870 161 268 87 580 70 668 13 590 882 444
1 812 20 838 102 378 353 038 47 414 169 422 79 426 42 884 10 268 827 480
4 530 34 428 148 886 134 390 21 140 47 414 19 932 8 456 5 738 424 914
3 926 69 460 60 098 21 744 3 624 9 362 3 926 4 228 4 530 180 898
9 966 81 540 14 798 5 134 1 510 3 624 2 718 2 718 5 436 127 444
21 140 10 268 2 114 2 114 0 1 510 906 1 208 3 624 42 884
16 912 3 322 302 0 302 302 302 1 208 3 926 26 576
–3 064
64
6 763
13 069
24 029
35 031
40 671
34.1
39.6
43.7
47.1
50.0
51.9
52.3
55.6
46.4
36.5
39.5
40.6
42.3
42.9
44.0
45.5
42.6
49.7
Total
3 020 66 440 302 244 620 0 398 942 302 1 518 758 0 783 690 0 546 318 0 252 774 302 199 018 1 208 87 278 5 134 4 097 838
108 536 –95 082
760
123
Source: Computations from the Danish Law Model.
5 134 604 0 0 0 302 302 302 302 6 946
>1m
124
Nielsen, Frederiksen and Lassen
Finally, we see that the average increase in disposable income for all tax payers is 760 Danish Kroner. This reflects the fact that in the LM experiments, the tax reform is slightly underfinanced by approximately 2.9 billion Danish Kroner.35
4.7
FURTHER ISSUES
In section 4.5, we mentioned the problem of transitional relief to existing corporate capital. This aspect reflects the fact that one would expect substantial asset revaluation effects when the tax treatment of various types of assets are changed. The virtue of the flat tax is that it leaves the opportunity cost of capital undistorted while taxing consistently inframarginal returns and risk premiums. This will, however, imply drastic changes in the tax treatment of various asset types as compared to the current system. Assets that yield the opportunity cost of capital will benefit from its non-taxation; accordingly, bond holders will gain. Assets yielding inframarginal returns that are currently imperfectly taxed – for example, some corporate assets – will suffer from higher effective tax rates. And assets that yield the opportunity cost of capital but currently face negative tax rates, such as housing, will suffer. Hence, to build political support for the flat tax it will likely be necessary to provide transitional tax relief to certain types of assets. In the case of, for example, owner-occupied housing, this should be fairly easy to administrate. Similarly, retaining tax depreciation associated with the existing capital of business firms is also reasonably straightforward. It becomes more difficult, however, to provide transition relief for business debt or tax losses accumulated under the income tax. The reason is that such provisions would open up tax arbitrage opportunities. 36 Another important aspect of comprehensive tax reform is its impact on international tax relations. Here we will mention three such aspects. First, by way of the non-taxation of financial flows, the incentives to engage in profitshifting activities through within-firm financial transactions would increase. That is, multinationals would benefit from shifting positive financial income into the flat tax jurisdiction, while negative net financial income is used to offset income taxes in other jurisdictions. Such incentives, however, also exist under the income tax, but the scope is somewhat larger when one country operates a regime of real cash-flow taxation. Second, when transitional relief is not granted to corporations, foreign multinationals face capital losses on their Danish subsidiaries. If foreign governments allow resident firms to credit Danish taxes against their income tax liability, this effectively implies a lump-sum income transfer from the foreign to the Danish treasury. This in turn could prompt foreign govern-
The Flat Tax in Denmark
125
ments into demanding renegotiation of double taxation treaties. In effect, this is one more argument in favour of granting transition relief. Third, tax competition may cause problems for the flat tax regime. If we allow for maintaining the individual income tax allowance, the flat tax rate ends up around 44 per cent. Current policy, on the other hand, is aimed at cutting the corporate income tax rate from 34 to about 25 per cent in the not too distant future. This intention is probably based mostly on tax competition considerations. In this context, it is crucial to distinguish between the marginal and statutory tax rates on investment. The marginal rate determines the capital–labour ratio chosen by firms. It is zero under the flat tax; investment neutrality obtains. The statutory tax rate may influence the locational choice of multinationals as well as the incentives to engage in transfer pricing. With a fairly high tax rate on corporate cash flows, multinationals face pronounced incentives to locate elsewhere or move surpluses to entities in other countries.
4.8
CONCLUDING REMARKS
The object of the present chapter was to consider a radically different tax system for the Danish economy. Instead of the current three-tier income tax system we proposed an examination of the so-called flat tax. This tax, which basically is a tax on comprehensive labour income, subjects both wages and cash flows in firms to tax at the same rate. The ordinary return to capital is in effect treated as tax-exempt in our investigation. For the particular year 1992 we computed the rate of flat tax which would have been necessary to generate the same revenue as that collected by means of existing income taxes, assuming unchanged behaviour. While the result depended on the precise assumptions as to personal allowances and deductions, we in our benchmark case with unchanged personal allowances derived a flat tax rate of 44 per cent, thus considerably below statutory marginal tax rates on personal labour (and capital) income in the system in place in 1992. The consistent tax treatment of labour income, the rather low flat tax rate, and the absence of tax on pure capital income would probably bring about significant efficiency gains associated with the reduction or removal of distortions to labour supply, saving, and capital accumulation. However, even if personal allowances in the current income tax system were maintained, the alternative flat tax will overall score somewhat less well on distribution. To counteract this, the flat tax could be supplemented by other initiatives to preserve the redistributive features of the current tax and transfer system. There are several ways to approach the findings of the chapter. First, the fact that a flat tax at a rate significantly lower than today’s marginal tax rates
126
Nielsen, Frederiksen and Lassen
in the personal income tax area could collect the same revenue as the current income tax system could be taken as evidence of pronounced ineffectiveness inherent in the present system. The Tax Reform Act of 1993, analyzed in detail by Lange, Pedersen and Sørensen (Chapter 5 in this volume), has led to some improvement in the functioning of the tax system, though. The inefficiencies probably have to do with unwarranted asymmetries in the taxation of otherwise similar income components, with tax arbitrage, with income shifting and with a non-transparent system of itemized deductions. Second, the high marginal tax rates on wages and the tax distortions on saving and investment generate sizeable efficiency losses which can be avoided through a fundamental reworking of the income tax system. Third, the flat tax system with its simple compliance and administrative features, and with its robustness vis-à-vis (domestic) tax arbitrage and income shifting, could contain elements of a desirable future income tax system.
Notes 1.
2. 3. 4. 5. 6.
7. 8. 9. 10.
We wish to thank Frederik Hansen for detailed comments and excellent assistance with the computations described in section 4.6. In addition, we have benefited from comments from Per Bach Jørgensen and Michael Lundholm, and from comments on an earlier version of this chapter from Roger Gordon and Robert Haveman. The activities of EPRU are financed by a grant from the Danish National Research Foundation. Nielsen is corresponding author: Nansensgade 19, 5., DK–1366 Copenhagen K. E-mail:
[email protected]. In a pure dual income taxation system, personal capital income is taxed at a flat rate, while personal labour income is subject to progressive taxation, with the lowest marginal tax rate on labour income equal to the capital income tax rate. By international comparison, the Danish property assessment system is rather effective, though. In the Danish Law Model described in section 4.6, the figures for positive and negative capital income in 1992 were 18.0 and 58.2 billion Danish Kroner, respectively. See Frederiksen (1997). Jørgensen and Pedersen (1992) estimated the total administration costs of central and local tax authorities to 4.2 billion Danish Kroner in 1992. In addition to this, costs associated with tax compliance would likely fall considerably (see Slemrod and Bakija, 1997, on this). A proportional flat tax system with a basic income payment was originally proposed for Denmark by Jepsen (1971, 1975). This specific tax/transfer system has been further analyzed by Atkinson (1995). Some problems in the assessment of taxes due will remain – for example, those related to the distinction between private consumption and input expenses for the self-employed, to fringe benefits, etc. An exception pertaining to pension contributions and benefits. Both the ACE business tax and the cash-flow tax are special cases of the general neutral business taxes conceived by Boadway and Bruce (1984).
The Flat Tax in Denmark 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
21.
22. 23.
24. 25. 26. 27.
28. 29.
127
A thorough investigation of the economic effects on the US economy of such fundamental tax reforms can be found in Aaron and Gale (1996). Put forward by David Bradford as the ‘X-tax’ (Slemrod and Bakija, 1997). 1992 is still the last year for which complete National Accounts statistics (including input–output tables) for Denmark are available. ‘Real’ transactions as opposed to ‘financial’, defined in the Meade Committee Report (Institute for Fiscal Studies, 1978). Firms with a too low turnover, firms traded on the Danish stock exchange and firms in selected industries are not required to fill out the form. This is a deviation from the procedure in Gordon and Nielsen (1996, 1997). Such as unemployment insurance premiums, mileage allowance and alimony. Note that the recently abolished wealth tax, still present in 1992, is not included in the income tax measure although it could be argued that it is a tax on capital income. Source: Taxes and Duties, Statistics Denmark (1995). Note that this is the statutory level of the personal allowance which differs slightly from the average utilized level, since some income earners earn less than the amount corresponding to the statutory allowance and therefore did not make use of the entire allowance. The figure reported earlier to be 137.7 is the aggregate, utilized level of personal allowance. This corresponds to a figure of 30,618 Danish Kroner per person liable to income tax. In this calculation we ignore that some individuals have incomes in the interval 31,500–40,000 Danish Kroner. These individuals cannot utilize the personal allowance to its fullest extent, and we thereby overestimate the reported tax rate slightly. In contrast to the pure dual income tax system, there was some progression in the personal capital income tax. The model constructed by Frederiksen (1997) is a perfect foresight, small openeconomy model, with the household sector based on the overlapping generations framework in Blanchard (1985). All markets are competitive, and market clearing is assumed throughout. Households choose labour supply and consumption of goods over time so as to maximize utility subject to an intertemporal budget constraint, while firms choose employment and investment, subject to costs of adjusment, so as to maximize the initial market value of the firms’ equity. The government is parametric and levies a number of taxes on households and firms, spending the revenue on wages, goods and lump-sum transfers under an intertemporal budget constraint. Considerable attention is also devoted to the consequences of pre-announcement of the tax reform, and to the transition from one system to the other. Meaning that the cost of capital equals the market rate of interest irrespective of capital structure and dividend policy. The latter is closely connected to the distributional issues considered in the next section. Source: Taxes and Duties, Statistics Denmark (1995). It should be noted, however, that the progressive rates in the taxation of personal income ‘bite’ at relatively moderate income levels in comparison to tax systems in other industrialized countries. This is partly due to the fact that high-income earners typically also make large deductions for mortgage interest payments in their tax returns. In 1992, the wealth tax raised revenue equivalent to 0.3 per cent of the total income tax revenue.
Nielsen, Frederiksen and Lassen
128 30. 31. 32.
33.
34.
35.
36.
The impact on different types of assets is discussed in section 4.7. We also include in this figure non-taxable transfers such as welfare benefits, etc. It is to be expected that the majority of share holdings in type (a) companies indeed is in the hands of individuals. Conversely, data compiled by Unibørs (1994) suggests that only a small portion, perhaps some 20–30 per cent, of holdings of shares in companies traded on the stock exchange can be directly attributed to individuals. Of the total value of traded companies at the end of 1992 (204 billion Danish Kroner, Copenhagen Stock Exchange, 1992) banks, insurance companies, mortgage institutions and pension funds together owned 38 per cent of outstanding shares. Mutual funds owned 0.8 per cent, foreigners 2.9 per cent, and the state/government 1.1 per cent. This leaves so-called ‘strategic holdings’ of 35.4 per cent and a residual of 24.1 per cent. Strategic holdings are defined as ‘covering major share holdings to be considered as long-run strategic investments constituting a fixed element in the ownership structure, or ownership on the part of foundations’. A large part of these strategic holdings cannot directly be connected to individuals. The residual group comprises all ownership shares below 5 per cent of every single company’s outstanding stock. Only part of it corresponds to individual holdings, the remainder being intercompany holdings, etc. As stated, all this suggests that some 20–30 per cent of holdings in type (b) companies can be directly attributed to individuals. Finally, for type (c) firms, those that are not traded on the stock exchange and not dominated by large owners, the share of assets directly attributable to individual households should be intermediate between the shares in type (a) and (b) companies. The value of total outstanding assets in the three groups of companies in the beginning of 1992 could be, respectively, close to 100 billion, about 200 billion and some 300 billion Danish Kroner. Again, no reliable statistics are available. For computational reasons we opt for a different tax treatment of pensions than the one used in the calculation of the flat tax rate of 44 per cent. Here, we maintain the deductibility of pension contributions while subjecting pension benefits/payments to tax. This in fact is the way pensions are handled in the flat tax proposal of Hall and Rabushka (1985). The large average reduction in disposable income for the top bracket (> 1 million Danish Kroner) is due to the fact that a few individuals receive very high dividends from shares and therefore will be hit very hard by the (assigned) cash-flow tax. This comes about mainly because of the alternative treatment of pension contributions and benefits. Since in 1992 the former exceeded the latter by some 15 billion Danish Kroner, this implies an immediate loss of revenue of about 6.5 billion Danish Kroner. Note, however, that the beneficial efficiency effects of the flat tax could easily lead to overfinancing instead, when the flat tax rate is kept constant, as argued in section 4.5. This is discussed in the Meade Committee Report (Institute for Fiscal Studies, 1978).
References Aaron, H.J. and W.G. Gale (eds) (1996) Economic Effects of Fundamental Tax Reform (Washington, DC: Brookings Institution Press). Atkinson, A.B. (1995) Public Economics in Action – The Basic Income/Flat Tax Proposal (Oxford: Clarendon Press).
The Flat Tax in Denmark
129
Blanchard, O.J. (1985) ‘Debt, deficits and finite horizons’, Journal of Political Economy, 93, pp. 223–47. Boadway, R. and N. Bruce (1984) ‘A general proposition on the design of a neutral business tax’, Journal of Public Economics, 24, pp. 231–9. Bradford, D.F. (1995) ‘Consumption taxes: some fundamental transition issues’, EPRU Working Paper, 1995–15. Cnossen, S. (1996) ‘Company taxes in the European Union: criteria and options for reform’, Fiscal Studies, 17(4), pp. 67–97. Cnossen, S. (1997) ‘Dual income taxation: the Nordic experience’, paper presented at the ISPE conference on ‘Reform and Harmonization of Company Taxes in the European Union’ (The Hague, 3–5 April). Copenhagen Stock Exchange (1992) Monthly Report of the Copenhagen Stock Exchange, 247 (December). Frederiksen, N.K. (1997) ‘Macroeconomic and efficiency effects of the flat tax system’, EPRU, manuscript. Gammie, M. (1992) ‘Corporate tax harmonisation: an ACE proposal’, European Taxation, 31, pp. 238–42. Gentry, W.M. and R.G. Hubbard (1997) ‘Distributional implications of introducing a broad-based consumption tax’, Tax Policy and the Economy, 11. Gordon, R.H. and J. Slemrod (1988) ‘Do we collect any revenue from taxing capital income?’, Tax Policy and the Economy, 2, pp. 89–130. Gordon, R.H. and S.B. Nielsen (1996) ‘Tax avoidance and value-added vs. income taxation in an open economy’, Appendix. Gordon, R.H. and S.B. Nielsen (1997) ‘Tax avoidance and value-added vs. income taxation in an open economy’, Journal of Public Economics, 66, pp. 173–97. Hall, R.E. and A. Rabushka (1985) The Flat Tax (Stanford: Hoover Institution Press). Institutes for Fiscal Studies (1978) The Structure and Reform of Direct Taxation: The Report of a Committee Chaired by Professor James R. Meade (London: George Allen & Unwin). Jepsen, G.T. (1971) ‘Negative income tax. Economic and social consequences of a negative income tax as a way of simplifying the tax system’ (in Danish: Negativ indkomstskat. Økonomiske og socialpolitiske konsekvenser af en negativ indkomstskat til forenkling af skattesystemet), Socialt Tidsskrift, 6–7, pp. 306–26. Jepsen, G.T. (1975) The Taxes before, now and in 1985 (in Danish: Skatterne før, nu og i 1985) (Albertslund). Jørgensen, P.B. and N.J.M. Pedersen (1992) The Public Sector (in Danish: Den offentlige sektor) (Copenhagen: Handelshøjskolens Forlag). Lange, K., L.H. Pedersen and P.B. Sørensen (1997) ‘The Danish Tax Reform Act of 1993: effects on the macroeconomy and on intergenerational welfare’, paper presented at the conference on ‘Macroeconomic Perspectives on the Danish Economy’ (Hornbæk, 19–20 June); see Chapter 5 in this volume. Ministry of Finance (1996) Annual Report on Fiscal Policy 1996 (in Danish: Finansredegørelse 96) (Copenhagen). OECD (1995) The OECD Jobs Study: Taxation, Employment and Unemployment (Paris: OECD). OECD (1996) Economic Surveys: Denmark (Paris: OECD) (February). Slemrod, J. and J. Bakija (1997) Taxing Ourselves – A Citizen’s Guide to the Great Debate Over Tax Reform (Cambridge, Mass.: MIT Press). Statistics Denmark (1995) Taxes and Duties (in Danish: Skatter og Afgifter) (Copenhagen).
130
Nielsen, Frederiksen and Lassen
Sørensen, P.B. (1990) Danish Tax Policy in the 1990s (in Danish: Dansk Skattepolitik i 1990’erne) (Copenhagen: DJØFs Forlag). Sørensen, P.B. (1994) ‘From the global income tax to the dual income tax: recent tax reforms in the Nordic countries’, International Tax and Public Finance, 1(1) (May), pp. 57–79. Sørensen, P.B. (ed.) (1997) Tax Policy in the Nordic Countries (London: Macmillan). Unibørs (1994) Ownership Structure of the Danish Stock Market (in Danish: Ejerforhold på det danske aktiemarked) (Copenhagen). US Department of the Treasury (1992) Integration of the Individual and Corporate Tax Systems: Taxing Business Income Once (Washington, DC: US Government Printing Office).
Comment Michael Lundholm
1
INTRODUCTION
In Chapter 4 the authors investigate the consequences of a tax reform in Denmark, implying a move from the dual income tax (DIT) system in its 1992 Danish variant to a flat tax system.1 The general result of the chapter can be summarized as an opportunity for substantial efficiency gains at the cost of quite strong distributional consequences. Depending on how these consequences are weighted, such a reform might enhance welfare or not. But even if it is the case that the suggested (marginal) tax reform improves welfare it does not necessarily follow that a flat tax system is a global optimum. Therefore, a move to a flat tax might not be the most obvious candidate for policy reform. In the following I will refer to three theoretical arguments, which can be found in the literature, against the hypothesis that a flat tax system in general is a global optimum:
●
the DIT is optimal differentiated treatment of different tax bases is (most often) optimal when the necessary information for implementation of differential treatment is available non-constant marginal income tax rates are optimal under asymmetric information.
2
THE DIT IS OPTIMAL
● ●
In Sørensen (1994, 1998) a number of arguments for the optimality of the DIT are presented. These arguments are based on efficiency as well as equity considerations; here I give only two examples. The first argument for the beneficial tax treatment of capital income compared to labour income in the higher tax brackets is that due to implementation problems the definition of capital income is made in nominal terms. Given this restriction, the effect of inflation, then, is compensated by a lower statutory tax rate than would have been the result if capital income had been defined in real terms. Another argument mentioned by Sørensen (1994, 1998) and modelled by Nielsen and Sørensen (1997) is that the flat tax discriminates against physical and financial investment compared to investment in human capital, a discrimination which the DIT eliminates through a higher marginal tax rate on the upper tax bracket for labour income (i.e. the returns to human capital).
3
DIFFERENTIAL TAX TREATMENT OF DIFFERENT TAX BASES
Should different tax bases be subjected to differential tax treatment? Atkinson and Sandmo (1980), among others, argue that capital income and labour income should be given differential treatment depending on the elasticities of the tax bases. Clearly,
131
Lundholm
132
this requires that the tax authority has information about what is capital income and what is labour income. If information is not available there is an opportunity for tax avoidance and/or tax evasion. This consideration led the Lindbeck commission on the Swedish economy (see Lindbeck, 1994), discussing the benefit rates in the Swedish social insurance system, to argue in favour of flat rates. However, if differential treatment is optimal under complete information (that is, when information is available at no cost), then under incomplete information (when information is available only at a cost) abundant information also has a value. Depending on the marginal gains and costs to attain additional information it might be optimal for the tax authority to spend resources to achieve the necessary information for differential tax treatment. (See Lundholm, 1996) for a similar discussion for the area of social insurance.) If it is the case, as the authors argue, that capital income taxation (excluding corporate income tax revenue) is negative, then this is an indication that the information available to tax authorities to implement the present tax system is insufficient. Therefore, instead of changing the actual tax system towards a flat tax, even if it is welfareimproving, an alternative reform of the tax system is to leave statutory tax rates unchanged and instead spend more resources to achieve information (for example, auditing).2 Of course, such an alternative reform must be evaluated according to the same principles as the flat tax reform, but one cannot ex ante determine that it is not welfare-improving.
4
NON-CONSTANT MARGINAL INCOME TAX RATES
Should the income tax base be subjected to non-constant marginal tax rates? Mirrlees (1971) first analyzed the general non-linear optimum income tax system under incomplete information. Sadka (1997) showed theoretically that the marginal tax rate for the highest-ability individual should be zero. Slemrod et al. (1994) showed that this result in a sense carries over to a two-bracket, piece-wise linear and continuous tax function; the optimal marginal tax rate should be lower in the upper bracket than in the lower bracket. These results therefore indicate that even under incomplete information fewer restrictions on the tax system will be exploited if the tax system is optimal. On the other hand, they also argue against the optimality of the DIT.3
5
CONCLUSIONS
My conclusion is that there are rather strong theoretical arguments against the hypothesis that flat tax constitutes a global optimum. Also, the authors’ analysis does not answer the question whether the marginal reform with a flat tax that they consider is welfare-improving. Even if it is, I have also argued that there is an alternative area of reform which includes increased auditing.
Notes 1. 2.
The DIT developed in the Nordic countries is in various aspects described by Sørensen (1998). The possibility of exploiting information, generated by the tax system itself, for differential treatment is analyzed by Dillén and Lundholm (1996).
Comment 3.
133
During the discussion of the chapter Søren Bo Nielsen correctly pointed out that the numerical simulations in Mirrlees (1971) and Slemrod et al. (1994) indicated that the optimal non-linear income tax was close to linear. First I want to object that a linear tax is different from a flat (proportional) tax because of its lump-sum component. Second, Toumala (1984) showed in a simulation that for the utilitarian criterion optimal marginal tax rates were decreasing for reasonable assumptions about the marginal rate of substitutions between consumption and leisure. Also, the difference between optimal marginal tax rates in the middle of the first and fourth quartiles were as large as in the order of 20 percentage points in some simulations.
References Atkinson, A.B. and A. Sandmo (1980) ‘Welfare implications of the taxation of savings’, Economic Journal, 90 (September), pp. 529–49. Dillén, M. and M. Lundholm (1997) ‘Dynamic income taxation, redistribution, and the ratchet effect’, Journal of Public Economics, 90 (1) (January), pp. 69–93. Lindbeck, A. (ed.) (1994) Turning Sweden Around (Cambridge, Mass.: MIT Press). Lundholm, M. (1996) ‘Reform of social insurance: problems and solutions (Reform av socialförsäkringen. Problem och lösningar)’, in M. Lundholm (ed.), Report 6: Challenges for the Public Sector (Rapport 6: Utmaningar för den offentliga sektorn) (Uppsala: Nordic Economic Research Council. Mirrlees, J.A. (1971) ‘An exploration in the theory of optimum income taxation’, Review of Economic Studies, 38 (April), pp. 175–208. Nielsen, S.B. and P.B. Sørensen (1997) ‘On the optimality of the Nordic system of dual income taxation’, Journal of Public Economics, 63 (3) (January), pp. 311–29. Sadka, E. (1986) ‘On income distribution incentive effects and optimal income taxation’, Review of Economic Studies, 43 (2) (June), pp. 261–7. Slemrod, J., S. Yitzhaki, J. Mayshar and M. Lundholm (1994) ‘The optimal two-bracket linear income tax’, Journal of Public Economics, 53 (2) (February), pp. 269–90. Sørensen, P.B. (1994) ‘From the global income tax to the dual income tax: recent tax reforms in the Nordic countries’, International Tax and Public Finance, 1 (1) (May), pp. 57–79. Sørensen, P.B. (ed.) (1998) Tax Policy in the Nordic Countries (London: Macmillan). Toumalam, M. (1984) ‘On the optimal income taxation: Further numerical results’, Journal of Public Economics, 23, pp. 351–66.
5 The Danish Tax Reform Act of 1993: Effects on the Macroeconomy and on Intergenerational Welfare1 Kathrine Lange, Lars Haagen Pedersen and Peter Birch Sørensen
5.1
BACKGROUND
In June 1993 the Danish Parliament enacted a tax reform to be phased in over a 5-year period starting in 1994. Following recent trends in the OECD area, the Danish tax reform combined a broadening of the tax base with general cuts in marginal tax rates on income from labour and capital. In terms of the magnitude of the drop in marginal tax wedges, the Danish reform was not as ambitious as the recent tax reforms in the neighbouring Nordic countries (see Sørensen, 1998). On the other hand, a substantial part of the income tax cuts was financed via a rise in various ‘green’ taxes, including excise taxes on energy and water use. The Danish reform bill may therefore be seen as one of the most serious attempts so far to implement an ‘ecological’ tax reform. The reform aimed to improve individual incentives for work and saving through lower marginal tax wedges, to moderate union wage claims by lowering the tax burden on labour income, to limit the scope for tax avoidance and tax evasion by broadening the tax base and by strenghtening tax enforcement and to encourage environmentally friendly economic behaviour via higher taxes on polluting activities. In this chapter we apply a dynamic overlapping-generations model of the Danish economy to evaluate the effects of the tax reform on economic performance and on the level of economic welfare enjoyed by different generations. Our analysis suggests that the reform will tend to raise the welfare of all generations, in particular the welfare of those who are not yet born. We start out in section 5.2 by giving a brief overview of the 1993 tax reform. Section 5.3 describes the model underlying the quantitative analysis of the tax reform presented in section 5.4, and section 5.5 summarizes our main findings. 134
The Danish Tax Reform Act of 1993 5.2
135
THE 1993 TAX REFORM ACT
The Danish tax reform implied substantial cuts in marginal personal income tax rates in all income brackets and a cut in the taxable imputed rent from owner-occupied housing. These tax cuts were financed through (1) a new payroll tax (in Danish: ‘arbejdsmarkedsbidrag’) levied on employers and employees, (2) excise taxes on certain polluting activities, in particular energy use and water use,2 (3) elimination or reduction of various deductions from the labour income and business income tax bases, (4) increased taxation of realized capital gains on shares, commercial buildings and intangible assets such as goodwill, (5) increased taxation of a number of ‘fringe benefits’, including company cars and (6) improved enforcement procedures to limit the amount of tax evasion. In an earlier tax reform bill of 1985, Danish policy makers made a halfhearted attempt to introduce a version of the Nordic ‘dual’ income tax which separates the taxation of capital income from the taxation of labour income (see Sørensen, 1994). Since that time the marginal tax rates on the two types of income have therefore differed. Table 5.1 shows how the 1993 Tax Reform Act will affect the marginal tax rates on labour income when the reform has
Table 5.1
Statutory marginal tax rates on labour income, before and after the 1993 tax reform
Before the tax reform (1993) Personal income1
0–Exemption2 Exemption–130,000 130,000–168,000 168,000–240,000 240,000–
Marginal tax rate (%)
0.0 50.0 51.5 57.5 68.0
After full implementation of the tax reform (1998) Personal income3
0–29,300 29,300–130,000 130,000–234,900 234,900–
Marginal tax rate Excluding Including payroll tax payroll tax4 (%) (%) 0.0 37.5 43.5 58.0
8.0 42.5 48.0 61.4
Notes: 1. Measured in 1993 prices. 2. The exemption level was 25,500 Kroner for local government tax and 32,600 Kroner for central government tax. 3. Income after deduction for payroll tax, measured in 1994 prices. 4. Payroll tax levied on the employee. The employer’s payroll tax of 0.8 per cent is not included.
Lange, Pedersen and Sørensen
136
been fully phased in by 1998.3 The marginal labour income tax rate varies with the level of the tax-payer’s ‘personal income’, defined in Danish tax law as the sum of income from labour, public transfers and private pension schemes. It can be seen from Table 5.1 that the 1993 Tax Reform Act implied cuts in marginal labour income tax rates of about 7–9 percentage points for all taxpayers with earnings above the (low) personal exemption level. As indicated in Table 5.2, the marginal tax rate on capital income depends on the sum of the taxpayer’s personal income and his or her net capital income. For taxpayers with positive net capital income above a certain level, there is a surtax which is also levied on personal income, whereas taxpayers cannot deduct negative net capital income from the tax base for the surtax. As a consequence, the top marginal tax rate is higher for positive than for negative capital income. We see from Table 5.2 that the tax reform caused a drop in the marginal tax rates on capital income roughly comparable to the fall in marginal labour income tax rates, except for the highest levels of capital income.
Table 5.2
Statutory marginal tax rates on capital income, before and after the 1993 tax reform
Before the tax reform (1993)
After full implementation of the tax reform (1998)
Personal income plus net capital income1
Personal income plus net capital income3
O–Exemption2 Exemption – 168,000 168,000–
Marginal tax rate on Positive Negative net net capital capital income income (%) (%) 0.0
0.0
51.5 57.5
51.5 51.5
0–29,300 29,300– 130,000 130,000– 234,900 234,900–
Marginal tax rate on Positive Negative net net capital capital income income (%) (%) 0.0
0.0
37.5
37.5
43.5 58.04
43.5 43.5
Notes: 1. Measured in 1993 prices. 2. The exemption level was 25,500 Kroner for local government tax and 32,600 Kroner for central government tax.
3. Income after deduction for payroll tax, measured in 1994 prices. 4. This high rate applies only to positive net capital income exceeding 20,000 Kroner (40,000 Kroner for married couples).
The Danish Tax Reform Act of 1993
137
Table 5.3 shows how the changes in statutory tax rates reported in Tables 5.1 and 5.2 are estimated to affect the various tax parameters in our simulation model of the Danish economy (the EPRU model).4 The marginal and average tax rates on labour income (including payroll tax) were calculated as weighted averages of the corresponding statutory tax rates in the different income brackets, with the weights being given by the proportion of income in each bracket in 1993. The marginal tax rate on capital income was estimated in a similar manner, with the weights calculated as the numerical value of net capital income in each income bracket relative to the total numerical amount of capital income in our data set. Table 5.3 includes estimated tax rates on dividends. ‘Dividends’ are interpreted broadly as all payments from unincorporated as well as incorporated firms to their owners, and the dividend tax rate is therefore calculated as a weighted average of the tax rates on corporate dividends and the tax rates applying to income from proprietorships and partnerships. The ‘dividend’ tax rate differs between the manufacturing and the construction sector because of the difference in the relative importance of unincorporated firms in the two sectors. As a parallel to our broad interpretation of dividends, we take ‘capital gains on shares’ to include all gains in the market value of non-corporate as well as corporate firms. However, the effective tax rate on these capital gains is estimated only on the basis of the statutory tax rates on realized gains on shares in corporate firms.5 In order to protect the ability of Danish firms to compete in international markets, the ‘green’ taxes included in the tax reform package were designed to fall mainly on the household sector, without any significant direct effect on the production costs of firms. We have therefore modelled the (non-environmental) effect of the green taxes as a gradual increase in the average rate of excise tax on non-durable consumption goods, as indicated in Table 5.3. The revenue effects of changes in various deductions and similar basebroadening measures are captured in the EPRU model via changes in the various types of lump-sum subsidies appearing in Table 5.3. For example, the income and revenue effects of lower deductions from the labour income tax base and from the assumed stronger enforcement of labour income taxes are reflected in the gradual fall in the subsidy to members of the labour force. By analogy, the broadening of the business income tax base (through less favourable tax rules for foreign direct investment and through a phasing out of deductions for inventory write-downs and for allocations to investment funds) is modelled as a lump-sum tax on firms. Despite the fairly high marginal capital income tax rate reported in Table 5.3, the revenue from Danish capital income taxes (excluding corporation tax) was actually negative in 1993. This was due to the
138 Table 5.3
Lange, Pedersen and Sørensen Estimated effective tax and subsidy rates, before and after the 1993 tax reform
Employee payroll tax rate Employer payroll tax rate Average effective tax rate on labour income Marginal effective tax rate on labour income Net replacement ratio (After-tax rate of unemployment benefit/ After-tax wage rate) Coefficient of Residual Income Progression1 Marginal tax rate on capital income Effective marginal tax rate on dividends from manufacturing firms Effective marginal tax rate on dividends from construction firms Effective marginal tax rate on accrued capital gains on shares Statutory tax rate on business profits Imputed rate of return on owner-occupied housing Aggregate excise tax rate VAT rate Subsidy to workers (Aggregate subsidy/Aggregate wage bill2) Subsidy to wealth owners (Aggregate subsidy/Aggregate capital income2) Subsidy to manufacturing firms (Aggregate subsidy/Aggregate net profits3) Subsidy to construction firms (Aggregate subsidy/ Aggregate net profits3) Subsidy to all citizens (Aggregate subsidy/ GDP at factor prices)
1993
1994
1995
1996
1997
1998
0.0000 0.0000
0.0500 0.0000
0.0600 0.0000
0.0700 0.0000
0.0800 0.0030
0.0800 0.0030
0.4410
0.4310
0.4240
0.4160
0.4090
0.3940
0.6040
0.5920
0.5780
0.5600
0.5530
0.5350
0.7800
0.7790
0.7750
0.7650
0.7630
0.7600
0.7080 0.5110
0.7170 0.4790
0.7330 0.4680
0.7530 0.4590
0.7560 0.4480
0.7670 0.4280
0.3520
0.3430
0.3370
0.3130
0.3090
0.3000
0.3520
0.3400
0.3310
0.3080
0.3020
0.2890
0.1500
0.2500
0.2500
0.2500
0.2500
0.2500
0.3400
0.3400
0.3400
0.3400
0.3400
0.3400
0.0250 0.1300 0.2500
0.0200 0.1380 0.2500
0.0200 0.1450 0.2500
0.0200 0.1520 0.2500
0.0200 0.1590 0.2500
0.0200 0.1660 0.2500
0.0371
0.0394
0.0360
0.0311
0.0298
0.0284
0.1450
0.1310
0.1330
0.1300
0.1230
0.1170
0.0000 –0.0072 –0.0048 –0.0126 –0.0270 –0.0548
0.0000
0.0000
0.0000 –0.0271 –0.0280 –0.0581
0.0127 –0.0125 –0.0002
0.0037
0.0075
0.0043
Notes: 1. Defined as (1-marginal effective tax rate on labour income)/(1-average effective tax rate on labour income). 2. 1993 income before tax. 3. 1993 profits before interest and tax, after deduction for economic depreciation.
The Danish Tax Reform Act of 1993
139
unlimited deductibility of interest expenses combined with the fact that the returns to pension saving and the returns to owner-occupied housing are taxed rather lightly according to special rules. To capture the revenue effect of these asymmetries in capital income taxation, the EPRU model assumes that wealth owners receive a lump-sum subsidy which is exogenous from the viewpoint of the individual saver. Our model simulations thus assume that the return to the marginal unit of saving is subject to the ordinary marginal capital income tax rate reported in Table 5.3. This seems reasonable, since Danish tax law and collective bargaining agreements impose limits on the amount of (pension) savings which can be made in taxsheltered form. Since the 1993 tax reform was intended to provide a stimulus to aggregate demand, it was not fully financed by the financing instruments mentioned at the beginning of this section. However, in order to separate the structural incentive effects of the tax reform from the macroeconomic effects arising from government debt dynamics, we have assumed that the residual revenue loss from the reform is neutralized through continuous adjustment of a general government lump-sum transfer which is paid to all age groups. In other words, the simulations reported below assume that the government budget is kept in continuous balance. The bottom row of Table 5.3 shows that the tax reform generated a revenue loss of about 2.5 per cent of 1993 GDP at factor prices in the fiscal year 1994, according to the model simulations described below.6 In the subsequent years the loss of revenue gradually fell below 1 per cent of GDP. Our simulation procedure implicitly assumes that this missing revenue can be recouped via some non-distorting government financing instrument. In practice, the Danish government may have to resort to distortionary taxation in order to satisfy the intertemporal government budget constraint. In that case, our simulations will obviously tend to overestimate the efficiency gains from the tax reform. On the other hand, some of the base-broadening measures which we have modelled as a reduction of lump-sum subsidies took the form of an elimination of deductions for expenditures that did not represent true economic costs of doing business (for example, excessive allowance for inventory write-downs) and which therefore distorted the investment pattern. Other initiatives to broaden the tax base included the full taxation of fringe benefits; a measure which would tend to eliminate distortions in the pattern of employee compensation. Ceteris paribus, the modelling of these financing elements as a reduction of lump-sum subsidies implies a downward bias in our estimates of the efficiency gains from the tax reform. In Lange, Pedersen and Sørensen (1997) we have analyzed how the welfare gains from the reform are affected if one assumes that the revenue shortfall must be financed through a higher rate of VAT rather than through non-distortionary
Lange, Pedersen and Sørensen
140
lump-sum taxation. It turns out that this alternative financing assumption makes little difference for our qualitative and quantitative results.
5.3
THE EPRU MODEL
General Assumptions Our quantitative assessment of the 1993 tax reform is based on the EPRU simulation model described in detail in Jensen et al. (1996). The behavioural equations in the model are derived from intertemporal optimization by households and firms with forward-looking expectations. The household sector of the model consists of an infinite number of overlapping generations which are divided into workers and retirees. An exogenous retirement age determines the relative size of these two groups. To facilitate aggregation, the EPRU model follows Blanchard (1985) in assuming a constant probability of death across generations and a birth rate equal to the death rate, implying a constant population. Households derive utility from goods consumption and incur disutility from work. Total goods consumption is a Cobb–Douglas aggregate of the consumption of housing services (taken to be proportional to the housing stock) and the consumption of non-durables which in turn is a CES aggregate of domestic and imported non-durables. The business sector of the EPRU model includes a ‘manufacturing’ sector and a ‘construction’ sector. Output from the manufacturing sector is used for domestic private and public consumption of non-durables, for exports and for investment in both production sectors. Domestic manufacturing output is an imperfect substitute for foreign (imported) goods, and export demand is given by a downward-sloping constant-elasticity world-demand curve, implying an endogenous determination of the terms of trade. The non-tradable output from the construction sector consists of new housing units serving to meet the demand for housing from domestic residents. In both production sectors total output is given by a CES function of imported materials and domestic value-added, with value-added being a Cobb– Douglas function of the inputs of capital and labour. Labour is perfectly mobile between the two domestic production sectors but immobile across borders. Financial capital is perfectly mobile internationally and capital income taxation is based on the residence principle, so the domestic interest rate before tax equals the exogenous pre-tax world interest rate. Because of adjustment costs in business investment, the domestic physical capital stock adjusts only gradually to its long-run equilibrium level. A distinctive feature of the EPRU model is that the labour market is imperfectly competitive. As we shall see below, this implies that the effects of
The Danish Tax Reform Act of 1993
141
taxation on employment are quite different from those emerging in a competitive labour market. We now describe the markets for labour and capital in more detail, focusing on the impact of taxation.
The Labour Market In each production sector total labour input is an aggregate of several imperfectly substitutable skill types. Since perfect competition prevails in all output markets, aggregate labour demand is given by the usual condition that the marginal product of aggregate labour input must equal the real product wage. Labour of each skill type is organized in a craft union monopolizing the supply of that particular skill. Since the model abstracts from employers’ associations, each union may dictate its wage rate to employers, subject to a downward-sloping labour demand curve deriving from the profit maximization of firms. The resulting total level of employment (hours) is shared equally between all members of the union. This assumption of work sharing eliminates intractable aggregation problems by avoiding a distinction between fully employed and fully unemployed workers. Unions set their wage rates with the purpose of maximizing the expected life-time utility of their members, but since the number of different skills (unions) is large, each union has a negligible influence on the general wage level and on the investment level of firms. In these circumstances maximization of life-time union member utility becomes equivalent to maximizing union member real income during each time period, adjusted for that period’s disutility of work, measured in aggregate consumption equivalents. The union therefore raises its hourly wage rate to the point where the ensuing marginal gain in the representative member’s real income equals the resulting marginal increase in the disutility of work. Given this behaviour, the real after-tax wage rate is a mark-up over the marginal opportunity cost of work as perceived by the union. This opportunity cost consists of the marginal disutility of work plus the real after-tax rate of unemployment benefit. The union’s wage claim will generate involuntary unemployment – that is, the union will ration the working hours allocated to each individual member – because all members gain collectively from the higher wage rate made possible through a restriction of individual labour supply. The union’s mark-up of the pre-tax wage rate over the marginal disutility of work will be higher the higher the net replacement ratio (defined as the ratio of the after-tax rate of unemployment benefit to the after-tax wage rate), the larger the average direct and indirect tax wedge and the higher the degree of tax progressivity. In particular, a rise in the average labour income tax rate will drive up the pre-tax real wage, whereas a rise in the marginal labour income tax rate (keeping the average tax rate constant) will reduce it.
142
Lange, Pedersen and Sørensen
These tax effects are consistent with a large number of recent empirical studies – see section 2.2 in Sørensen (1997) for a brief review – and they differ from the effects of taxation in a competitive labour market, where a higher marginal tax rate would drive up the wage rate through a negative substitution effect on the supply of working hours. The intuition for these tax effects in the EPRU model is that a higher average tax rate on labour income makes the work option less attractive relative to the collection of unemployment benefits, thus inducing unions to reduce employment through more aggressive wage claims, whereas a higher marginal tax rate makes it less costly for unions to raise membership employment through wage moderation, since a given fall in the pre-tax wage rate will now lead to a smaller fall in the after-tax wage rate. The mark-up of wages over the marginal disutility of work will also be larger the lower the elasticity of substitution between different skill types, since less substitutability between the members of different unions increases the monopoly power of each individual union. A deterioration of the international terms of trade will likewise raise the real domestic product wage claimed by unions, thus increasing the labour costs of domestic firms, because a higher relative price of imported consumer goods requires a rise in the nominal wage relative to the price of domestic goods if workers are to maintain an unchanged real consumer wage. The Markets for Business and Housing Capital Over time, taxation will affect the economy’s supply side not only via its impact on the labour market, but also via its effect on the accumulation of capital. In the EPRU model firms choose their investment level with the purpose of maximizing the discounted value of after-tax dividends, taking account of the fact that gross investment involves installation costs which are given by a convex cost function. An exogenous fraction of net investment is financed by debt, while the remaining investment is financed through retained profits. The physical capital stock depreciates at a constant exponential rate, but the tax code allows deduction for depreciation at an accelerated rate exceeding true economic depreciation. In long-run equilibrium where gross investment equals total economic depreciation, the optimal capital stock of the representative firm in each production sector is determined by the requirement that the marginal product of capital adjusted for profits tax must equal the marginal steady-state cost of capital which consists of the sum of the tax-adjusted weighted average cost of debt and equity finance, the tax-adjusted cost of depreciation, and the marginal installation cost arising from the fact that an extra unit of capital generates more replacement investment. A rise in the tax rate on interest income will stimulate investment by lowering the steady-state cost of capital, since a higher
The Danish Tax Reform Act of 1993
143
tax burden on interest reduces the opportunity cost of investing in stocks rather than interest-bearing assets. On the other hand, a rise in the corporate income tax rate or a rise in the effective tax rate on capital gains on shares will raise the required pre-tax rate of return on capital by raising the tax burden on equity-financed investment. It should be mentioned that the arbitrage condition determining the market value of firms assumes the marginal shareholder to be a domestic resident subject to domestic personal tax. This assumption may be problematic when it comes to corporations quoted in the stock exchange, where the marginal investor may be an institutional or a foreign investor. To the extent that these latter investor groups determine stock prices at the margin, the EPRU model will tend to overestimate the effects of changes in domestic personal tax rates on stock market values and corporate investment. We turn now to the housing market. The demand for housing investment follows from the demand for housing services which varies negatively with the user cost of owner-occupied housing.7 In Danish tax law nominal mortgage interest payments are deductible against the capital income tax base while an imputed rate of return on housing is subject to capital income tax. Since the imputed rate of return is set below the real interest rate, and since nominal capital gains on houses are tax exempt, a rise in the capital income tax rate reduces the user cost of housing as the additional tax cut arising from interest deductibility will exceed the extra tax on the imputed return. The EPRU model assumes that households face no costs of adjusting the size of their housing stock. The adjustment of this stock is nevertheless constrained from the supply side of the housing market, due to the installation costs associated with investment in the construction sector. In the model there is always some new construction going on, if only for replacement investment, so the prices of existing houses are tied to the price of a newly constructed housing unit. If the economy were seriously depressed, bringing all construction activity to a halt, the price of old housing units might in practice fall below the costs of constructing a new house, but the EPRU model cannot accommodate such a depression scenario. Because of perfect international mobility of financial capital, total domestic investment is not constrained by the level of total domestic saving, since any excess of investment over saving may be covered through capital imports. The evolution of the domestic capital stock is thus determined entirely by the variables affecting investment demand, whereas the evolution of the domestic stock of wealth is determined by the level of domestic saving which is found by subtracting consumption from income. The solution to the intertemporal optimization problem of households implies that total real private consumption (adjusted for the disutility of work) is proportional to the sum of the real stocks of human and non-human wealth, where ‘nonhuman wealth’ includes the market value of the housing stock as well as the
144
Lange, Pedersen and Sørensen
market value of business firms plus private sector bond holdings. The propensity to consume out of wealth varies positively with the after-tax interest rate, reflecting our assumption that the intertemporal elasticity of substitution in consumption is less than unity so that the income effect dominates the substitution effect. However, a rise in the after-tax interest rate will also reduce the stock of wealth, since it will cause future after-tax earnings to be discounted more heavily. In the EPRU model this wealth effect dominates the effect on the propensity to consume, so the net effect of a higher aftertax interest rate on total consumption is negative.
Calibration The values of the most important parameters in the EPRU model (besides the tax parameters given in Table 5.3) are reported in Table 5.4. Parameter values have been chosen such that the initial long-run equilibrium of the model reproduces the relevant shares and ratios for Denmark in the early 1990s as accurately as possible (see Jensen et al., 1996, Table 2, for details). For example, the retirement age and the probability of death have been chosen to imply a realistic proportion of pensioners to workers, and our combined choice of the utility discount rate and the intertemporal substitution elasticity in consumption generates a realistic stock of private wealth relative to GDP. Table 5.4 also reports some important elasticities implied by our choice of parameter values. The implicit elasticity of working hours w.r.t. the real aftertax consumer wage is equal to 0.10, corresponding quite well to the empirical estimate for Denmark found by Smith (1995). As mentioned on p. 142, union monopoly power – and hence the ability of unions to shift labour taxes onto pre-tax wages – depends on the elasticity of substitution between different labour skills. By choosing the value of this elasticity, we can therefore calibrate the response of the pre-tax consumer real wage to changes in the average retention ratio (defined as 1 minus the average labour income tax rate) as well as the wage response to changes in the marginal retention ratio (defined as 1 minus the marginal labour income tax rate). Our chosen substitution elasticity implies an elasticity of the pre-tax real wage w.r.t. the average retention ratio equal to –1.09, and a real wage elasticity w.r.t. the marginal retention ratio equal to 0.09. For comparison, the extensive empirical Swedish study by Holmlund and Kolm (1995) estimated the real wage elasticity w.r.t. to the average retention ratio to be between –0.73 and –1.15, while the elasticity w.r.t. the marginal retention ratio was estimated to be between 0.10 and 0.15. Indeed, the Danish study by Hansen, Pedersen and Sløk (1996) found considerably higher tax elasticities in Danish blue-collar wage formation, so we consider the tax elasticities implied by our calibration of the EPRU model to be rather conservative.
The Danish Tax Reform Act of 1993
145
Table 5.4 Calibration of the EPRU model (selected exogenous variables and parameters) Intertemporal elasticity of substitution in consumption Numerical elasticity of substitution between foreign and domestic non-durables Share of housing consumption in total private consumption Elasticity of marginal disutility of work w.r.t. working hours Numerical elasticity of exports w.r.t. the real exchange rate Rate of time preference Years of active labour market participation per generation Probability of death Real rate of interest before tax Rate of economic depreciation of the housing stock Rate of economic depreciation of the capital stock in the manufacturing sector Rate of depreciation for tax purpose in the manufacturing sector Rate of economic depreciation of the capital stock in the construction sector Rate of depreciation for tax purpose in the construction sector Elasticity of marginal installation costs w.r.t. investment Elasticity of substitution between imported materials and domestic value-added Elasticity of substitution between different labour skills Capital share in domestic value-added Ratio of business debt to replacement value of business capital stock
0.600 2.000 0.155 10.000 2.500 0.010 36.000 0.040 0.050 0.025 0.064 0.170 0.119 0.259 1.000 0.670 9.000 0.240 0.550
(Implicit elasticities and ratios) Elasticity of labour supply w.r.t. the real net wage Elasticity of real wage claims w.r.t. the average retention ratio Elasticity of real wage claims w.r.t. the marginal retention ratio Ratio of pensioners to workers Ratio of public consumption to GDP at factor prices Ratio of public debt to initial GDP
0.100 –1.085 0.085 0.299 0.300 0.511
Economic Distortions Included in the EPRU Model: A Summing Up To understand the welfare effects of the Danish 1993 tax reform predicted by the EPRU model, it is useful to summarize the various economic distortions which exist in the model equilibrium prevailing before the reform. First of all, there is a wedge between the marginal pre-tax real consumer wage and the marginal disutility of work. The wedge arises from direct and indirect taxes, from unemployment benefits and from union monopoly power. Because of all these distortions, a rise in employment will ceteris paribus generate a first-order welfare gain.
146
Lange, Pedersen and Sørensen
Second, the personal tax on capital income distorts consumer choices between present and future consumption by driving the private net return to saving considerably below the social return which equals the exogenous world rate of interest before tax. As a consequence, private savings are inoptimally low in the initial equilibrium, so a policy-induced rise in savings will tend to be welfare-improving. Third, the marginal product of capital (adjusted for depreciation and installation costs) differs from the social cost of capital which is also given by the world interest rate.8 The wedge between the pre-tax marginal return to capital and the pre-tax world interest rate is due to accelerated depreciation for tax purposes and to an asymmetric tax treatment of the return to equity and debt. Given our estimates of the pre-reform tax parameters, the marginal social (that is, pre-tax) return to capital was below the social cost of capital, implying – ceteris paribus – that a reallocation of domestic savings away from domestic business investment and towards international portfolio investment would tend to improve economic efficiency, thereby raising consumer welfare. As a fourth distortion, housing investment is subsidized by the income tax system since the imputed taxable rate of return on housing falls short of the pre-tax interest rate. The value of this subsidy is higher, the higher the capital income tax rate, and the lower the imputed rate of return. The consumption of housing services is further encouraged relative to consumption of non-durables by the fact that only the latter goods are subject to excise taxes. Given the specification of household preferences in the EPRU model, it would be efficient to have a uniform indirect tax on all goods and services, including housing services, since they are all equally substitutable with leisure and all have uniform income elasticities. Abstracting from the possible beneficial environmental effects, a rise in excises on non-durables thus tends to exacerbate the distortion of the pattern of consumption. On the other hand, a switch from income taxation to consumption taxes will raise the overall savings level in the EPRU model and will thereby tend to offset the initial saving distortion. With these observations in mind, we are now ready to consider the effects of the Danish tax reform.
5.4 MACROECONOMIC AND WELFARE EFFECTS OF THE TAX REFORM Decomposing the 1993 Tax Reform It is illuminating to decompose the tax reform package into those elements which mainly affect the static distortions in the labour market and those
The Danish Tax Reform Act of 1993
147
which mainly impact on the intertemporal distortions in the capital market. The Danish 1993 tax reform may thus be seen as a combination of a labour tax reform and a capital tax reform. The labour tax reform includes (L1) the cut in the marginal and average tax rates on labour income and the associated change in the net replacement ratio, (L2) the introduction of payroll taxes on employers and employees, (L3) the measures to broaden the labour income tax base (modelled as a fall in lump-sum subsidies to workers), (L4) the rise in various ‘green’ excise taxes and (L5) the cut in the rate of imputed return on owner-occupied housing. The latter element is counted as part of the labour tax reform because it affects the cost of housing services. The tax on imputed rent may thus be treated as an excise tax which impacts on wage formation via its effect on the consumer price index. The remaining elements of the 1993 reform may be seen as a capital tax reform comprising (K1) the cut in the marginal tax rate on capital income, (K2) the rise in the effective marginal tax rate on capital gains, (K3) the cut in the effective marginal tax rate on dividends and (K4) the broadening of the business income tax base (modelled as a fall in lump-sum subsidies to firms). Before discussing the effects of the total tax reform, we consider the isolated effects of the labour tax reform and of the capital tax reform.
Effects of the Labour Tax Reform Figures 5.1–5.7 illustrate the simulated effects of the labour tax reform on some key macroeconomic aggregates and on intergenerational welfare. All our simulations assume (somewhat heroically) that the Danish economy was in long-run equilibrium in the pre-reform year of 1993. The Tax Reform Bill was enacted in June 1993, with a 4-year phase-in period starting on 1 January 1994. Since expectations are forward-looking in the EPRU model, all changes in tax rates from the beginning of 1994 are therefore treated as fully anticipated. Figure 5.1 predicts that the labour tax reform generates a total rise in employment of about 1.5 percentage points over the phase-in period. This development occurs despite the positive impact on real labour costs arising from the increase in green taxes (which raises the indirect tax wedge), from the introduction of the (small) payroll tax on employers and from the fall in the degree of labour tax progressivity documented in Table 5.3. While all of these reform elements tend to generate additional wage pressure, they are outweighed by the wage-moderating effect of the fall in the average labour income tax rate combined with a (modest) fall in the net replacement ratio (see Table 5.3).
102
101. 5
101
100.5
100
99.5
Aggregate employment (hours)
148
Figure 5.1
Figure 5.2
Aggregate human wealth
149
Aggregate non-human wealth
150
Figure 5.3
Figure 5.4
Physical consumption of non-durables
151
Aggregate physical stock of business capital
152
Figure 5.5
Figure 5.6
Aggregate physical stock of housing capital
153
154
Figure 5.7
Intergenerational welfare effects (compensating variation in per cent of initial wealth)
5.00
4.00
3.00
2.00
1.00
0.00 –25
–1.00
–2.00
–20
–15
–10
–5
0
5
10
15
20
25
30
35
40
45
50
The Danish Tax Reform Act of 1993
155
The rise in employment caused by the fall in real product wages generates an increase in the stock of human capital which stimulates consumption. The rise in the demand for housing services is particularly large, since the cut in the taxable rate of return imputed to owner-occupiers and the imposition of green taxes on certain non-durables induce substitution towards housing consumption. As a consequence, the stock of housing capital increases. The stock of business capital also goes up somewhat, since higher employment raises the marginal productivity of the initial capital stock above the cost of capital, thereby increasing the profitability of business investment. Consumption of non-durables fluctuates during the phase-in period as a result of intertemporal price speculation. As consumers realize that nondurables will gradually become more expensive because of rising excise taxes, they react by raising the consumption of non-durables in the initial part of the phase-in period (where excises are still relatively low) at the expense of consumption during the subsequent years. Figure 5.7 highlights the effects of the tax reform on the welfare of current and future generations. Welfare is measured by the compensating variation, expressed as a percentage of the total initial human and non-human wealth of the generation in question. The recorded welfare gain for a particular generation thus indicates the proportionate amount of wealth which could be taken away from that generation without reducing its welfare below the prereform level, given the after-tax prices and wages prevailing after the tax reform. On the horizontal axis of Figure 5.7, Generation 0 is the generation entering the labour market at the beginning of the first year of the tax reform – that is, 1994. Generations entering the labour market in subsequent years (future generations) are ordered to the left of Generation 0. To the right of Generation 0 we find generations who became active in the labour market prior to the tax reform. To interpret the welfare figures in Table 5.7, it is important to keep in mind that our assumption of a constant probability of death implies that all generations have the same expected remaining lifetime, independent of their age. All households therefore continue to save throughout their lives, although saving occurs at a reduced rate after the time of retirement. Consequently, a generation’s stock of non-human wealth is systematically larger the older it is. At the same time a generation’s human wealth tends to decline with age as a result of fewer remaining years of labour market participation. These are the reasons why the welfare gains from the labour tax reform are smaller and are ultimately turned into a loss, as we consider generations of rising age in Figure 5.7: since the rise in the indirect green taxes erodes the purchasing power of previously accumulated wealth, and since generations with fewer remaining years of labour market participation experience a smaller gain from the lower tax burden on labour and from the
156
Lange, Pedersen and Sørensen
rise in employment, the total gain is smaller, or the proportionate loss is greater, the older the generation is. We also see from Figure 5.7 that future generations will experience a (slightly) larger gain, the further into the future they enter the labour market. The reason is that, as time proceeds, households are on average wealthier, since there are more people in the economy who have lived longer under the new and more efficient system of labour income taxation. Aggregate consumption demand therefore rises over time, and this tends to raise the relative price of domestically produced goods, implying an improvement of the international terms of trade which will further stimulate employment by lowering the real product wage. With employment and real consumer wages rising slightly over time due to improving terms of trade, new generations will attain ever higher levels of human wealth and welfare as time passes by. According to Figure 5.7 generations who have entered the labour market shortly before the reform will gain more than the generations entering the market shortly after the reform, because the former generations benefit from a short-term capital gain on their pre-existing housing wealth, because of the increase in housing demand mentioned above. Of course, the older living generations also experience a capital gain on their houses, but with fewer remaining years of labour market participation they benefit less from the labour tax cuts and therefore gain less than future generations. Effects of the Capital Tax Reform As illustrated in Figures 5.1–5.7, the effects of the capital tax reform are very different from those of the labour tax reform. In the short run, the capital tax reform is contractionary. This is mainly because of the fall in the marginal tax rate on capital income, which has a number of depressing short-run effects. First of all, the higher after-tax interest rate causes future labour and transfer income to be discounted more heavily, thereby cutting into human wealth. Secondly, the lower capital income tax rate raises the user cost of housing (since the taxable imputed rate of return is lower than the deductible rate of interest), causing a gradual decline in the housing stock by forcing housing investment below the replacement level. The falling demand for housing also induces a fall in the price of existing housing units, implying a drop in nonhuman wealth which further depresses consumption. As a third depressing factor, the higher after-tax interest rate tends to reduce non-human wealth and consumption by reducing the present value of future dividend flows from the existing capital stock, and it also discourages business investment by raising the opportunity cost of equity finance. The higher effective tax rate on capital gains likewise tends to reduce business investment by raising the tax burden on new equity-financed investment.9 Nevertheless we see from Figure 5.5 that there is a slight initial rise in the
The Danish Tax Reform Act of 1993
157
business capital stock before the contractionary effects on investment manifest themselves. This initial investment stimulus reflects an intertemporal substitution effect caused by the gradual fall in the effective dividend tax rate over the phase-in period. Ceteris paribus, if the dividend tax rate is falling over time, it becomes attractive for shareholders to forego dividends today in order to obtain more dividends tomorrow. In other words, the falling dividend tax rate in the initial stage of the capital tax reform induces corporations to retain and reinvest more profits in the short run to be able to pay out more dividends later when the dividend tax rate is lower. However, since the negative short-run effect on consumption and housing investment is much larger than the small expansionary short-run effect on business investment, the initial impact of the capital tax reform is a fall in aggregate demand which reduces the price of domestic relative to foreign goods. Because of the ensuing deterioration of the international terms of trade, the domestic real product wage goes up, causing a slight fall in employment. After a few years, consumption demand starts to recover, as consumers build up non-human wealth by raising their savings out of current income, because the higher net interest rate provides a permanent stimulus to savings, as we explained on p. 144. The gradual rise in non-human wealth and consumption pulls up aggregate demand, thus reversing the fall in the terms of trade and the concomitant drop in employment. The demand for housing also gradually increases as consumers become wealthier, and in the long-run aggregate private wealth rises considerably, reflecting a positive net effect of higher financial and housing wealth combined with a lower stock of domestic business capital. Since the marginal social return to business investment is initially below the social cost of capital (cf. p. 146), and since the initial capital income tax wedge drives the private return to saving below the social return in the initial equilibrium, the capital tax reform will generate an aggregate efficiency gain by raising saving and lowering business investment. Figure 5.7 shows that all generations will share in this gain, although to varying degrees. Because older living generations have a larger stock of non-human wealth than younger generations, the older households will gain more than the younger ones from the fall in the capital income tax rate. However, despite their negative financial wealth (mortgage debt), even the youngest living generations will gain life-time utility from the fact that the capital tax reform improves the allocation of resources. Future generations will also gain, and their gain will be larger the further into the future they enter the labour market. This is due to three factors: (1) the slight initial labour market recession generated by the capital tax reform; (2) the fact that the costs of dismantling the existing capital stock cause the efficiency-improving reallocation from domestic real business assets to foreign financial assets to be stretched out over time and – most
158
Lange, Pedersen and Sørensen
importantly – (3) the gradual rise in aggregate wealth which enables the government to gradually increase its lump-sum transfers to households without raising any tax rates.
Effects of the Total Tax Reform Having explained the separate effects of the capital tax reform and the labour tax reform, we can be very brief in reporting the effects of the total tax reform, since the total effects are essentially found by summing the effects of the two former reform elements. Specifically, the welfare curve termed ‘Sum’ in Figure 5.7 is found by simple vertical addition of the curves for the capital tax reform and the labour tax reform, and it is seen to be very close to the welfare curve denoted ‘Total tax reform’ which emerges when both reform elements are simulated simultaneously. It is interesting to note from Figure 5.7 that the total tax reform seems to generate a Pareto-improvement in the sense that all generations experience a welfare gain. It is also worth noting that the distribution of the welfare gains among currently living generations is much more equal under the total tax reform than under any of the two separate components of the reform. In short, the labour tax reform is a ‘pro-young’ policy whereas the capital tax reform is a ‘pro-old’ policy. The combination of the two reform elements seems to strike a reasonable balance between the interests of the young and those of the old. In Lange, Pedersen and Sørensen (1997) we have analyzed the sensitivity of the above results to our choice of strategic parameter values. It turns out that our findings are not crucially dependent on the particular parameter values indicated in Table 5.4. In particular, variations within a realistic range of the implicit labour supply elasticity, the intertemporal elasticity of substitution in consumption, or the elasticity of substitution between different labour skills do not overturn our result that the tax reform generates a Pareto-improvement for all generations. A Pareto-improvement also materializes even if we assume that the initial revenue loss from the reform is recovered through a rise in the rate of VAT rather than through nondistortionary lump-sum taxes.
5.5
CONCLUDING REMARKS
This chapter considered the macroeconomic and intergenerational (nonenvironmental) welfare effects of the Danish 1993 Tax Reform Act. To understand these effects, we found it useful to decompose the total reform into a labour tax reform and a capital tax reform. While the capital tax
The Danish Tax Reform Act of 1993
159
reform element was seen to generate disproportionate welfare gains for the old generations, the labour tax reform was shown to favour the young. In combining these two elements, the total tax reform appears capable of raising the welfare of all current and future generations. Our simulations indicate that future generations will gain relatively more than current generations, so the tax reform may help to redress the generational imbalance which is apparently implied by current fiscal policies in Denmark (see Jensen and Raffelhüschen, 1997). These results do not appear to be crucially dependent on our particular choice of strategic parameter values. However, in order to isolate the structural effects of the tax reform, our simulations assumed that the initial revenue loss from the reform was recouped via higher lump-sum taxes. In reality, the Danish government allowed part of the initial tax cuts to be financed via additional accumulation of government debt. Since government debt tends to redistribute income from future to current generations, there is no doubt that our simulations overstate the gains to future generations resulting from the combination of tax and debt policies actually pursued during the phasing-in of the reform. To put it differently, the Danish government decided to transform part of the potential gains to future generations into gains for the generations currently alive. To allow exact aggregation based on optimizing behaviour at the micro level, the EPRU model relies on strong simplifying assumptions regarding the demographic structure. This is another reason why the results reported in this chapter should not be taken as a literal quantitative estimate of the welfare gains for particular generations currently living in Denmark. Hence one should not overemphasize our finding that the 1993 tax reform seems capable of benefiting all generations. Indeed, the study by Knudsen et al. (1997) – which was based on the more elaborate ‘DREAM’ model of the Danish economy – found that a few groups will tend to lose from the 1993 tax reform. Nevertheless, we were reassured by the fact that the overall effects of the tax reform predicted by the DREAM model are very similar to those described in the present study.
Notes 1.
2.
This chapter is an abbreviated and revised version of Lange, Pedersen and Sørensen (1997). Without implicating them in any remaining errors and shortcomings, we thank Peter Broer and Niels Kleis Frederiksen for valuable comments on the earlier version. The rise in public charges for water use was motivated by the fact that pure sub-soil water is becoming scarce in Denmark. However, the water charges are not differentiated geographically to account for differences in the scarcity of pure water.
Lange, Pedersen and Sørensen
160 3.
4. 5.
6.
7. 8. 9.
The figures in Tables 5.1 and 5.2 reflect the local income tax rate prevailing in an average Danish municipality in 1993. Since then there has been an increase in local government tax rates. At least a part of this increase was probably due to an erosion of the local government tax base following from the tax reform. The present analysis ignores this incentive to raise the local income tax rate. A detailed description of the assumptions and data underlying the estimates in Table 5.3 is provided in Lange (1997). The capital gains of the owners of unincorporated firms will be taxed via the specific taxes on gains on machinery, buildings and intangible assets. It should be stressed that our estimate of the average capital gains tax rate in the EPRU model is rather uncertain. The EPRU model overestimates the revenue loss from the fall in the taxable imputed return on owner-occupied housing by counterfactually assuming that all imputed rents are taxed at the marginal capital income tax rate. The model also seriously exaggerates the 1994 revenue loss resulting from the initial capital losses generated by reform, because the model neglects the various limits to loss offsets embodied in Danish tax law. According to the official government estimate, the revenue loss from the tax reform was only 0.6 per cent of GDP in 1994, compared to our model estimate of more than 2 per cent. For the subsequent years, the difference between the officially estimated revenue losses (which mechanically assumed unchanged economic behaviour) and the losses predicted by the EPRU model is considerably smaller. As a simplification, the EPRU model assumes that all dwellings are owneroccupied. In practice, only about two-thirds of the Danish population are home-owners. This statement obviously assumes free capital mobility and abstracts from possible externalities associated with investment. Paradoxically, the rise in the capital gains tax rate tends to generate a shortterm capital gain on existing shares. This is explained as follows: in equilibrium the shareholder must be indifferent between the two options of (1) receiving an extra krone of dividend and using the after-tax proceeds to buy additional shares, and (2) allowing the corporation to retain an extra krone to generate a rise in the value of his existing shares. When the capital gains tax rate goes up, the latter option becomes less attractive, so to maintain indifference between the two modes of increasing shareholder wealth, option (1) must also be made less attractive through an increase in stock prices.
References Blanchard, O.J. (1985) ‘Debt, deficits and finite horizons’, Journal of Political Economy, 93, pp. 223–47. Hansen, C.T., L.H. Pedersen and T. Sløk (1996) ‘Indkomstskatteprogression, aktivitet og løn: teori og dansk empiri’ (Income tax progresivity, activity and wages: theory and Danish evidence)’, Nationaløkonomisk Tidsskrift, 134, pp. 153–74. Holmlund, B. and A.-S. Kolm (1995) ‘Progressive taxation, wage setting and unemployment’, Tax Reform Evaluation Report, 15, National Institute of Economic Research and Economic Council, Sweden. Jensen, S.E.H. and B. Raffelhüschen (1997) ‘Public debt, welfare reforms, and intergenerational distribution of tax burdens in Denmark’, in A. Auerbach, L. Kotlikoff and W. Leibfritz (eds), Generational Accounting Around the World (Chicago: University of Chicago Press).
The Danish Tax Reform Act of 1993
161
Jensen, S.E.H., S.B. Nielsen, L.H. Pedersen and P.B. Sørensen (1996) ‘Tax policy, housing, and the labour market: an intertemporal simulation approach’, Economic Modelling, 13, pp. 355–82. Knudsen, M.B., L.H. Pedersen, T.W. Petersen, P. Stephensen, and P. Trier (1997) ‘Modelling structural reform: a dynamic CGE analysis of the Danish Tax Reform Act of 1993’, paper presented at the OECD conference on ‘The Effects and Policy Implications of Structural Adjustments in Small Open Economies’ (Amsterdam, 23–24 October). Lange, K. (1997) ‘Implementing the tax reform’, Economic Policy Research Unit, Copenhagen Business School, mimeo. Lange, K., L.H. Pedersen, and P.B. Sørensen (1997) ‘The Danish Tax Reform Act of 1993: effects on the macroeconomy and on intergenerational welfare’, paper presented at the conference on ‘Macroeconomic Perspectives on the Danish Economy’ (Hornbæk, 19–20 June). Smith, N. (1995) ‘A panel study of labour supply and taxes in Denmark’, Applied Economics, 27, pp. 419–29. Sørensen, P.B. (1994) ‘From the global income tax to the dual income tax: recent tax reforms in the Nordic countries’, International Tax and Public Finance, (1) (May), pp. 57–79. Sørensen, P.B. (1997) ‘Public finance solutions to the European unemployment problem?’, Economic Policy, 25 (October), pp. 221–65. Sørensen, P.B. (ed.) (1998) Tax Policy in the Nordic Countries (London: Macmillan Press).
Comment Peter Broer Chapter 5 by Lange, Pedersen and Sørensen is an interesting addition to a substantial body of literature dealing with the evaluation of tax reforms by means of an Applied General Equilibrium (AGE) model. It argues that the 1993 tax reform in Denmark was actually Pareto-improving through the use of a well chosen mix of reductions in marginal tax rates and a broadening of the tax base. It is of course pleasing to learn about the existence, and implementation, of Pareto-improving reforms, but the chapter itself is also a well-written and informative exposition of the basic mechanisms of the reform. The model used is familiar from previous publications and contains a number of potentially relevant extensions of the standard AGE literature. Most AGE studies of tax reforms suffer from the drawback that the model, while internally consistent, describes an economy with perfectly competitive markets, in which the only distortion is the tax levy of the government. This, to most people, does not appear as a realistic description of the economy, as it cannot explain the persistent underutilization of labour resources of the last two decades. In fact, many policy measures address labour market problems in particular. As a result, the conclusions of AGE-based tax reform studies often lack policy relevance. The EPRU model explicitly considers labour market imperfections and as such it appears to be well suited to evaluate the consequences of the recent Danish tax reform, which was geared, among other things, towards a reduction of the roll-off of wage taxes through higher wage claims by unions. The explicit consideration of market imperfections is of course one of the reasons that, according to the model, the tax reform was actually Pareto-improving. The inclusion of a separate housing sector, which is treated quite differently before taxes, and the explicit consideration of finitely lived individuals, which allows for an analysis of population ageing, also add to the realism of the model. Any AGE model that matches the theoretical standards closely must make a large number of simplifying assumptions, and the EPRU model is no exception. In the sequel I shall try to indicate a number of issues where, in my view, a different specification might have been chosen, with possibly different results. As the outstanding characteristic of the EPRU model is the labour market, I will consider this point first. (1) The EPRU model makes a somewhat special assumption about the preferences of households, where the elasticity of substitution between consumption and leisure is effectively infinite. This specification implies that the supply of labour of households depends only on their current real wage. In particular, the supply of labour does not depend on the level of income. This aspect of the model seems debatable. There is substantial empirical evidence that labour supply depends also on current income and wealth – for example, the early retirement decision is strongly affected by pension wealth (see Disney, 1996). Since life-time income of households increases as a result of the tax reforms, life-cycle labour supply effects may therefore be overstated by the model. It would be interesting to also include the standard formulation of household preferences in the model, that treats consumption and leisure symmetrically, and allows for income effects in labour supply.
162
Comment
163
(2) The description of wage formation in the EPRU model deviates somewhat from the standard model, which describes a wage-bargaining process between employers and unions. In this standard model, the objective of the union is slightly different from the one in the EPRU model, because it is formulated in terms of the excess of utility over the fall-back level that would arise if no agreement between both parties could be reached. This small change makes an important difference in the resulting wage equation, because the effect of tax rates on wages vanishes. This absence of real wage resistance in the standard model is empirically disputed, but many studies find that in the long run wage taxes and terms of trade losses are indeed borne exclusively by labour (see Layard et al., 1991). In the EPRU model, on the other hand, a decrease of the wedge lowers the real product wage. This effect may generate a large part of the computed welfare gain of the tax reform. It would therefore be desirable to perform a sensitivity analysis with respect to this aspect of the model as well. (3) The nature of the imperfection in the labour market is, according to the EPRU model, that the labour union for each skill group can organize a closed shop. Individual workers want to work more, but they are prevented from doing so because of the economy-wide control of the unions. This description of the Danish labour market strikes me as rather static. First, there is probably also a substantial non-unionized sector. This sector should be able to accommodate a substantial part of the excess labour supply of the unionized sector. For example in the United States, the role of unions has gradually diminished because of the growing importance of the non-unionized sectors. Second, the assumption of work sharing implicitly assumes that workers stick to the same job for their entire life. This of course, is at variance with the huge amount of job reallocation that is continuously occurring. This job reallocation means that a substantial portion of the working population is negotiating its employment conditions – and, indeed, its skill category – anew each period. There now exist a few CGE models that include a description of job flows (for example, Merz, 1995), and it might be useful to extend the EPRU model in this direction. Third, explicit consideration of the source of the monopoly power of the unions also calls for a dynamic approach. As Caballero and Hammour (1996) argue, the reasons why labour can appropriate part of the return to capital are to be found in the sunkcost character of investment made by the firm, be it in physical capital goods, firm-specific human capital, or search costs of the stock of employees. Tax reforms may alter the degree of appropriability of fixed assets – for instance, a labour tax reform that increases labour supply and search effort of the unemployed reduces the search costs of employers and thereby also the monopoly power of unions. (4) In the chapter only one market imperfection is explicitly considered, the monopoly of unions on the labour market. Of course, imperfections may be considered on all markets – for example liquidity constraints due to asymmetric information on financial markets, or imperfect competition on the goods market due to increasing returns. The latter imperfection seems particularly relevant in the present context, since it affects the conclusions of the chapter with respect to the inefficient taxation of capital income. According to recent OECD estimates (see Martins, Scarpetea and Pilat, 1996), the average mark up in Denmark is around 15 per cent. This mark up distorts the relation between the social cost and the social revenue of capital in the direction of a too low utilization of capital. The same does not hold for labour, as the presence of the unions pushes up the wage above the marginal disutility of work. So one may argue that in particular capital should be subsidized by means of a tax credit or a low tax
164
Broer
rate. The low effective tax rate on capital income may then be interpreted as an attempt by the Danish policy makers to correct for this inefficiency (see Keuschnigg and Kohler, 1996, for an AGE model that makes a similar point). (5) In practical policy making, the predominant issue is often that of the distributional consequences of policy measures. In the EPRU model only intergenerational distribution issues can be studied. The outcome of the model that the reform was Pareto-improving notwithstanding, I therefore wonder whether the 1993 Danish tax reform has met with unanimous approval, especially from the lower-income groups. To study these issues, it would be desirable to abandon the symmetric treatment of the skill classes and explicitly distinguish the skill levels by labour productivity.
References Caballero, R.J. and M.L. Hammour (1996) ‘The macroeconomics of specificity’, NBER Working Paper, 5757. Disney, R. (1996) Can We Afford to Grow Older? (Cambridge, Mass.: MIT Press). Layard, R., S. Nickell and R. Jackman (1991) Unemployment: Macroeconomic Performance and the Labour Market (Oxford: Oxford University Press). Keuschnigg, C. and W. Kohler (1996) ‘Commercial policy and dynamic adjustment under monopolistic competition’, Journal of International Economics, (40), pp. 373–409. Martins, J.O., S. Scarpetta and D. Pilat (1996) ‘Mark-up pricing, market structure, and the business cycle’, OECD Economic Studies, 27, pp. 71–105. Merz, M. (1995) ‘Search in the labour market and the real business cycle’, Journal of Monetary Economics, 36, pp. 269–300.
6 Minimum Wage Contracts and Individual Wage Formation: Theory and Evidence from Danish Panel Data1 Lars Haagen Pedersen, Nina Smith and Peter Stephensen 6.1
INTRODUCTION
The persistent high levels of registered unemployment in Denmark and other European countries remain one of the main challenges both to economic policy and to economic theory. Labour market reforms, tax reforms and other so-called structural reforms are central policy issues in most European countries in the 1990s. The effects of such reforms to a large extend depend upon how the reforms affect the wage determination in the economy. Against this background this chapter derives and estimates wage equations for six major union groups in the Danish labour market. Wage formation depends upon the institutions that are present in the economy in question. In Denmark, two sets of institutions have major effect on wage formation. The two are a general and tax-financed unemployment benefit system and the collective-bargaining system that covers most of the private labour market in Denmark. Wage formation affects unemployment in two very different ways. First the wage may be so low for some low skilled individuals that getting a job is not considered a net welfare gain. These persons are therefore not actively seeking a job and by definition they are voluntarily unemployed. One explanation of the existence of voluntary unemployment in welfare states is that it to some extend is possible to be affiliated with the (tax-financed) unemployment insurance system without actively seeking job. Some low skilled workers may consider this situation at least as good as accepting a low wage job, and thus choose to become voluntarily unemployed. Second, there may exist imperfections in the labour market that make the wage higher than the market-clearing wage. In this case there exist unemployed 165
166
Pedersen, Smith and Stephensen
persons who are willing to work at the going wage and are actively seeking a job since getting job represents a net welfare gain. These persons are called involuntarily unemployed. Voluntary and involuntary unemployment may coexist in the economy, since the two phenomena refer to different segments of the labour market. One view on the Danish labour market – which is not refused by the estimated equations presented here – is that the combination of the institutions of the tax and the benefit systems effectively determine the wage for the lowskilled workers, implying that the net income gain for these persons from being employed is zero. Pedersen and Smith (1996) find that 25 per cent of employed members of unemployment insurance funds gains less than 500 DKK per month by being employed. For workers with higher skills there exists involuntary unemployment for example due to the presence of unions that have a sufficient bargaining power in the wage formation. As mentioned involuntary unemployment is explained by referring to some kind of market failure, typically in the labour market. There exist 3 types of labour market theories that explain the presence of involuntary unemployment. These are search theory, efficiency wage theory and bargaining theory.2 The three theories focus on different elements of imperfections in the labour market and are not mutually exclusive. Search theory explains involuntary unemployment by referring to inefficiencies in the allocation of job offers and unemployed persons. Since there is friction in the labour market a vacant job and an unemployed person is not necessarily matched immediately. The friction implies that unemployment prevail even in the (long-run) equilibrium. Pissarides (1990) gives a rigorous presentation of search theory. Efficiency wage theories focus on situations where it is in the interest of the employer to pay a wage that is higher than the market-clearing wage. The reason is that a wage, which is higher than the market wage, will increase the productivity of the employed labour force (by making people work harder), and therefore it may be in the interest of firms to pay this higher wage. If all firms react to this incentive to increase the wage, then involuntary unemployment will prevail in equilibrium. The mechanism which makes people work harder is either based on a psychological argument called the ‘partial gift exchange’, where the bare fact that the worker receives a higher wage implies that she finds it fair to increase work effort. As another mechanism to generate the same effect Shapiro and Stiglitz (1984) consider the so-called ‘shirking model’, where the worker is assumed to shirk on the job. For a given level of monitoring the firm will observe less shirking by increasing the wage since this increases the cost of being caught and fired. Finally, involuntary unemployment may prevail due to imperfect competition in the labour market. Standard assumptions are that wages are negotiated between confederations of employers and unions of employees.
Minimum Wage Contracts and Individual Wage Formation
167
The negotiations lead to wages that are higher than market-clearing wages if the unions have sufficient bargaining power. Recent Danish policy issues include a restructuring of the unemployment benefit system and of the tax system. A reduction in the level of unemployment benefits after tax will tend to reduce unemployment irrespectively of whether it is voluntary or not. The mechanism at work is that reducing unemployment benefits reduces the welfare of the unemployed. For the voluntarily unemployed this implies that the relative gain from being employed at a given wage rate is increased. In case of involuntary unemployment a reduction in the unemployment benefits also tend to increase employment by moderating the wage rate of the employed. The mechanism behind this effect is an increase in the cost of being unemployed, which tends to reduce unemployment in all the 3 different types of explanations of involuntary unemployment. Politicians in Denmark and most West-European countries are however reluctant to reduce the level of unemployment benefits due to the increase in the income inequality which is expected to be the consequence. To avoid the inequality problem for example the European Commission (1997) and OECD (1997) have proposed to fight the unemployment problem through changes in the tax structure.3 The idea is to reduce the tax burden of low wage incomes while keeping the level of unemployment benefits after tax constant. The financing of the tax reduction is obtained by an increase in the tax burden of the part of the employed that has the highest salaries. Thus the overall effect on the labour income taxation is that progressivity is increased. The idea is that this shift in the tax structure will increase the incentives for low skilled unemployed to become employed. Thus voluntary unemployment may be reduced without reducing the standard of living for the unemployed. Involuntary unemployment may also be reduced since the pre tax wage for the low-income groups may be reduced. Whether this positive effect is outweighed by the increase in the tax burden and the derived increase in the pre-tax wage for high-income groups is a central policy issue. Those in favour of the ‘employment tax cut’ proposal argue that there will be a net reduction in the involuntary unemployment due to the restructuring. The positive net-effect appears because the progressivity of the tax system is increased. The argument is that for given levels of the tax burdens an increase in the progressivity (defined as marginal tax rates divided by average tax rates) will reduce the wage level in the economy. This effect is sometimes called ‘the Robin Hood effect’. However the effects on wage formation of pure increases in progressivity depend on whether the labour market is perfectly competitive or not. In models of perfect labour markets increased progressivity tends to reduce labour supply and therefore reduce activity and increase the equilibrium wage rate. On the contrary standard models of all 3 types of labour market
168
Pedersen, Smith and Stephensen
theories with involuntary unemployment imply that increased progressivity tends to reduce the wage rate. Consider for example the labour market with bargaining over wages. The wage claim of the union is determined such that the sum of the welfare gains to the employed members (due to the increased wages after tax) is equal to the sum of the welfare losses of those members who will become unemployed if the pre-tax wage is raised. The number of members who will become unemployed depends solely on the pre-tax wage. The welfare gain for the employed members depends on the after-tax wages. Therefore high marginal tax rates tend to reduce the welfare gains of the employed members and the union tends to profit by having lower wage claims than in the case of low marginal tax rates. Malcomson and Sator (1987) find this effect of progressive taxation in case of unionized economies. Similar results are obtained in the case of search theory (see Pissauro, 1990, and efficiency wage theory, see Hoel, 1990). As observed by Hansen, Pedersen and Sløk (1995) the reported effect of progressive taxation in labour markets with imperfections ignores the labour supply effect that is responsible for the opposite result in the models with competitive labour markets. They consider a labour market with bargaining over both wages and working hours. The result is that the effect of a pure increase in the progressivity of the tax system is indeterminate. If union bargaining power is weak and the labour supply elasticity is sufficiently high, then an increase in progressivity may increase the wage which is the outcome of the bargain. Given this result it remains an empirical question whether a pure increase in progressivity reduces or increases the wage rates in the economy. Therefore it remains an empirical question whether ‘employment tax cuts’ will in fact reduce involuntary unemployment. In this chapter we deduce and estimate wage equations for persons belonging to six major unions in the Danish labour market. The data used is a panel sample of Danish wage earners covering the period 1980–90. The data is a representative 1 per cent sample of the Danish population. The wage equations are derived in a theoretical model which is an extension of the traditional bargaining model, (see, for example, MacDonald and Solow, 1981), designed to yield a better description of the institutions in the Danish labour market. The main extension to the traditional model is that each union organizes persons with different levels of productivity, such that some workers are more efficient than others. Wages in the economy reflect these differences in productivity. The wage of an individual worker is given as the minimum wage for the specific union and a personal rise. The union indirectly affects the wage level of the worker by its bargaining power in the negotiations over the minimum wage. This leads to individual wage equations that contain both measures of personal productivity and standard opportunity costs of the union.
Minimum Wage Contracts and Individual Wage Formation
169
Our empirical estimations reveal some general tendencies: First of all that variables representing personal productivity get a higher weight in the wage equation for unions representing higher income groups. Furthermore the effect of variables measuring personal productivity is larger for men than for women. Similarly the weight of the unemployment benefits is lower in the wage equations for unions representing higher income groups. Again, there is a difference between the effects in the relation for men and women, where the effect of unemployment benefits on the wage is larger for women than for men. The average tax rate tends to have a significant increasing effect on the wage for all groups. Whereas the effects of marginal tax rate in the case of male workers have a significant wage-reducing effect except for members of the building and construction union. For female workers increases in the marginal tax rate tend to be wage increasing. However, for the large group of unskilled females organized in KAD, the effect is insignificant. The result from Hansen, Pedersen and Sløk (1995), that increased marginal tax rates may increase wages if the union bargaining power is low and the labour supply is sufficiently elastic, seems to be in line with the observation that positive effects of marginal taxes on wages are more concentrated in the estimations for females. Although the effect of an increase in the marginal tax rates tend to be wage-reducing for most groups of male workers, the effect on the pre-tax wage is much less in this study than in previous studies based on macro data.
6.2
THE STRUCTURE OF THE DANISH LABOUR MARKET
The labour market of the private sector in Denmark is dominated by collective-bargaining between the individual Employers’ Associations organized in the Confederation of Danish Employers’ Associations (DA) and the Unions in the Confederation of Trade Unions (LO).4 The DA has rules implying that the contracts negotiated by the individual members are to be approved by a majority of the members of the confederation. A similar formal set of rules does not apply in the LO. The coverage of the employers’ associations measured as employment in organized firms relative to total private employment is 52.7 per cent. The coverage of DA alone is 44.3 per cent. The total union coverage measured as union members relative to total number of wage earners is 88.0 per cent. The coverage of LO alone is 66.6 per cent (Dansk Arbejdsgiverforening, 1998). The coverage of collective bargaining agreements defined as the number of employed in firms where more than 50 per cent of the workforce is covered by collective bargaining agreements relative to the total private employment is 83 per cent (Statistics Denmark, 1997).
170
Pedersen, Smith and Stephensen
Contracts and working conditions for workers who are not covered by the member organizations of DA and LO to a large extend mimic the outcome of the bargaining in the organized labour market. In the organized part of the labour market there exists the following types of non-piecework contracts with provisions concerning wages: (1) A standard wage-contract, where the wage of a given category of workers is set in the contract. (2) A Minimum wagecontract, where union–employer negotiations determine the minimum wage. The total wage of a given employee is given as the sum of the minimum wage and a personal rise. (3) A so-called minimum payment-contract, where there is no direct connection between the minimum wage and the total wage of a given employee, except for the fact that the wage has to be at least as high as the minimum wage. From 1961 to 1989 the coverage of the different types of contracts was almost constant with the standard wage contract covering approximately 50 per cent of the non-piecework employment covered by contracts between the DA and the LO member organizations. Minimum payment-contracts and minimum wage-contracts covered approximately 40 per cent and 10 per cent respectively. From 1990, there has been a rather dramatic increase in the coverage of non-standard wage-contracts. By 1997 the coverage of the standard wage-contract was reduced to only 16 per cent. 67 per cent of total contracts are either of the minimum wage or the minimum payment type, whereas 17 per cent of the contracts do not have sections on wages (see Nicolaisen, 1997; Dansk Arbejdsgiverforening, 1998). Unfortunately our data set only allows us to estimate wage equations for the period from 1980 to 1990, where approximately 50 per cent of the organized labour market was covered by non-standard wage contracts. In Figure 6.1 below the distribution of wages in 1981 and 1990 are shown for the unions that are represented in the estimated wage equation at the end of the chapter. Observe that wage dispersion tends to increase with the length of the education needed to require the skill in question. Observe also that even if 50 per cent of the organized labour market use a standard wage contract there does not seem to be a single wage that is dominating in the distributions.
6.3
THE THEORETICAL MODEL
Decentralized Bargaining In this section we discuss a theoretical model of decentralized wage bargaining between unions and employers federations. A formal derivation of the theoretical model is found in Pedersen, Smith and Stephensen (1998). The assumption of decentralized bargaining implies that the parties of the
Minimum Wage Contracts and Individual Wage Formation
171
Figure 6.1 Frequency of wage rates. Wage rates of 1990 are deflated by consumers price index SID, male Union of unskilled workers (SID), male 0.12
Frequency
0.10 0.08
Freq 81 Freq 90
0.06 0.04 0.02 0.00 50
70
90
110
130
150
170
190
210
230
250
270
Wage
METAL, male(METAL), male Union of metal workers 0.12
Frequency
0.10 0.08
Freq 81 Freq 90
0.06 0.04 0.02 0.00 50
70
90
110
130
150
170
190
210
230
250
270
290
Wage
Union of clerical workers (HK), male HK, male 0.12
Frequency
0.10 0.08
Freq 81
0.06
Freq 90
0.04 0.02 0.00 50
70
90
110
130
150
170
Wage
190
210
230
250
270
290
Pedersen, Smith and Stephensen
172 Figure 6.1
(Cont’d) BYG, male Unions of building and construction workers (BYG), male
0.12
Frequency
0.10 0.08
Freq 81
0.06
Freq 90
0.04 0.02 0.00 50
70
90
110
130
150 170
190
210
230
250
270
290
Wage
AC, male Unions of academic workers (AC), male 0.08 0.07
Frequency
0.06 0.05
Freq 81 Freq 90
0.04 0.03 0.02 0.01 0.00 50
70
90 110 130 150 170 190 210 230 250 270 290 310 330
Wage
SID, female Unions of unskilled workers (SID), female 0.14 0.12
Frequency
0.10 0.08
Freq 81 Freq 90
0.06 0.04 0.02 0.00 50
70
90
110
130
150
170
Wage
190
210
230
250
270
Minimum Wage Contracts and Individual Wage Formation Figure 6.1
173
(Cont’d) female Union of unskilledKAD, female workers (KAD), female
0.16 0.14
Frequency
0.12 0.10
Freq 81
0.08
Freq 90
0.06 0.04 0.02 0.00 50
70
90
110
130
150
170
190
210
230
250
270
Wage
HK, workers female (HK), female Union of clerical 0.10 0.09 0.08
Frequency
0.07 0.06
Freq 81
0.05
Freq 90
0.04 0.03 0.02 0.01 0.00 50
70
90
110 130 150 170 190 210 230 250 270 290 310
Wage
collective bargaining consider the macroeconomic variables as independent – that is, unaffected of the outcome of the bargaining process. This assumption is admittedly questionable with respect to the Danish labour market institutions. First, one may argue that the number of parties performing collective bargaining is not sufficiently large to assume that it is rational for each bargaining unit to ignore the effect on the macro variable of the outcome. Second, because of the rule in the DA that each contract has to be approved by a majority of the employers federations one may argue that at least on the employer side, there is a tendency to take the effects on the macro-economy into account. By assuming decentralized bargaining (and Nash equilibrium) one introduces a too large incentive for the specific union to increase the wage. There are three reasons for this: first, the union ignores the effect on
174
Pedersen, Smith and Stephensen
the overall price index and therefore overestimates the positive effect on the real wage of the employed. Second, the union underestimates the negative consequences on the aggregate demand for labour and therefore also on the demand for the kind of labour organized in the union. Third, the union ignores the increase in government expenditures that follows from the unemployment benefits paid to the members of the union. Effectively, the union sets the cost of these marginal funds to zero for its own members. Minimum Wage Contracts and Different Productivity Levels Standard theoretical models with imperfect labour markets do not allow for distributions of wages between different workers working within the same collective bargaining area. In bargaining models unions and employers organisations negotiate a unique wage rate for a given part of the labour market. In this chapter we want to explain the existence of wage distributions within a given part of the labour market covered by a single collective bargaining agreement. As our data does not allow us to discriminate between types of wage-contracts, we assume that the minimum wage contract may describe the data. Observe that the standard wage contract is a special case of a minimum wage-contract where the personal rise is identical for all workers. In the specification we use, this corresponds to a situation where the productivity of all workers within a specific union is identical. Therefore we need only to consider the more general case where unions set minimum wages. From Figure 6.1 one observes that the less skilled the type of worker the lower is the variation in the distribution of wages – that is, the unskilled worker differ less in productivity than skilled workers and especially academics. We assume that the productivity of individual workers differs and that this difference is observable both to the union and the employers. Thus the firms know the productivity of a given worker in advance and on this basis the firm decides whether or not to employ the worker in question. We assume that the firms have ‘right to manage’ such that for a given wage rate the employment decision is made unilaterally by the firm. This implies that the minimum wage set in the bargaining determines a minimum level of productivity, which is required to beome employed. All workers who meet this requirement are employed. The salaries of the workers, who have a productivity level which is higher than the required level, is determined as marketclearing wages subject to the minimum wage. Thus unemployment in the economy is entirely due to lack of productivity relative to the minimum wage. The result of the theoretical model that for all groups, where the individual productivity is sufficiently high, a non-zero unemployment rate is observed, is of course too simplistic. The effect appears in the theoretical model because we have ignored the market imperfections described in for example search theory.
Minimum Wage Contracts and Individual Wage Formation
175
The Outcome of the Bargaining over Minimum Wages Each union chooses its minimum wage claim to maximize the sum of utilities of the members. The incentive for the union to raise the minimum wage is that this raises the total wage of those who become employed and this may outweigh the loss due to the unemployment that is the consequence of the minimum wage. The increase in the total wage of the employed appears because the productivity of a given worker depends both on her individual productivity and on the overall level of employment in the economy. The mechanism from total employment to individual productivity is that increased employment reduces the capital–labour ratio, which reduces productivity of the individual worker. The optimal minimum wage is given as a mark up on the opportunity cost of work as perceived by the union. The opportunity cost is equal to the gross welfare loss of an unemployed worker who becomes employed. This loss is the sum of the unemployment benefits corrected for effects of the tax system and the money value of the amount of leisure the worker has to give up by taking a job. This result concerning the minimum wage in our setting where labour is heterogeneous is analogous to the result concerning the total wage in a rightto-manage union model with standard wage contracts. This is so since the standard wage-contract is a special case of the minimum wage-contract. Observe that, since the minimum wage determines the level of unemployment in the economy, any reform that reduces the opportunity cost of the union will tend to increase employment. Thus tax reforms, reforms of the unemployment benefit system and reforms that affect the welfare effect of the amount of leisure available to the unemployed may have a positive effect on employment in this model. Determination of the Wage The total wage is determined as the sum of the minimum wage and a personal rise, which is set such that the productivity corrected wage is the same for all workers. As the workers with lowest productivity are laid off first, it follows that a high rate of unemployment means that the marginally employed worker has higher productivity than if unemployment is lower. Therefore the personal rise of a given employed worker is a decreasing function of the level of unemployment within the specific union. Finally the tax system matters for the level of the pre-tax wage. First, an increase in the average tax rate increases the product wage. This effect is due to the non-competitiveness of the labour market. There would be no effect present if the labour market had been competitive, since (by assumption of the theoretical model) no income effect is present in the labour
176
Pedersen, Smith and Stephensen
supply. The effect appears because the alternative costs of being employed depend upon the benefits per hour net of tax adjusted to the allowed tax deduction in labour income for employed workers. The partial effect of an increase in the marginal tax rate for a given level of the average tax rate may be both positive and negative. This is in contrast to the results of, for example, Lockwood and Manning (1993), who under quite general conditions show that if the wage is an interior solution to the Nashbargaining problem of the right-to-manage type, then an increase in the marginal tax rate will reduce the wage. The feature that distinguishes this result from Lockwood and Manning’s result is that the number of working hours is subject to bargaining. This feature re-introduces a variant of the labour supply effect of the competitive models of the labour market. The result is that an increase in the marginal tax rate increases the pre tax wage if the labour supply is sufficiently elastic, since in this case the labour supply effect may dominate the bargaining effect and the result becomes a variant of the competitive case.
Determinants of the Wage Rate This completes the description of determination of the wage rates in the economy. To sum up we found that the wage rate is determined by the following variables: ● ● ● ● ● ●
personal productivity (positive effect) average labour income tax rate (positive effect) marginal labour income tax rate (positive or negative effect) consumer price index (positive effect) potential unemployment benefits net of tax (positive effect) unemployment level in the union (negative effect).
The inclusion of the personal productivity in our view improves the realism of the model and also ‘solves’ one of the main criticisms of the so-called union model where the wage is determined in a negotiation between the union and the employers’ federation. As mentioned the standard right to manage model implies that the wage is set as a mark up on the opportunity cost of work as perceived by the union. This result leaves little room for effects of productivity increases on the wage. Our approach implies that increased personal productivity – for example, due to increased human capital – will increase the actual wage without affecting the negotiated minimum wage. Thus the present model embeds both the standard result of labour union theory that wages reflect the opportunity cost of the union and the stylized empirical fact that increases in wages reflect the increases in productivity of the workers.
Minimum Wage Contracts and Individual Wage Formation
177
Furthermore we would like to stress that the present modelling of union behaviour is neither based on insider–outsider behaviour of the union nor stochastic rationing of the union members. The fact that bargaining is over minimum wages and that heterogeneity is observed implies that the union in principle is aware of the fact that the part of the members who has the lowest productivity is rationed in their labour supply. This has some of the same implications as the insider–outsider model except for the fact that the present model allows for hiring of outsiders who are sufficiently productive to meet the requirements of the minimum wage. This seems to be consistent with the fact that flows from unemployment to employment are relatively large in the Danish economy (see Daugaard, 1997).
6.4
DATA AND EMPIRICAL RESULTS
The sample used is a panel sample of Danish wage earners covering the 11-year period 1980–90, but since lagged variables are included in the analysis, the estimation period is 1981–90. The sample is a sub-sample of the Danish longitudinal database which is a representative 5 per cent sample of the Danish population. Self-employed persons and assisting wives have been excluded from the sample because reliable wage data are available only for wage earners. Further, the analysis is restricted to workers employed in the private sector, since we expect the wage formation in the public sector to be generated by other (political) forces than in the private sector. We do not have explicit information on the type of wage-contract or union-membership for each individual. However, we have information on which unemployment insurance fund the individual belongs to. In Denmark unemployment insurance (UI) funds are administered by the unions. Thus membership of a given UI fund is closely related to being member of a given union. The data includes information on seven union (UI fund) groups: unskilled workers (SID); female union (KAD); Metal Workers (METAL); Clerical workers (HK); Building and Construction workers (BYG); and Academic workers (AC); and all other unions. The sample is restricted to members of these union groups, but the very heterogeneous group of ‘all other unions’ is excluded. The sample is an unbalanced panel, where (mainly young) people enter the sample over time and (mainly older) people leave the sample. Only observations with an observed hourly wage rate are included in the analysis. This means that people who leave the labour force or people who are unemployed for the whole year are excluded from the estimations. This may give rise to a self-selectivity problem, which is often handled by the traditional Heckman selectivity correction. However, we do not correct for the potential selectivity problem in this study.
178
Pedersen, Smith and Stephensen
The hourly wage rate is not observed directly but has to be constructed from annual wage income, divided by working hours, which is calculated from the register on supplementary pension payments (ATP). Annual earnings include overtime pay and to the extent that overtime hours vary over time (or business cycles) variations in the hourly wage rate may also reflect variations in overtime work. The wage rate includes holiday payments, but does not include pension payments, which the employer pays to a pension scheme. Since the growth of labour market pension schemes did not take place until late in the 1980s, we do not expect that the observed wage rates during the 1980s (contrary to the 1990s) are influenced much by employers’ pension payments. The marginal and average tax rates (tm and ta) are calculated for each individual on the basis of information on taxable income, and other income information which are used by the tax authorities when they calculate the tax amounts to be paid. Since the information is based on the same administrative registers as the registers used by the tax authorities, we expect the information on marginal and average tax rates to be quite reliable. The wage income of the individual is considered marginal to all other income. Thus, the marginal tax rate is the marginal tax rate on marginal income for the person after taxes on positive or negative capital income and other sources of income taxes have been paid. The average tax rate is calculated as the average tax rate paid on the individual’s observed wage income. The aggregated unemployment rate is measured by the unemployment rate in the union, which the person belongs to. This variable is based on public statistics covering the whole labour market, including the public sector. The unemployment insurance compensation is measured by the potential hourly compensation rate net of taxes. The potential compensation rate is calculated as 90 per cent of the individual’s wage rate up to a maximum, which corresponds to the maximum compensation in the UI scheme for each year. The tax rate on UI payments is calculated as the average tax, which the individual would have to pay if he or she received no wage income and was unemployed for the whole year. All variables are measured in nominal terms. Therefore, we include the consumer price index as an explanatory variable in the wage regression. Furthermore, we include different human capital variables in order to capture individual variation in productivity and productivity growth. The traditional variables to include are experience, experience squared, and education. Experience squared is entered into the estimation to allow for a non-linear relationship between wage and experience. One would expect a negative sign of the squared experience coefficient reflecting a decreasing marginal importance of experience in the determination of wages. As we estimate a fixed effect transformation, the time invariant variable for length of education disappears from the regression. Beside these variables the
Minimum Wage Contracts and Individual Wage Formation
179
lagged individual unemployment variable is included in order to reflect individual human capital effects, either because previous unemployment imply depreciation on human capital or because of productivity signal effects to the employer. The variables representing the lagged individual unemployment rate and the aggregated union unemployment rate are measured on a scale from 0 to 1. Since we exclude individuals who are unemployed for the whole year in our analysis because we have no wage information for these individuals, the aggregated unemployment rate in a given union is higher than the average individual unemployment rates in the union. Sample means for the variables used in the estimations are shown in Table 6.1 for each union group. The sample means concern the year 1990. The explanatory variables, marginal and average tax rates, are endogenous since the tax rates depend on the wage level of the individual. Following the considerations in section 6.3, the same holds for the aggregated union unemployment rate. In order to take care of this problem, these variables are instrumented. The instruments used are: Age, age squared, length of education, other non-wage income, negative wealth, positive wealth, indicators for residence in the province, marriage, ownership of house and year-specific indicators. We regress the calculated average and marginal tax rates on the instruments in an OLS (pooled cross-sections). Since the union unemployment rate does not vary across unions, by definition, and since the explanatory power of the variables used as instruments for the tax variables is extremely low, the aggregated union unemployment variable has been instrumented slightly differently. Instead of year specific indicators, we have used a trend variable (and second- and third-order polynomials of the trend variable), and further, the lagged union unemployment rate has been added as an instrument. In order to test for the validity of each of the instruments we have regressed the calculated residual term from the wage equation on the instruments. If the coefficient of an instrument is significant in this residual regression, the instrument may not be valid. As it is shown in Table 6.2, most of the instruments pass this test but there seem to be minor problems with the age variables. In a number of other experiments we have tried different specifications of the model, mainly by varying the price and productivity variables included in the wage regressions and experimenting with the instruments used. If (average) productivity is included in the wage regression, the regressions become extremely unstable with respect to the size and sign of the other included variables in the wage equation. As an alternative to the fixed-effect estimation we have estimated the model in one year differences. As expected, this does not change the results notably, and since the fixed-effect estimation is exploiting the panel information in the sample more intensively and efficiently, we prefer the fixed effect transformation.
180 Table 6.1
Sample means in 1990, standard errors in parentheses
Men
Age
Unskilled (SID)
38.209 (11.714)
Metal
Clerical Building Academic (HK) and (AC) Constr.
36.609 35.160 36.310 (10.848) (11.838) (10.299)
38.280 (9.167)
Other income/DKK100,000
0.083 (0.162)
0.089 (0.190)
0.080 (0.202)
0.121 (0.168)
0.220 (0.370)
Negative wealth/DKK100,000
1.895 (1.940)
2.164 (2.028)
2.209 (2.253)
2.375 (2.113)
3.963 (2.907)
Positive wealth/DKK100,000
2.553 (2.557)
2.994 (2.680)
2.733 (2.808)
3.168 (2.726)
4.761 (3.471)
Province (0/1)
0.795 (0.404)
0.761 (0.427)
0.706 (0.456)
0.748 (0.434)
0.428 (0.495)
Years of education
9.334 (2.094)
11.256 (1.607)
11.578 (1.805)
11.536 (1.405)
16.656 (2.028)
Married (0/1)
0.480 (0.500)
0.478 (0.500)
0.422 (0.495)
0.496 (0.500)
0.548 (0.498)
Owner of house (0/1)
0.571 (0.495)
0.636 (0.481)
0.544 (0.499)
0.679 (0.467)
0.737 (0.441)
15.879 15.181 (9.480) (10.976)
14.392 (7.879)
13.047 (8.812)
Experience
16.691 (11.237)
Hourly wage, DKK
123.773 128.322 119.421 132.786 206.642 (32.686) (28.390) (41.127) (34.958) (59.787) 4.021 4.037 4.020 4.050 4.113 (0.147) (0.156) (0.160) (0.165) (0.192)
Log hourly potential compensation rate after tax, DKK Lagged individual 0.082
0.032
0.041
0.077
0.026
(0.162)
(0.101)
(0.137)
(0.148)
(0.094)
0.160 (–)
0.071 (–)
0.091 (–)
0.110 (–)
0.124 (–)
Average tax rate, tave
0.377 (0.086)
0.370 (0.081)
0.369 (0.083)
0.373 (0.086)
0.448 (0.089)
Marginal tax rate, tm
0.585 (0.057)
0.595 (0.057)
0.589 (0.067)
0.572 (0.029)
0.673 (0.048)
1558
866
419
655
407
unemployment Union unemployment rate
Number of individuals in 1990
181 Table 6.1
(Cont’d) Unskilled (SID)
Female union (KAD)
37.909 (11.909)
38.746 (11.886)
36.530 (11.341)
Other income/DKK100,000
0.083 (0.188)
0.068 (0.137)
0.066 (0.185)
Negative wealth/DKK100,000
1.076 (1.636) 1.142 (1.831)
1.081 (1.585) 1.182 (1.894)
1.230 (1.682) 1.463 (1.993)
Province (0/1)
0.801 (0.400)
0.778 (0.416)
0.627 (0.484)
Years of education
9.307 (2.083)
8.879 (1.914)
11.086 (1.810)
Married (0/1)
0.494 (0.501)
0.574 (0.495)
0.488 (0.500)
Owner of house (0/1)
0.301 (0.460)
0.320 (0.467)
0.362 (0.481)
Experience
10.981 (7.084)
11.360 (7.332)
12.164 (6.913)
103.444 (18.861)
108.396 (25.170)
105.116 (33.674)
Log hourly potential compensation rate after tax, DKK
3.930 (0.122)
3.951 (0.115)
3.933 (0.145)
Lagged individual unemployment
0.095 (0.170) 0.309 (–)
0.087 (0.169) 0.196 (–)
0.038 (0.119) 0.126 (–)
Average tax rate, tave
0.390 (0.086)
0.364 (0.087)
0.365 (0.083)
Marginal tax rate, tm
0.556 (0.037)
0.551 (0.042)
0.558 (0.054)
176
397
1214
Women Age
Positive wealth/DKK100,000
Hourly wage, DKK
Union unemployment rate
Number of individuals in 1990
Clerical (HK)
182 Table 6.2
Test of instruments, dependent variable residual error term, standard errors in parentheses
Men Age Age squared/100 Other income/DKK100,000 Negative wealth/DKK100,000 Positive wealth/DKK100,000 Province (0/1) Years of education Married (0/1) Owner of house (0/1) R2 Women Age Age squared/100 Other income/DKK100,000 Negative wealth/DKK100,000 Positive wealth/DKK100,000 Province (0/1) Year of education Married (0/1) Owner of house (0/1) R2
Unskilled (SID)
Metal
Clerical (HK)
–0.001 (0.001) 0.000 (0.001) 0.027 (0.009) 0.001 (0.001) –0.000 (0.001) 0.001 (0.003) –0.000 (0.001) –0.003 (0.003) –0.007 (0.004) 0.002
–0.008 (0.001) 0.009 (0.001) –0.001 (0.001) –0.001 (0.001) 0.001 (0.001) 0.001 (0.003) –0.000 (0.001) –0.005 (0.003) –0.004 (0.004) 0.017
–0.004 (0.001) 0.004 (0.002) –0.002 (0.012) –0.001 (0.001) 0.004 (0.001) –0.001 (0.004) –0.000 (0.001) –0.004 (0.005) –0.016 (0.006) 0.010
Unskilled (SID) 0.004 (0.003) –0.005 (0.003) –0.033 (0.024) 0.002 (0.004) 0.004 (0.003) 0.012 (0.008) –0.000 (0.002) –0.012 (0.008) –0.021 (0.012) 0.008
Building Academic Constr. (AC)
Female union (KAD) –0.002 (0.002) 0.001 (0.002) 0.068 (0.019) 0.010 (0.003) –0.003 (0.003) –0.006 (0.005) –0.002 (0.001) 0.005 (0.005) –0.022 (0.009) 0.009
–0.008 (0.001) 0.008 (0.002) –0.004 (0.012) 0.000 (0.001) 0.002 (0.001) 0.001 (0.004) 0.000 (0.001) –0.001 (0.004) –0.012 (0.006) 0.011
0.000 (0.002) –0.000 (0.002) 0.006 (0.006) 0.001 (0.001) 0.001 (0.001) 0.002 (0.004) 0.000 (0.001) 0.006 (0.005) 0.006 (0.008) 0.005
Clerical (HK) –0.001 (0.001) 0.000 (0.001) 0.014 (0.010) 0.001 (0.002) –0.001 (0.001) –0.006 (0.003) 0.001 (0.001) 0.011 (0.003) –0.003 (0.005) 0.002
Minimum Wage Contracts and Individual Wage Formation
183
The estimated wage equation has the form log = t + xit + zi + ε it
(6.1)
where xit and zi are vectors of explanatory time-varying and time-constants variables, respectively, and (are vectors of parameters, and i is an individual-specific constant term which captures unobserved time constant heterogeneity. The fixed-effect transformation, given by log it − log i = ( xit − xi ) + ε it − ε i
(6.2)
where xi is the individual-specific average of xit, eliminates the individual specific intercepts as well as the time-constant variables.
6.5
RESULTS
The results from the fixed-effects estimations of wage functions for separate gender and union categories are shown in Table 6.3. In other empirical studies of human capital functions (see, for example, Rosholm and Smith, 1996), it is usually found that the coefficients to the experience and experience squared variables are positive and negative, respectively, and numerically larger for men than women. The results in Table 6.3 are in line with the results found in other empirical studies. The coefficient of the experience-squared variable is significantly negative for men, but insignificant for women. However the career profile may be captured relatively poorly because we do not have information on firm-specific tenure, promotion, etc. Therefore it might be suspected that career effects are also partly captured by the consumer price index variable, log CPI, if variation in this variable is positively correlated with individual wage increases. A priori the coefficient to the CPI variable should be close to one, but for the academic union, the coefficient is significantly larger than one. This may indicate that individual wage increases over the career profile during the observation period have been captured by the macro variable for the consumer prices. The CPI coefficient is significantly smaller than 1 for the female unions. Lagged individual unemployment generally has a negative influence on the wage rate. The general impression from Table 6.3 is that the union unemployment rate is dominating the individual unemployment rate in the male unions, but still individual unemployment seems to be a strong negative signal or have negative human capital effects for academic and clerical male workers. For women the negative effects of individual
184 Table 6.3
Fixed-effect estimation of wage function, dependent vaiable log hourly wage rate, standard errors in parentheses
Men
Experience Experience squared/100 Log CPI Log hourly potential UI compensation after tax, DKK Lagged individual unemployment [0, 1] Union unemployment [0, 1] Log (1 – tave) Log (1 – tm) R2 Number of observations Explanatory power, instrument regression R2, log (1 – tave) regression R2, log (1 – tm) regression R2, union unemployment regression Women Experience Experience squared/100 Log-CPI Log hourly potential UI compensation after tax, DKK Lagged individual unemployment [0, 1]
Unskilled (SID)
Metal
Clerical (HK)
Building Academic and (AC) const.
0.016 (0.002) –0.027 (0.003) 0.752 (0.044)
0.018 (0.003) –0.036 (0.004) 0.952 (0.063)
0.049 (0.004) –0.076 (0.005) 0.929 (0.087)
–0.000 (0.004) –0.036 (0.005) 1.100 (0.070)
0.046 (0.004) –0.100 (0.006) 1.397 (0.091)
0.461 (0.018)
0.307 (0.022)
0.352 (0.027)
0.376 (0.026)
0.212 (0.025)
–0.007 (0.011) –0.403 (0.110) –0.435 (0.032) 0.185 (0.083) 0.58 14335
–0.036 (0.017) –0.632 (0.190) –0.359 (0.044) 0.426 (0.106) 0.67 7697
–0.150 (0.028) –1.451 (0.286) –0.197 (0.056) 0.388 (0.108) 0.70 4085
0.024 (0.018) –3.065 (0.290) –0.164 (0.042) –0.173 (0.078) 0.63 6233
–0.159 (0.032) –0.387 (0.099) –0.270 (0.040) 0.498 (0.078) 0.79 2987
0.42 0.14
0.55 0.19
0.51 0.31
0.51 0.24
0.53 0.31
0.87
0.73
0.83
0.76
0.93
Unskilled (SID)
Female union (KAD)
Clerical (HK)
0.005 (0.010) 0.010 (0.017) 0.397 (0.147)
–0.001 (0.005) 0.011 (0.009) 0.794 (0.084)
0.002 (0.003) –0.008 (0.006) 0.304 (0.047)
0.618 ((0.061)
0.476 (0.037)
0.935 (0.019)
–0.099 (0.038)
–0.007 (0.025)
–0.085 (0.018)
Minimum Wage Contracts and Individual Wage Formation Table 6.3
Women Union unemployment [0, 1] Log (1 – tave) Log (1 – tm) R2 Number of observations Explanatory power, instrument regression R2, log (1 – tave) regression R2, log (1 – tm) regression R2, union unemployment regression
185
(Cont’d)
Unskilled (SID)
Female union (KAD)
Clerical (HK)
–0.281 (0.573) –0.864 (0.114) –0.366 (0.189) 0.61 1109
0.819 (0.433) –0.651 (0.068) 0.201 (0.163) 0.57 3419
–0.467 (0.179) –0.947 (0.042) –0.138 (0.073) 0.64 10339
0.29 0.18 0.73
0.25 0.11 0.64
0.35 0.23 0.75
and union unemployment are mainly pronounced for the very heterogeneous group of clerical workers, while the unemployment variables are insignificant for the female union (KAD). In all union groups, male as well as female, the log hourly compensation rate of the unemployment insurance scheme has a highly significant and positive effect. The effect is considerably larger for the female union groups. This result may indicate that the unemployment compensation rate, which in Denmark is 90 per cent of the wage rate for low-wage groups (but considerably lower for high wage groups due to the relatively low flat level) works like a floor on the wage distribution. Since a larger fraction of the women in the three union groups included in this study are situated in the lower part of the wage distribution, this may be an explanation of the large positive coefficient of the compensation rate. In all gender and union groups the coefficient of the average tax variable (1–ta) is significantly negative. Thus, an increase of the average tax rate tends to increase the wage level in line with the predictions of the theoretical model. However, the result is a rather uncontroversial finding since this is in agreement with most theoretical models based on either union behaviour, search theory, efficiency wage setting or competitive labour markets. The results in Table 6.3 show that the empirical results concerning the effect of the marginal tax rate is not unambiguous. In all male union groups except building and construction, the coefficient of (1 – tm) is significantly positive in line with the union model. In the three female groups the
186
Pedersen, Smith and Stephensen
coefficient is insignificant. Finally, the coefficient is significantly negative for male workers of the building and construction union, indicating either a competitive labour market or an imperfect labour market dominated by strong labour supply effects for these groups. Thus, the union model is mainly confirmed for the male unions while the female unions are either dominated by labour supply effects of marginal tax changes or a more competitive labour market. It is not obvious that the union power should differ between men and women in the unions which organize both gender, for example unskilled workers’ union (SID) and clerical workers’ union (HK). In contrary it is more plausible that the results in Table 6.3 are driven by differences in labour supply elasticities. In recent Danish empirical studies of labour supply functions it is found that the supply elasticity (compensated and uncompensated substitution elasticities) is considerably higher for women than for men, and the elasticity is decreasing with income, implying that low wage groups have the highest labour supply elasticities (see Graversen and Smith, 1997). The female union groups included in this study are all traditional low wage groups, and therefore the results in Table 6.3 may reflect a relatively strong labour supply effect, dominating the union effect from changes of the marginal tax rate. For workers in the building and construction union this explanation probably does not apply. This union group is a relatively well paid group. On the other hand the building and construction union is a special group concerning its wage and employment sensitivity to the general business cycle. For this union there is a very high correlation between the business cycle and the union’s unemployment level and the wage drift component of wage increases. Thus, the empirical results found in this study seem to give a more ambiguous picture compared to earlier Danish empirical research based on macro data presented in Hansen, Pedersen and Sløk (1995) and Lockwood, Sløk and Tranæs (1995) concerning support for the view that progressive taxes may be wage-moderating. This result seems to hold only in some of the Danish unions and mainly for the male labour market. Notes 1. 2. 3. 4.
The authors would like to thank our discussants Donald Parson and Jan Rose Sørensen, as well as Troels Østergaard Sørensen for useful comments on an earlier draft of the paper. In addition to these theories involuntary unemployment may be explained by imperfections in the goods market (see, for example, Silvestre, 1996). In Denmark the Economic Council has put forward a similar proposal (see Economic Council, 1997). Except for a very minor Christian Union all unions of blue collar workers in Denmark are organized in the LO.
Minimum Wage Contracts and Individual Wage Formation
187
References Danmarks Statistik (1997) ‘Overenskomstundersøgelsen 1995’, Nyt fra Danmarks Statistik, 225 (Copenhagen) (Statistics Denmarks, 1997) ‘Collective bargaining agreements 1995’, in Danish)). Dansk Arbejdsgiverforening (1998) Arbejdsmarkedsrapport 1998 (Copenhagen) (Cooperation of Danish Employers’ Associations (1998) ‘Labour market report 1998’, in Danish). Daugaard, S. (1997) ‘Labour market flexibility at micro and macro level – experience from Denmark’, proceedings of the international workshop on The Effects and Policy Implications of Structural Adjustment in Small Open Economies. Economic Council (1997) The Danish Economy, Autumn 1997 (Copenhagen) (in Danish). European Commission (1997) Growth, Competitiveness, Employment: The Challenges and Ways forward into the 21st Century (Brussels). Graversen, E. and N. Smith (1997) ‘Labour supply, overtime work and taxation in Denmark’, Paper presented at the EALE conference (Aarhus). Hansen, C.T., L.H. Pedersen, and T. Sløk (1995) ‘Progressive taxation, wages, and activity in a small, open economy’, EPRU Working Paper, 1995–21. Hoel, M. (1990) ‘Efficiency wages and income taxes’, Journal of Economics, 51(1), pp. 89–99. Lockwood, B. and A. Manning (1993) ‘Wage setting and the tax system. Theory and evidence for the United Kingdom’, Journal of Public Economics, 52, 1–29. Lockwood, B., T. Sløk and T. Tranæs (1995) ‘Progressive taxation and wage setting: some evidence for Denmark’, EPRU Working Paper, 1995–20. MacDonald, I. and R. Solow (1981) ‘Wage bargaining and employment’, American Economic Review, 71, pp. 896–908. Malcomson, J.M., and N. Sartor (1987) ‘Tax push inflation in a unionized labour market’, European Economic Review, 31, pp. 1581–96. Nicolaisen, S. (1997) ‘Individuel løn vinder frem’, (Individual wages in progress), Arbejdsmarkedspolitisk Agenda, (DA), (Copenhagen) (in Danish). OECD (1996) Employment Outlook (Paris: OECD). Pedersen, L.H., N. Smith, and P. Stephensen (1998) ‘Individual wage formation based on minimum wage contracts: theory and evidence from Danish panel data’, Economic Modelling Working Paper Series, 1998:5, Statistics Denmark. Pedersen, P.J. and N. Smith (1995) ‘The welfare state and the labour market’, Working Paper 95–17, Center for Labour Market and Social Research, University of Aarhus and Aarhus School of Business. Pisauro, G. (1991) ‘The effect of taxes on labour in efficiency wage models’, Journal of Public Economics, 46, pp. 329–45. Pissarides, C.A. (1990) Equilibrium Unemployment Theory (Oxford: Blackwell). Rosholm, M. and N. Smith (1996) ‘The Danish gender wage gap in the 1980s: A panel data study’, Oxford Economic Papers, 48, pp. 254–79. Shapiro, C. and J. Stiglitz (1984) ‘Equilibrium unemployment as a discipline device’, American Economic Review, 74, pp. 433–44. Silvestre, J. (1993) ‘The market-power foundation of macroeconomic policy’, Journal of Economic Literature, 31, pp. 105–41. Statistics Denmark (1997) ‘Collective Bargaining Agreements 1995’ (in Danish) (Danmarks Statistik (1997) Overenskomstundersøgelsen 1995, Nyt fra Danmarks Statistik, 225 (Copenhagen).
Comment Donald O. Parsons The determinants of individual wages in Denmark and in other highly syndicalized labour markets remain poorly understood. In the United States, where wages are largely set by market forces, the human capital model fits the data well – wages are systematically higher where schooling and training investments are higher – but this is less true in the union sector. A complete wage formation model of the US economy will require that we first explain which jobs are unionized, which not, and then answer key union distributional questions, including (1) how much economic rent (wage payments in excess of market productivity) can the union extract from the firm; and (2) how are these rents distributed across union members? We have only broad insights on many of these latter issues. The locus of union membership is surely driven by a cost and return computation – are there union rents to extract from the firm and at what cost? – but unions have rarely flourished in a strictly voluntary contracting environment, and the locus of union jobs in the United States has a strong political element with which most labour economists are not especially comfortable. Unions flourish in locations where they have the greatest political support – that is, in local political constituencies, states and cities, in which workers in other high potential-rent industries reside. Most analysts believe that unions increase member wages in the United States by an average of 5–15 per cent, although ardent proponents of alternative wage theories, specifically efficiency wages, argue that unions raise wages not at all. 1 We know even less about individual wages – how these union rents, should they exist, are distributed across union members. If the union is a homogeneous craft union, that might not be much of an issue, but the distribution question is more serious in industrial unions representing multiple crafts. The United States evidence suggests that unions distribute rents in an egalitarian way across skills. Freeman and Medoff (1984) attribute this egalitarianism to the voting power of low-skilled majorities, although the evidence is that the same wage patterns hold when high wage workers hold a majority (Parsons, 1992). This is not to say that high skilled workers in the United States are fully supportive of transfers to the less skilled; when in the majority, they frequently reject formal unionism, relying instead on implicit unionization threats to extract rents that are distributed in a more elitist fashion (Parsons, 1991). But these findings give only a hint of the full process of wage formation in unionized labour ‘markets’. I emphasize the limited progress we have made in understanding individual wage formation in syndicalized work places in the United States in order to highlight the potential value of the Pedersen, Smith and Stephensen contribution in Chapter 6. The bulk of Danish wages are set directly or indirectly by collective decisions and this is just the area where our understanding is most limited. The allocation rules are unclear, especially on the important question of skill differentials. Because individuals remain free to invest in skills or not, certain human capital relationships must hold – the private returns to investing in skills must justify the private costs of investing – but this balance can be obtained by reducing the private costs of investing by subsidizing schooling and training without requiring higher wages for skilled workers. Indeed Denmark and other egalitarian states heavily subsidize schooling in situations that the
188
Comments
189
United States does not. The international labour market may also place weak constraints on the structure of wages – dissatisfied workers may migrate abroad. Still in general the individual wage structure in Denmark is what union syndicates and politicians want it to be and the Pedersen, Smith and Stephensen study is a most welcome exploration of this apparently uncharted area (the authors cite only a few, highly specialized prior investigations of individual wage formation in the Danish economy). The chapter is largely empirical, as one might expect at this level of development. The authors estimate the wage-setting process for a sample of individuals in six unions – Unskilled (SID), Metal (METAL), Clerical (HK), Building and Construction (BYD), Academic (AC) and Female (KAD). They partition the analysis by sex as well as union identity and estimate wage equations for the sex-union pairs which have sufficiently large sample sizes as to be interesting – a total of five male wage models and three female models. They use panel data over the period 1980–90 and estimate fixed-effects models, which eliminates potentially troubling problems with individual unobservables, although at the cost of suppressing estimates of relatively permanent wage factors such as education and training. They particularly stress estimation of government policy variables such as unemployment benefit levels and tax parameters on the Danish wage structure. Although the theoretical characterization of the wage-setting process is not the focus of the chapter, the authors motivate interpretation of their results using a minimum wage union contract in which unions set a minimum wage for the least skilled members, with market forces then determining the wages of workers whose salaries are above the minimum. This contract form has become more widespread in the period after their data was drawn, but probably is not the best model to employ to analyze the sample they in fact analyze. I am also a little sceptical of the validity of the model, even in the presence of a minimum wage contract. The assumption that workers above the minimum receive competitively determined wages would probably come as a surprise to union members, the vast bulk of whom are in this category, and is also inconsistent with what other experts report. Jensen (1995, p. 8) argues, for example: The contracts negotiated at central level not only specify the adjustment of the minimum wage during the contract period, they also signal a benchmark for the regulation of wages to be negotiated at firm level. The critical question then is to what extent the local part of the total wage adjustment differs from the ‘guidelines’ indicated by the unions and employers’ organizations? The available evidence seems to indicate that the centrally determined guidelines for local wage adjustments actually do constitute an important benchmark. Intuitively, this is easy to understand….This intuition is borne out by the fact that many firms have introduced automatic wage adjustments equal to the negotiated guidelines. Fortunately the estimation is not constrained in any serious way by this wage model, so one can reject the simple minimum wage model without requiring any change in the estimation. I do have a more direct concern about the way the authors implement their empirics, however. The results indicate that individual wages are powerfully driven by the unemployment insurance (UI) benefit levels for which the worker qualifies. Not only are the regression coefficients on this variable large, they are extraordinarily significant, with t-values that range from 10 to 50 (for clerical women). One wonders if this is because the UI benefits measure is simply a good proxy for the wage rate, the dependent variable in the right-hand side of the equation. As they describe the construction of the UI measure: ‘The potential [UI] compensation rate is
Parsons
190
calculated as 90 per cent of the individual’s wage rate up to a maximum, which corresponds to the maximum compensation of the UI scheme for each year’ (p. 178). Were it not for the benefit maximum, the UI measure and the dependent variable would be perfectly correlated and the regression model inestimable. Much of the data, including the sample means, is intriguing, even if one discards the regression results. For example, the average wage is virtually identical across skills for men and also for women despite large differences in education (although wage differentials between men and women remain). Unskilled male workers are paid the same as metal workers or building and construction union members. Only male members of the academic union are paid significantly more on average. One hardly needs progressive taxation in a syndicalized labour market with this degree of egalitarianism. One may not need much of an economic theory to explain this phenomenon either. Once the potential problem with the UI benefit measure is sorted out, I would be interested in seeing additional structural analyses of the wage functions. Important questions can be confronted with appropriate structural tests. For example, how strong is the coordinating power of the umbrella labour union organization? Do specific unions differ in their wage-setting practices or can the individual union samples be pooled for more precise estimates? Of special interest, can the hypothesis of identical wage functions for male and female workers in the same union be rejected? Are there different wage processes for men and women in the clerical union (HK), or among the unskilled (SID). Some members might be surprised if there are, especially in HK where the working conditions are likely to be similar for men and women. Clearly there is much yet to learn about individual wage-setting in the Danish economy and the authors are on a very promising track for providing us with new information on that process.
Note 1.
See the rather unconvincing Econometrica paper by Summers and Krueger (1988).
References Freeman, R. and J. Medoff (1984) What Do Unions Do? (New York: Basic Books). Jensen, S.E.H. (1995) ‘Recent developments in the Danish economy: problems, policies, and prospects’, EPRU Analyses, 13, Economic Policy Research Unit, Copenhagen. Krueger, A.B. and L.H. Summers (1988) ‘Efficiency wages and the inter-industry wage structure’, Econometrica, 56 (March), pp. 259–93. Parsons, D.O. (1991) ‘The distribution of worker rents in union and strategically unorganized work places’, Research in Labor Economics, 12, R.G. Ehrenberg (ed.), pp. 191–212. Parsons, D.O. (1992) ‘The Internal Distribution of union rents: An empirical test of the voting power model’, Review of Economics and Statistics, 74 (August), pp. 439–45.
Comment Jan Rose Sørensen This Chapter 6 is a nice piece of work with relevant information on the functioning of the Danish labour market. Moreover, the authors follow a sensible approach with a theoretical model which makes the background for estimations of the Danish wage formation for different groups of workers. It is not made totally clear what the purpose of the chapter is, but, since a lot of effort is used on explaining the tax effects on wage formation, I interpret it as if the main purpose is to empirically find these tax effects. In particular, the focus seems to be on what has been termed the ‘Robin Hood’ effect. This effect implies that an increase in the marginal tax rate gives rise to a lower wage rate and higher employment. In, for instance, a wage-bargaining model, where trade unions are concerned about the after-tax wage as well as employment, an increase in the marginal tax rate makes it relatively more attractive for trade unions to choose a higher employment at the cost of a lower wage. This is very different from what happens in a model of a competitive labour market where a higher marginal tax rate gives rise to a lower labour supply and in turn an increase in the pre tax wage. The theoretical model which the authors set up as a background for their estimations is only presented verbally but, in spite of that, I find the model very relevant and interesting. It is assumed that trade unions play an important role in the determination of wages, but firms unilaterally determine employment. Hence, the model is of the ‘right-to-manage’ type, but relative to the existing literature there is an interesting change which probably makes this trade union model relevant for a lot more labour markets than is the case for most existing trade union models. It is assumed that the trade union sets only a minimum wage and the workers are then free to get a higher wage rate if they are able to do so. This implies that there is a ‘required’ minimum level of productivity which determines whether or not a worker gets a job. If her productivity is below the ‘required’ level, she is not worth the minimum wage and will be unemployed. If her productivity is above the ‘required’ level, she will be employed, and she will be paid according to her productivity. However, the wage of an employed worker will still be proportional to the minimum wage. Hence, the trade union indirectly determines the level of all wages. What also makes this chapter very interesting is that the data used in the estimations are panel data on 1 per cent of the Danish population. Moreover, the workers in the data set can be divided according to which trade union they belong to. This division turns out to be very fruitful as there are some important differences in the estimated parameters for the different groups of workers – and, in particular, between men and women. For all the male groups of workers, except for building and construction workers, there seems to be a ‘Robin Hood’ effect. Whereas, for the female groups, except those belonging to KAD, a higher marginal tax rate gives rise to a higher wage rate. This is a surprising difference, but the authors offer a possible explanation. They argue that the labour supply elasticities are typically found to be higher for women than for men, and this tends to make the ‘normal’ labour supply effect of higher marginal tax rates the dominating one for women. The problem with this explanation is that it is difficult to argue that there should be a big difference between the women who are members of SID and those who are members of KAD as
191
192
Sørensen
both groups are unskilled and they work in very similar types of jobs. Another result which I find strange and difficult to explain is that, for the male groups, the ‘Robin Hood’ effect is most pronounced in the wage equation for academics. Theoretically, an increase in the marginal tax rate gives rise to a lower wage rate only if trade unions play an important role in wage determination, and I would expect that trade unions play a much more important role in the wage determination for unskilled than for academics. In conclusion, I find the discussion about the effects of higher marginal tax rates to be very important, and not at least for Denmark. The chapter gives valuable information on this topic, but I am still not convinced that the ‘Robin Hood’ effect is something to take very seriously in policy discussions. There is still too much in the empirical results we do not understand and cannot explain.
7 Government Solvency, Social Security and Debt Reduction in Denmark1 Ninette Pilegaard Hansen, Svend E. Hougaard Jensen and Martin Junge 7.1
INTRODUCTION
From a macroeconomic perspective, the Danish economy has performed very well in recent years. After a long period of recession, the rate of economic growth since 1993 has been relatively high, the unemployment rate has fallen and no significant wage–price pressure has been evident. The discretionary impact of fiscal policy has on the whole been neutral – with the period 1993–5 being a possible exception, as a result of the so-called fiscal ‘kick start’ – so the upturn has mainly been driven by non-government demand. Also the track record of public finances has been good: the ratio of public debt–GDP has fallen considerably, and the budget turned into surplus in 1997. If it so wished, Denmark could join EMU by 1999. Despite the Danish reservations about participating fully in the final stage of EMU, fiscal policy has been conducted in accordance with a commitment to the principles of nominal convergence and fiscal discipline. The position of the Danish government is to adhere to public debt reduction, not only relative to GDP but also to run budget surpluses in absolute amount. For example, the Finance Minister has recently aired an ambitious plan of eliminating the current level of public debt over the next 10–15 years (see Lykketoft, 1995). The official target set by the Danish government (see Ministry of Economic Affairs, 1997), is less far-reaching, thus involving a reduction in the debt ratio from its current level of almost 70 per cent of GDP to 40 per cent of GDP by year 2005.2 Moreover, the government has declared that it would strive to gradually reduce the size of government so as to pave the way for a lower tax burden, currently among the highest in Europe. The purpose of this chapter is to examine whether similarly optimistic conclusions can be drawn about the medium-to-long-term stance of public finances and whether the above mentioned targets are likely to be achieved through the current path of fiscal policy. Among the questions we try to 193
194
Hansen, Jensen and Junge
answer are the following: Is current fiscal policy sustainable? If not, what is the magnitude of the fiscal imbalance? To what extent are we passing fiscal burdens onto future generations? What would be an appropriate design of a fiscal adjustment aiming at debt reduction? In particular, what is the significance of the duration and the composition of a fiscal tightening for the macroeconomic response and for the distribution of welfare across different current and future generations? There are several reasons why we think these questions need to be addressed: first, outlays to transfer payments to people of working age (notably unemployment and early retirement benefits) have increased significantly, which has undoubtedly raised the size of the structural deficit. Second, the process of changing demographics is likely to lead to a substantial increase in the number of elderly, a phenomenon which will almost certainly put upward pressure on public expenditures. Third, a tendency toward earlier retirement has been observed which, unless tempered by adequate welfare reforms, will reinforce the budgetary pressure caused by population ageing. Fourth, in many countries a considerable amount of fiscal consolidation would be needed in order not to violate EMU’s fiscal criteria, both for current members and possible entrants. Although these challenges in themselves seem enormous, others could be added. For example, as a result of the rise in transfer payments, other expenditures have been cut back, including public service production and investment in infrastructure. A pressure for more resources to be allocated to these areas might therefore arise. One could also envisage that it will be increasingly difficult to raise the revenues needed to sustain the financial viability of the welfare state. For example, increased mobility within Europe may induce some members of the labour force to flee to lower-tax jurisdictions, thereby eroding the tax base. From here we proceed as follows. Section 7.2 seeks to give a brief overview of some historical trends in the behaviour of public finances. Section 7.3 presents some calculations of the medium-to-long-term sustainability of current fiscal policy, and section 7.4 tries to assess the country’s fiscal position from the perspective of generational accounting. Using an inter-temporal simulation model, section 7.5 examines some macroeconomic and welfare effects of fiscal adjustments aiming at public debt reduction. Section 7.6 concludes.
7.2
TRENDS IN FISCAL PERFORMANCE
The extent of government involvement in the Danish economy has increased significantly over the last 35 years. Table 7.1 summarizes the basic facts. The share of general government expenditures to GDP has more than doubled,
Table 7.1
Principal budget components, unemployment, interest and growth rates, 1960–95 1960
1965
1970
Total expenditures – Government consumption – Transfer payments – Government investment – Interest payments Total revenues – Personal income taxes – Social security contributions – Indirect taxes – Other revenues General government deficit Government debtb
24.8 13.5 7.1 2.7 1.4 26.6 11.1 1.2 9.0 5.3 –1.8 2.1
30.0 16.4 8.8 3.6 1.2 31.2 13.1 1.6 11.6 4.8 –1.2 3.2
39.5 20.0 13.5 4.7 1.3 41.0 19.6 1.6 14.7 5.0 –1.5 5.2
Unemployment rate Real interest ratec GDP growth rate Growth-adjusted real interest rate
1.8 0.6 6.1 –5.6
0.9 –0.2 4.5 –4.7
1.3 –0.4 2.0 –2.4
1975
1980
(Per cent of GDP) 48.6 56.9 24.6 26.7 19.1 22.9 3.8 3.4 1.2 3.9 47.3 53.6 23.1 23.6 0.5 0.8 13.1 16.8 10.5 12.5 1.4 3.3 10.5 32.1 (Per cent) 5.3 7.0 –3.3 6.6 –0.7 –0.4 –2.6 7.0
1985
1990
1995a
59.9 25.3 22.5 2.3 9.9 57.9 24.6 1.9 15.5 15.9 2.0 61.2
59.3 25.3 24.7 2.0 7.3 57.8 25.6 1.5 15.6 15.1 1.5 61.1
62.5 25.2 28.6 2.0 6.7 60.6 27.7 1.6 15.1 16.2 1.9 83.0
9.1 3.8 4.3 –0.5
9.6 5.2 1.4 3.8
10.3 4.7 2.7 2.0
a
Preliminary numbers. 1995 incl. creditors (3.4 percentage points). c Effective long-term nominal interest rate minus average inflation rate over preceding six years. Sources: ADAM bank (April 1997); Statistical Yearbook (1962, 1968, 1972); Ministry of Finance. b
195
196
Hansen, Jensen and Junge
from 25 per cent in 1960 to 62 per cent in 1995. Though the share of government revenues also has risen strongly in this period, increases have not been sufficient to balance the government budgets. As a result of persistent deficits, primarily since the mid-1970s, there has been a substantial increase in the (gross) government debt–GDP ratio, from 5 per cent in 1970 to almost 80 per cent in 1995.3 On the expenditure side, it is mainly within the ‘core’ areas of the welfare state that outlays have increased more than proportionally with income. The most expanding area has thus been transfer payments, which have increased fourfold. Government consumption, including health and education costs, grew rapidly in the 1960s, but has grown more or less in line with GDP since the mid-1970s. Reflecting the steady increase in public debt since the early 1970s, interest expenditures have also risen. Government investment, on the other hand, has been in decline as a share of GDP. On the revenue side, it is mainly personal income taxes that have increased more than GDP. Unlike in many other European countries, contributions of employers and employees have remained almost negligible. A recent tax reform (1993) (see Chapter 6 in this volume) will, however, gradually shift the taxation from income taxes to payroll and environmentally related taxes. Indirect taxes have also increased their share of GDP, particularly VAT. The significant slowdown in economic growth after the early 1970s is an important source of the rise in the expenditure ratio. Many welfare programmes were designed in the 1960s with rapid economic growth and ‘optimistic’ expectations. The fall in economic growth rates, in combination with slower population growth, has made these schemes significantly more expensive than originally envisaged. Therefore, the tax rates needed to maintain the benefit rates expected by recipients have become very high. The growth in public debt may partly reflect the failure of adjusting welfare programmes to changing economic conditions. Public finances have also been adversely affected by the rise in the structural rate of unemployment. Not only have almost 10 per cent of the labour force become registered as unemployed; in addition, a large number of working-age people have become recipients of some sort of income compensating public transfer payment other than unemployment benefits. Evidence suggests that about 18 per cent of all working-age people are nonemployed but should in principle have the ability to work (see Ølgaard, 1995). With the corresponding number in the 1960s being almost negligible, two worrying tendencies have been identified: the workforce is retiring earlier, and a larger number of (mainly low-skilled) workers have been expelled from the labour market. These developments have clearly had budgetary effects: the costs of benefits for the unemployed and retirees have increased and tax revenues have fallen.
Solvency, Social Security and Debt Reduction
197
The period considered has also been characterized by a remarkable rise in the real interest rate. The combination of a rising real interest rate and a falling growth rate has indeed led to a ‘double’ upward pressure on the socalled growth-adjusted real interest rate (rg for short), which approximately equals the difference between the real interest rate and the real economic growth rate. The sign of rg is critical for the extent to which the level of public debt has a negative effect on the government’s financial position. For example, if rg is negative, the debt–GDP ratio can be held constant only if the government runs a (primary) deficit: it can borrow, pay the interest on this loan with new debt, and still have a surplus left over that can be used either to increase expenditures or to lower taxes. This basic insight may help to explain why the debt–GDP ratio remained stable through the 1960s and the first half of the 1970s and then rose sharply in the 1980s and 1990s. Until the late 1970s nominal interest rates were relatively low, possibly reflecting that actual inflation may have been higher than expected inflation, and with high real economic growth rg was negative (ex post). The rise observed in the debt–GDP ratio in the second half of the 1970s was thus exclusively the result of primary deficits (the deficit excluding interest paid). Thereafter, however, nominal interest rates rose, reflecting that inflation was increasingly anticipated, and inflation became a less powerful device for eroding the real value of government debt. In combination with a falling real economic growth rate, rg turned positive and has remained at very high levels since. The dynamics of debt accumulation have thus been enforced by the effects of primary deficits. More recently, however, primary balances have been positive, so the positive rg has worked against them. Although a rising share of government spending to GDP is a phenomenon that has been observed in most industrial countries (see Masson and Mussa, 1995), there may be some distinct features of the Danish economy accounting for the particularly strong growth in the size of government. Denmark has, like other Scandinavian countries, undoubtedly carried the welfare state further than most other European countries, both in terms of coverage and generosity (see Hagen et al., 1997). For example, safety nets have not only been provided for the less fortunate in society, but as a result of the so-called ‘universalist’ approach to the welfare state, relatively generous social programmes have been extended to the general population.4 The generosity of the Danish welfare programmes may also be reflected in the way many transfer payments have been indexed. For example, in the area of public pensions to the elderly, evidence shows that the purchasing power of public pensions did increase by more than real wages of blue-collar workers over the period 1970–92 (see Socialkommissionen, 1993). Also a certain ‘sense of entitlement’ may have become particularly widespread in Denmark and other Scandinavian welfare states. For example, early retirement benefits were introduced in 1979 as a temporary measure to improve
198
Hansen, Jensen and Junge
job opportunities for young workers by stimulating elderly workers to leave the workforce. However, most people now feel themselves entitled to receive public retirement benefits at the age of 60, regardless of their health status (or other conditions). It has, therefore, become difficult to scale back these benefits.
7.3
DEBT SUSTAINABILITY
Analyses of sustainability of a fiscal programme typically start with the annual budget identity, stating that the change in the stock of debt between (end-of-years) t and t–1 is equal to the deficit run by the government through year t. By iterating the annual identity forwards, we get the intertemporal budget identity. It says that the difference between the present value of the stock of debt in some future year and the current stock of debt is equal to the present value of all primary budget balances in between. The intertemporal identity becomes a constraint when imposing the condition that, as the terminal date goes to infinity, the present value of outstanding debt must tend to zero. A fiscal programme is thus said to be sustainable if the current stock of debt equals the present value of all future primary balances. In this study, however, a narrower interpretation of sustainability is adopted, reflecting the fact that we only consider fiscal programmes with finite horizons (see Appendix 1, p. 214). Specifically, with debt and deficits being scaled by the general price level and GDP, we calculate a tax rate, –i,n, which is held constant between the implementation date, i, and the terminal date, n, and which ensures that the terminal debt–GDP ratio, bn, will not be higher than the initial debt ratio (bn ≤ b0). The difference between the permanent tax ratio and the current tax ratio, i,n (≡ –i,n – 0), serves as an intelligible indicator of sustainability: if i,n < 0, a fiscal tightening (of that magnitude) is required to make fiscal policy sustainable.5 Since the fiscal adjustment could equally well take the form of a permanent spending reduction, a more precise term for the required fiscal adjustment would be the ‘immediate, permanent reduction in the primary deficit’ (see Auerbach, 1994). Given the forward-looking nature of i,n, a number of assumptions must be made. First, the terminal date is important, not only for the expected magnitude of the required fiscal adjustment but also for the uncertainty associated with the calculation. The role of the time horizon may be of particular importance in the Danish case, where the demographic projections suggest that the next decade will be characterized by a declining share of elderly, whereas the years following this ‘breathing-space’ will be characterized by strong population ageing. We shall consider three alternative terminal dates
Solvency, Social Security and Debt Reduction
199
– namely, year 2005 (the ‘medium term’); year 2030 (the ‘long term’); and year 2070 (‘the very long term’). Second, assumptions must be made about the indexation of future per capita primary expenditures. In view of our presumption that population ageing will have a great impact on sustainability, we distinguish between expenditures allocated to the elderly and other expenditures. As to the latter, we assume that they constitute a constant share of GDP. In relation to oldage expenditures, we distinguish between income transfers and other costs, including health. Most public debate has focused on pensions, but health and other services provided for the elderly may be important, too, especially because the number of very old elderly is expected to increase (‘double ageing’). In the base case we assume that both categories of expenditures go up in line with earnings, an assumption fairly consistent with how these expenditures have evolved in the past. From a policy perspective, however, it would be interesting to examine a scheme where ‘other costs’ are indexed at, say, a rate 1 percentage point below the growth rate of productivity. After all, the growth in old-age expenditures can be ‘decoupled’ from the demographic shifts only through indexation schemes providing less-than-full coverage of productivity growth. Third, the government’s debt target is important. Two alternative debt targets are considered, each reflecting possible scenarios from the Danish policy debate. The first is a ‘soft’ one where the ultimate debt–GDP ratio is the same as in the base year. This would not only ensure sustainability, but it would also be consistent with full Danish participation in EMU. The second scenario is an ambitious one where all outstanding debt has been eliminated at the end of the relevant horizon. The objective of public debt elimination seems to be shared across a broad political spectrum, although the positions differ with respect to what would be a realistic time horizon for achieving that objective. Finally, the growth-adjusted real interest rate rg is critical for sustainability. We consider only cases where rg is positive. The baseline involves an annual real interest rate of 5.0 per cent and a real GDP growth rate of 2.0 per cent. Since some scholars are of the impression that the objective of price stability would gain credibility if Denmark joined EMU, and thereby eventually would enjoy a lower nominal (and real) interest rate, we also consider a scenario where the real interest rate exceeds the growth rate by only 1 percentage point, which is low compared to current levels. The calculations can only be meaningfully performed if the debt and deficit ratios are linked in a stock–flow consistent manner. Among the several measures of debts and deficits being offered, only two satisfy this condition – namely, the so-called ‘registered government debt’ (‘registrerede statsgæld’), which is approximately stock–flow consistent with the so-called ‘DAU balance’ (‘DAU saldo’).6 The problem is, however, that the latter does not include
Hansen, Jensen and Junge
200
those expenditures and revenues which are directly controlled by lower-tier authorities (‘kommuner/amtskommuner’). Since we want not only to use ‘DAU’ as our deficit measure but also to use expenditure and revenue ratios for the entire public sector, we make a level displacement equal to the expenditures and revenues governed by lower-tier authorities under the (relatively harmless) assumption that these run balanced budgets. Table 7.2 shows the composition of the three alternative measures of public debt used in this chapter: ‘gross government debt’, ‘EMU debt’ and ‘registered government debt’. As to the last-mentioned, we report numbers for both 1994 and 1995, as well as the difference between them. Our estimate of the DAU ratio is –3.2 per cent, reflecting public sector revenue and expenditure ratios of, respectively, 56.8 and 60.0 per cent in 1995. Since we shall assume that the output gap is zero in 1995, our estimate of the DAU ratio also equals our estimate of the structural deficit ratio for 1995.7 The corresponding debt ratio, b0, equals 59.8 per cent. As the future pressure on
Table 7.2
Alternative measures of public debt and the deficit, 1995, bn DKK ‘Gross ‘EMU debt’ government debt’
State Domestic bonds Foreign bonds Creditors Government holding of own bonds Danmarks Nationalbank Social Pension Fund (DSP) Government bonds Non-govt bonds Lower-tier authorities Consolidation item Funded debt for ‘kasser’ Correction for trade credits Total % of GDP ‘DAU-balance’ (deficit) % of GDP
658.9 111.1 32.8
‘Registered government debt’ 1994
1995
∆
658.9 111.1 32.8
618.8 131.6
658.9 111.1
39.9 –24.2
–6.6 –
–2.8 –55.3 –146.8 –50.1 –96.7
–6.6 –33.7 –151.4 –68.9 –82.5
–3.8 21.6 –4.6
545.5 56.4
578.3 59.8 31.3 3.2
32.8 3.4
–68.9 53.1 –36.5 0.4 –46.1 802.8 83.0
698.2 72.2
Note: Source: Ministry of Finance, Danmarks Nationalbank.
Solvency, Social Security and Debt Reduction
201
public finances is likely to come mainly from ageing populations, the (net) expenditures to the elderly constitute the most interesting part of the government budget. For 1995 we find that the share of total old-age (60+) expenditures to GDP is 13.4 per cent, with the ‘transfer’ and ‘service’ (health) ratios constituting, respectively, 9.5 and 3.9 per cent. These numbers should, however, be adjusted for the fact that old-age pensions are subject to taxation. In Ministry of Finance (1996, p. 126), the taxes paid by the elderly have been estimated to 2.6 per cent of GDP. If proper account is also taken of the fact that some elderly people work – and hence pay labour income taxes – and are wealth owners – and hence pay property and capital income taxes – we have estimated the number to 4.7 per cent of GDP. Table 7.3 summarizes our findings. Suppose, as a benchmark, that fiscal policy is conducted within a balanced budget framework, implying that the debt–GDP ratio is smoothed throughout the relevant horizon. To prevent public debt from rising in the next century, non-neglible increases in the tax ratio are required. The profile of necessary tax adjustments is hump-shaped: modest in the medium term, significant in the long term – by 2030, for example, the tax ratio is projected to be 3.8 percentage points above its current level – and then falling as the ‘baby-boomers’ pass away. The results reported in the next four rows correspond, under different contingencies, to cases where fiscal adjustments are based on longer time horizons. Consider first the magnitude of the required fiscal adjustment
Table 7.3 Horizon Debt target
Magnitude of required fiscal adjustment 2005 b n = b0 bn = 0
2030 bn = b0 bn = 0
2070 bn = b0 bn = 0
A rg = 3 Balanced budget Baseline Delay Diff. indexation Demographics only
0.86 0.93 – 0.75 –0.26
– 6.12 – 5.95 –
i,n – 0 3.82 1.73 2.18 1.21 0.54
– 2.71 3.72 2.19 –
3.33 2.30 2.79 1.43 1.11
– 2.52 3.08 1.65 –
B rg = 1 Balanced budget Baseline Delay Diff. indexation Demographics only
–0.33 –0.27 – –0.45 –0.27
– 5.43 – 5.25 –
i,n – 0 2.62 0.71 1.11 0.12 0.71
– 2.14 3.13 1.55 –
2.13 1.49 1.83 0.34 1.49
– 2.02 2.47 0.88 –
202
Hansen, Jensen and Junge
under baseline assumptions. In order for the debt ratio to return to its initial level by 2030, the tax ratio would have to be raised, immediately and permanently, by 1.7 percentage points above its current level. As a result the public sector would begin to save, and the debt–GDP ratio would be brought down during the ‘breathing space’. Thereafter, as the ‘baby-boomer’ effect sets in, the government would run deficits until the initial debt ratio is recaptured in 2030. If instead the debt is eventually eliminated by 2030, there is a need for an additional tax increase of 1 per cent, a relatively modest extra burden. However, an ambitious debt target could also mean that the government wishes to get rid of its debt in a medium-term perspective. Both the following demographic wind and the economic boom currently prevailing would facilitate such a strategy. Suppose, therefore, that the debt ratio gradually falls from its current level until the year 2005. This necessitates an increase in the tax ratio by 6 percentage points above the current level. If this additional tax burden is levied solely on labour income, the wage tax should, with a wage share of about two-thirds, be raised by about 9 percentage points. For the same time horizon, if the government does not aim at a lower debt–GDP ratio the current tax ratio is almost sustainable, a permanent deficit reduction of 1 percentage point would be required. One could also envisage a scenario where the government, for some ‘political economy’ reasons, decides to delay serious action until some later day. Suppose the government follows a balanced budget policy up to year 2005 before taking action. This scenario could be relevant if a government feels tempted to derive short-term benefits from the reduced pressure on public expenditures. If the required adjustment is postponed by a decade, the permanent tax ratio has to be increased by more than two percentage points above the current level, against an increase of 1.7 percentage points had it been implemented from the very beginning. To what extent represents differentiated indexation a solution to this sustainability problem? If old-age services are indexed at a lower rate than cash transfers, the fiscal imbalance is reduced somewhat, particularly in the long term. For example, while the effect is almost negligible in the medium term, the magnitude of the required fiscal contraction is almost halved if the horizon extends to the year 2070. However, attempts to adjust old-age services more in line with prices than with wages are likely to imply a lower quality of these provisions. In fact, given that the main costs of taking care of the elderly are wage costs (and medical costs), it seems rather unlikely that these costs could be kept in line with the general price level. At the very least, rising wage costs would have to be compensated by productivity increases. Alternatively, the quality of services, and presumably the living standard of the elderly, would fall.
Solvency, Social Security and Debt Reduction
203
The need for a fiscal adjustment has several causes, including ‘high’ primary expenditures, ‘large’ interest payments, a ‘big’ structural deficit in the base year, and ‘ambitious’ plans for debt reduction. Given our emphasis on old-age social security costs, it would be interesting to isolate that part of the adjustment which is attributable to changing demographics. Two tax rates would then have to be calculated, namely (a) the base year tax ratio for the hypothetical case where both interest payments and the structural deficit are zero, and (b) the permanent tax ratio which covers the current and (discounted sum of) future primary expenditures. The need for fiscal adjustment which is due to demographics only can then be found as the difference between the two. This calculation shows that there is indeed a demographic ‘breathing space’ in Denmark: changes in the age structure exert no pressure on public finances in Denmark over the next decade, on the contrary. Evaluated for the long term, however, the required increase in the tax ratio as a result of changing demographics is found to be about half a percentage point. The exact answers also depend on the relative magnitudes of the real interest rate and the growth rate of GDP. Suppose, for example, that the growth-adjusted real interest rate, rg, falls to 1 per cent, accomplished through a fall in the real interest rate of 2 percentage points. While this would lower the burden of servicing the public debt, it also raises the discount factor, thereby raising – in present value – the costs which are concentrated in the medium-to-long-term. To illustrate the importance of the ‘debt service’ effect, suppose the government aims at keeping the debt ratio in year 2005 at its 1995 level. Under what is otherwise baseline assumptions, a fall in rg of 2 percentage points is seen to require a fall in the permanent tax ratio by a quarter of a percentage point, in contrast to the need for a rise of almost 1 percentage point if rg equals 3 per cent a year. While the medium-to-long-term indicators of fiscal policy presented above provide useful information about the extent to which fiscal policy needs adjustment, a number of important questions have been left open. In particular, we have not addressed the issue of how fiscal contractions of the magnitudes demonstrated would impact on different (current and future) generations, and we have also ignored how the macroeconomy would be affected by such actions. These are the issues to be taken up in the remainder of the chapter.
7.4 INTERGENERATIONAL DISTRIBUTION OF FISCAL BURDENS In this section we address the fiscal imbalance from the perspective of generational accounting, an approach motivated by the traditional concern about debt and deficits, namely that they may unduly burden future genera-
204
Hansen, Jensen and Junge
tions. This method is also based on the government’s intertemporal budget constraint which, rather than explicitly expressing the restrictions on the future path of primary deficits, can be stated in terms of the net taxes that have to be levied on current and future generations in order to pay for the government’s spending programme and service its initial debt. Specifically, generational accounts show, in present value, what a representative member of each – current and future – generation can expect to pay in taxes net of transfers over its remaining life span.8 For the implementation (see Appendix 2, p. 216) we need a long-term population projection (1995–2200); an aggregate government budget for 1995; cross-section survey data (in order to distribute aggregate taxes and transfers to existing generations, as defined by age); and an estimate of the government’s current debt position. As to the exogenous parameters, we maintain the baseline assumptions of an annual real GDP growth rate of 2.0 per cent and a real interest rate of 5.0 per cent.9 Current and future generations are represented by, respectively, a current newborn and someone born in the year after. This ensures, since generational accounts are forward-looking, that net payments refer to entire life-times of both ‘representative individuals’. In principle, as the total burden on all future generations is assumed to be equally distributed on each of them, except for growth adjustment, future generations could be represented by anyone yet unborn. Table 7.4 shows the results of five alternative simulations, corresponding to different assumptions about the course of fiscal policy. We first report the baseline scenario, assuming no change in current fiscal policy. Any burden associated with fiscal adjustments initiated so as to meet the government’s intertemporal budget constraint is assumed to be placed on future generations. While this may not be a very realistic assumption, it serves as a useful benchmark for measuring the degree of fiscal imbalance. We find that the generational account of a current newborn, GAt,t, is actually negative.10 This result is robust to alternative combinations of growth and real interest rates (see Jensen and Raffelhüschen, 1997b). Consequently, the net taxes levied on someone born in the future, GAt,t + 1, is positive. The difference between the two accounts shows, in absolute amount, the ‘excess burden’ passed on to future generations. The magnitude of these numbers indicates that (at least) some of the burden of restoring fiscal sustainability will have to be levied on current generations. It would, undoubtedly, be perceived as both unfair and infeasible to meet the government’s intertemporal budget constraint by passing such heavy burdens on to future generations. Against this background, we run three alternative simulations leading to generational balance, defined as policies which equalize the net tax payments of future generations and the (growth-adjusted) net tax payments of newborns – that
Solvency, Social Security and Debt Reduction Table 7.4
Generational accounts of current and future generations
Generations
Generational balance
age in 1995
Baseline
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90
–6 188 560 914 1271 1376 1314 1156 869 476 1 –417 –862 –1020 –1088 –1128 –1160 –1149 –277
Future Gen. % change
205
277 –
LIT(1995)
LIT(2005)
(thousands of DKK) 74 101 281 311 668 701 1034 1059 1405 1409 1516 1505 1449 1428 1284 1252 984 943 574 535 78 45 –354 –384 –815 –840 –982 –1005 –1060 –1078 –1107 –1123 –1144 –1157 –1136 –1149 –274 –277 75 7.5
103 8.9
Elderly
REF
32 232 612 971 1335 1446 1394 1249 976 600 145 –246 –657 –807 –885 –940 –982 –983 –236
41 235 621 1001 1369 1469 1403 1244 950 548 61 –372 –832 –996 –1068 –1111 –1148 –1141 –277
33 15
142 –
is (1 + g)GAt,t = GAt,t + 1. Policies with this property fulfil what Kotlikoff (1993) has called the ‘fiscal balance rule’. We shall again emphasize the role of timing, but here we also pay attention to the composition of the fiscal adjustment. The first scenario involves an immediate and permanent adjustment of the labour income tax (LIT) levied on living generations. In this case an additional revenue equal to 7.5 per cent of the existing tax revenue is required. Since labour income taxes constitute about a third of GDP, this result is in line with the result found in section 7.3 of a permanent rise in the total tax ratio of about 2 per cent of GDP. The second experiment also leads to generational equity through a higher labour income tax, but it is now assumed that the government delays this
206
Hansen, Jensen and Junge
adjustment until year 2005. This would just magnify the required adjustment, with the permanent increase in the labour income tax increase being 1.5 percentage points higher than if implemented immediately. The third experiment assumes that the net tax payments of current newborns and future generations are equalized by cutting expenditures allocated to the elderly. This may well be an unfair policy and is hardly realistic. Indeed, the magnitude of the required adjustment equals an across-theboard cut in pensions, health, etc. of 15 per cent relative to the current expenditure level. Compared to the baseline scenario, restoring generational balance would clearly raise the net tax burdens placed on current generations. An interesting question is how sensitive the generational account of each current generation is to the specific form of achieving generational balance. Table 7.4 also illustrates these effects. If the wage tax is increased immediately, the adjustment will leave the generational accounts of all individuals above working age almost unaffected. If the action is delayed, however, the adjustment will also leave unaffected some of the generations currently active in the labour market – namely those who will retire in about a decade. Instead, extra burdens will have to be placed on those who are of working age now and those who become labour market active during the years with higher wage taxes. It follows from the very nature of the present value concept that by delaying action, higher revenues will have to be raised at the time when action is taken in order to raise the required payments. If the burdens are placed exclusively on the elderly, current newborns would also feel this adjustment in their generational account but not by nearly as much as they would had the adjustment taken the form of higher wage taxes. We finally consider a scenario (REF) which is based on rather optimistic assumptions about the Danish economy. It is not designed so as to bring about fiscal balance per se; rather, it is based on the assumption that the labour market and tax reforms implemented in recent years would lead to a significant fall in the structural rate of unemployment and thereby ‘endogenously’ bring about stronger public finances. Specifically, the rate of unemployment is assumed to fall by a total of 3.4 percentage points over the years 1996–9. The increase in employment will not only reduce expenditures on unemployment and cash benefits, it also leads to higher tax revenues, thereby expanding the generational accounts of current generations. In our calculations we use official estimates of the budgetary effects of a fall in the rate of unemployment (see Ministry of Finance, 1994). The question arises whether the extra revenues should solely be used to reduce public debt, or whether it should also translate into a fall in the tax burden on living generations. We assume that over a period of 10 years, the extra revenues are used to reduce the government’s financial debt, so no
Solvency, Social Security and Debt Reduction
207
discretionary fiscal adjustment (such as a cut in tax rates) will take place. By the end of year 2010 the government has almost got rid of its outstanding debt. In view of the strength of public finances in year 2011, we assume that a permanent tax cut is implemented so as to ensure that the debt level prevailing in that year can be kept constant in the future. This is seen to offset much of the fiscal imbalance associated with the baseline scenario. The generational accounts of current generations increase and those of future generations fall. In short, policy reforms which lead to lower unemployment will also yield a ‘generational dividend’. Finally, our calculations are based on those stock and flow components which are currently included in the government’s accounting practice. This excludes such implicit government assets as, for example, deferred tax revenues associated with private pension schemes, the present value of which has been estimated to more than a third of GDP. An example of an implicit liability is the government’s commitment to pay pensions to civil servants (see Bovenberg and Petersen, 1992). The present value of these future transfer payments has been estimated to about 10 per cent of GDP. We have made some preliminary experiments where these implicit assets and liabilities are included in the calculations of the generational accounts. Our results suggest that this would more than halve the magnitude of the required fiscal adjustments.
7.5 MACROECONOMIC AND WELFARE EFFECTS OF FISCAL ADJUSTMENTS While many of the recent attempts to built more flexibility into the Danish economy would also help overcoming the fiscal imbalance, there is likely to be a need for further discretionary fiscal adjustments. Against this background, we try in this section to remedy a serious shortcoming of the analysis carried out so far – namely, the implicit assumption that such fiscal adjustments can be undertaken without having macroeconomic repercussions. More specifically, we shall examine how the generational and macroeconomic effects depend on whether the adjustments take the form of a higher wage tax or a cut in the pension benefit, and whether the adjustments are implemented over the medium or the long term. Recent evidence seems to suggest that the composition of fiscal adjustments influences both their duration and their macroeconomic consequences. It has thus been found (see Alesina and Perotti, 1997), that fiscal adjustments which rely primarily on spending cuts on transfers and the government wage bill stand a good chance of lasting long and being expansionary. On the other hand, fiscal adjustments which rely primarily on tax increases and cuts in public investment tend not to last and may be contractionary.
208
Hansen, Jensen and Junge
Our analysis is based on simulations using an intertemporal CGE model with optimizing agents, the EPRU model for short.11 It portrays an initial steady state of the Danish economy, and it offers a description of the dynamic adjustment towards a new steady state after a policy change has been initiated. The fiscal adjustments are designed such that the current level of public debt will eventually be eliminated. Specifically, the debt target is achieved smoothly through intertemporal fiscal adjustments in either the wage tax or the pension benefit rate to the elderly. As in section 7.4 we shall assume that fiscal policy is tightened temporarily. As soon as the target debt ratio has been achieved – that is, when there is no more public debt in the economy – fiscal policy will be relaxed again. To ensure comparability with the results presented above, we distinguish between a medium-term approach (10 years) and a long-term approach (35 years). The latter strategy would be largely consistent with the government’s target of bringing the debt ratio down to 40 per cent by the year 2005. What are the magnitudes of the intertemporal fiscal adjustments needed to accomplish a gradual elimination of public debt? This is clearly sensitive to both the composition and the duration of these adjustments. If public debt is settled through a heavier taxation of labour income (and unemployment benefits) over a period of 10 years, the (average) wage tax must be raised by 9.9 percentage points on impact. When the debt has eventually been settled, a cut of 13.7 percentage points can be implemented (which is almost 4 percentage points below the initial steady state). If instead the debt is settled over a horizon of 35 years, much smaller adjustments in fiscal policy is required. Indeed, it would be enough to raise the average wage tax by 1.8 percentage points and, when there is no more debt after 35 years, the tax ratio can be set at a level 5.2 percentage points below the initial steady state. Suppose instead fiscal policy is tightened in the form of lower pension benefits. Rather than raising the benefit rate again when the public debt has been eliminated, we assume that the (distortionary) wage tax is reduced, keeping the pension benefits at their low level. Effectively, this amounts to a partial conversion from a public to a private pension system without preannouncement. Within a medium-term horizon, the benefit rate – defined as the pension benefit relative to the average wage minus payroll taxes – would have to fall by 79.2 per cent! Subsequently, however, the wage tax could be cut by 19.2 per cent (relative to the initial steady state). Obviously, if the process is stretched over a longer time horizon, the intertemporal profiles become smoother. The corresponding numbers for the long-term horizon are thus 11.8 and 8.4, respectively. The design of a debt reduction programme thus involves a number of trade-offs: How should the long-term benefits be weighted against the transitional losses? Who should carry the burden among current generations, workers or pensioners? And how should the gains and pains be distributed
Solvency, Social Security and Debt Reduction
209
between current and future generations? Although we do not explicitly formulate a social welfare function to answer these questions, some indications can be offered. For example, in the case of tax adjustments, since the deadweight loss imposed by a distortionary tax is believed to increase by more than in proportion to the tax rate, a smooth tax profile can be motivated on efficiency grounds (see Barro, 1979). Similarly, a fiscal adjustment involving a cut of almost 80 per cent in the pension benefits would not be sensible, at least if the adjustment has not been announced well in advance. Figure 7.1 shows some macroeconomic trajectories for the scenario (a) where the fiscal adjustment only involve the wage tax and for the scenario (b) where the public debt is eliminated through a temporary cut in the pension benefit rate followed (after 10 or 35 years) by a permanent cut in the wage tax so as to maintain the debt ratio at zero. The first important question is where the economy ends up if the government undertakes to eliminate its outstanding debt. We observe that the lower wage tax rates implied by both scenarios lead to a downward pressure on wages. This stimulates demand for labour and hence employment. The Macroeconomic effects of debt elimination (a) wage tax
(b) pension benefit rate cum wage tax
year
40
35
30
Consumption, fixed prices
year
45 50
35 40
35
30
25
20
15
10 10
1.2 1.16 1.12 1.08 1.04 1 0.96 0.92 0.88
0 5
45 50
35 40
35
30
25
20
15
10
0 5
10 year
35
year
Consumption, fixed prices 1.2 1.16 1.12 1.08 1.04 1 0.96 0.92 0.88
25
20
10
10 0 5
45 50
35 40
35
30
25
20
15
10
0 5
10
GDP (at factor prices) 1.12 1.1 1.08 1.06 1.04 1.02 1 0.98 0.96 0.94
45 50
GDP (at factor prices) 1.12 1.1 1.08 1.06 1.04 1.02 1 0.98 0.96 0.94
15
Figure 7.1
Hansen, Jensen and Junge
210 Figure 7.1
(Cont’d) Before-tax real-wage
Before-tax real-wage
1.025
1.025
1.015
1.015
1.005
1.005
0.995
0.995
45 50
35 40
35
30
25
20
15
10 5
0.975
10
45 50
40
35
35
30
25
20
15
5
10
0
0.975
0.985 10
0
1.985
year Employment
Employment 1.06 1.04 1.02 1.00 0.98
year
45 50
35 40
35
30
25
20
15
10 10
0.96 0.94
5
45 50
35 40
35
30
25
20
15
10
5
10 0
0.98 0.96 0.94
0
1.06 1.04 1.02 1.00
year
effects on output and employment are relatively small, however, compared to the upsurge in consumption. The reason for this is that not only has public debt been eliminated; the country has changed its position vis-à-vis the rest of the world from being a net debtor to a net creditor. Rather than paying interest to foreigners as in the initial steady state, interest incomes are received from abroad, so GNP exceeds GDP. To what extent are these long-term gains offset by transitional losses? Like the steady-state effects, the transitional effects depend on the composition and the duration of the adjustments undertaken. As to the composition, the labour market response is most pronounced if wage taxes are raised. Higher wage taxes would induce workers to demand more leisure, which is not subject to taxation. This effect is clearly not present if the public debt is brought down through lower transfer payments to the elderly. As to the role of timing, we find that a stretched-out elimination period (n = 35) generates a much smoother macroeconomic outcome. While a ‘tough’ approach may generate output losses of more than 3 per cent relative to the initial steadystate, the ‘soft’ approach at no time brings losses above 1 per cent. In any case, our results do not support the so-called ‘expansionary fiscal contraction’ hypothesis, as offered by Giavazzi and Pagano (1990).
Solvency, Social Security and Debt Reduction
211
The desirability of alternative debt strategies may also be evaluated in terms of their effect on the intergenerational distribution of consumer welfare (see Figure 7.2). The percentage welfare gains are measured along the vertical axis, while the different generations are ordered along the horizontal axis, with generations becoming successively younger as we move from left to right. Generation 0 is the generation entering the labour market Figure 7.2
Welfare effects of debt elimination Relative compensating variation, 35 years
percent
10
0
25
20
15
10
5
0
–5
–10
–15
–20
–25
–30
–35
–40
–45
–10
–50
wage-tax pension cum wage-tax
Generation Relative compensating variation, 10 years
0
Generation
25
20
15
10
5
0
–5
–10
–15
–20
–25
–30
–35
–40
–40
–45
wage-tax pension cum wage-tax –50
percent
40
Hansen, Jensen and Junge
212
in the period where the specific debt policy is launched; generation –50 is the generation which entered the labour market 50 years before that time, while generation 25 comprises yet unborn individuals who will start their working careers in 25 years. If the debt target is achieved through a higher wage tax, the currently living generations of working age are the real losers. However, since pension benefits go untaxed, retired people are broadly left unaffected. In the case of a short elimination period, the most severely hit generation is the one which entered the labour market 25 years before the introduction of a new debt policy and which has to pay the high wage tax each year over its remaining 10 years as a member of the workforce. Unlike previously born generations, it never gains from a lower wage tax. Not all future generations win; indeed, generations born right after the higher tax rates have been introduced would also lose. A very different distribution of losses across living generations is found if the public debt is brought down by cutting transfer payments to the elderly. While a true gesture to yet unborn generations, debt reduction achieved through this composition means that the gains gradually diminish with age. All retirees would clearly suffer from such a programme, in particular if a short elimination period is chosen. The magnitude of the losses incurred by the elderly suggests that if debt policy is combined with a (partial) conversion from a public to a private pension scheme, this should be announced well in advance to avoid strong intergenerational welfare distributions.
7.6
CONCLUSIONS
The stance of fiscal policy in Denmark may not be as good as a casual glance at the current levels of debt and deficits might suggest. Since the upsurge in unemployment in the early 1970s, the behaviour of public finances shows an important asymmetry, namely that the surplus on the government budget in ‘good’ years is significantly smaller than the deficit in ‘bad’ years. If public finances were sufficiently sound, a balanced budget would typically come out in a ‘normal’ year. It could therefore be argued that there is a need for fiscal retrenchment in Denmark. Indeed, it is somehow revealing that one can be impressed by the prospect of a balanced budget after 4 unbroken years with economic upturn. The case for fiscal adjustments seems particularly strong in view of the projected rise in the demographic dependency burden. In the absence of a tighter fiscal policy, public debt is likely to rise in the medium-to-long-term, and tax burdens will be passed on to future generations. Delay in addressing these imbalances may ultimately necessitate very tough policies. Our analysis suggests that the need for additional discretionary fiscal actions would
Solvency, Social Security and Debt Reduction
213
diminish significantly if recently introduced initiatives to bring down the structural rate of unemployment prove successful. Labour market reforms thus serve as a good surrogate for fiscal retrenchment. Although we agree to the need for a fiscal tightening, we are also aware of the risk that a fiscal contraction may cause some transitional distress. Our findings suggest that adherence to very ambitious debt targets may undermine macroeconomic stability in the short run. This is particularly so if the fiscal adjustment takes the form of higher wage taxes which may lead to wage pressure. This shows, in essence, the familiar choice between sound finance and the use of fiscal policy to manage demand. From a policy perspective, this suggests that fiscal policy should be designed as a long-term strategy of public debt elimination. This would not only be relatively lenient to the macroeconomy but it would also stand a good chance of adding to credibility.
Appendix 1 The Intertemporal Budget Constraint and Sustainability Consider first the government’s annual budget identity, written in levels and nominal terms:
Bs − Bs−1 = (Gs + H s − Ts ) + is Bs−1 = Ds + is Bs−1 ; s = 1, K, n
( A 7.1)
where B is the outstanding stock of government bonds; G is government spending on goods and services; H is transfer payments; T is the tax revenue; D (≡ G + H – T) is the primary deficit (net of interest payments), and i is the interest paid on government debt. We restrict attention to the case where budget deficits are financed by issuing bonds to the general public. Deficits could also be monetized or financed by running down the foreign reserves, but these alternative options are not considered. The variables in (A7.1) can also be expressed in real terms and as ratios to GDP. For example, the ratio of real government debt to GDP is defined as b ≡ B/(PQ), where P is the domestic price level, and Q is real output (GDP). After some tedious manipulations, (A7.1) can be written as bs − bs −1 = d s + rsg bs −1
(A7.2)
where rsg ≡
rs − s 1 + s
; rs ≡
is − π s 1+πs
; πs ≡
Ps − Ps −1 Ps −1
and s ≡
Ys − Ys −1 Ys −1
By iterating (A7.2) forward, and upon discounting to time zero, we obtain the intertemporal budget identity
R0, n bn − b0 = ∑s=1 R0, s d s ; R0, s = ∏ is=1
1
n
1 + ri g
( A 7.3 )
where R0,s is the growth-adjusted discount factor. (A7.3) says that the difference between the debt–GDP ratio prevailing at the end of the terminal year (discounted to time zero) and the initial ratio at time zero is equal to the discounted sum of primary deficits over the years in between. A sustainable fiscal policy can be defined as a policy which ensures that the debt ratio eventually converges back to its initial level. Fiscal adjustments implied by more ambitious debt targets could also be considered. Hence we will simply express the terminal condition as bn ≤ b0. Suppose the future courses of the growth-adjusted real interest rate and the government expenditures are known. The corresponding tax and debt profiles are then implied by the intertemporal budget constraint. For example, if government expenditures are financed along the way (pay-as-you-go), the debt ratio is smoothed
214
Solvency, Social Security and Debt Reduction
215
at its current level. Alternatively, if the tax burden associated with the same expenditure programme is smoothed, the debt ratio becomes endogenous. A switch from ‘pay-as-you-go’ to tax smoothing can be implemented either immediately or with a delay of, say, i periods (i < n). From (A7.3) it is straightforward to compute the tax rate in the case of debt smoothing. Under tax smoothing, however, the expression is more complicated n t i ,n = ∑ s = i +1 R0,s
−1
{∑
i s =1
R0,s ( g s + hs − t s ) +
}
∑s = i +1 R0,s ( g s + hs ) + (1 − R0,n )b0 + R0,n (b0 − bn ) n
(A7.4)
where –i,n reflects the primary balance up to period i (first term); the average primary expenditures after time i (second term); accrued interest payments on the initial debt (third term) and the cost associated with realizing a certain debt target (last term).
Appendix 2 The Method of Generational Accounting Generational accounting also starts with the government’s intertemporal budget constraint. It goes one step further, however, in asking not simply how large fiscal adjustments are required to ensure a sustainable fiscal policy, but how different changes in taxes and spending, at different dates, will affect different age groups in the population. The starting point is therefore a reformulation of the intertemporal budget constraint, saying that the government must ultimately pay for its spending and service its initial indebtedness with resources obtained from current and future generations. Formally, we have
∑s =0 N t,t − s + ∑s =1 N t,t + s = ∑s = t Gs (1 + r)t − s + Dtg ∞
T
∞
(B7.1)
where Nt,k represents the present value in the base year t of the remaining lifetime net payments to the government be the generation born in year k. Generations refer to individuals by specific years of age. Net payments are the difference between the government tax receipts and the government transfer payments. Present values reflect discounting at a pre-tax real interest rate, r. The maximum lifetime of a particular generation is denoted T. The first term on the left-hand side of (B7.1) thus adds together the present value of payments by existing generations, ranging from the newborn generation (s = 0) to the oldest generation alive (s = T). The second term comprises the present value of net payments by yet unborn generations. On the right-hand side, the first term is the present value of government consumption, with Gs denoting government consumption in year s, and the second term is the current level of government debt. The term Nt,k is defined as follows N t ,k = ∑ s =max( t ,k ) Ts ,k Ps ,k (1 + r )t − s (B7.2) – where T s,k denotes the average net amount that a representative individual of the generation born in year k is projected to pay to the government in year s, and Ps,k is the number of surviving members of the generation in year s who were born in year k. The summation begins in the base year t for currently living generations, and in year k for future generations. Regardless of the generation’s year of birth, the discounting is always back to year t. By adding the values of Nt,k, one for each current and future generation, one gets exactly the right-hand side of (B7.1). – More specifically, T s,k is an aggregate of all relevant taxes and transfers given by k+T
Ts ,k = ∑ i hsi ,k
(B7.3)
where his,k represents each category of taxes and transfers which a representative individual born in year k either pays or receives. The indexation of each item follows the rule
hsi, k = hti, k (1 + g ) s− t ; s > t
(B7.4 )
where g is the exogenous (steady-state) growth rate of GDP.
216
Solvency, Social Security and Debt Reduction
217
A set of gender-specific generational accounts can now be found by simply dividing the value of Nt,k by the corresponding number of members of each generation nt , k =
N t,k Pt ,k
(B7.5)
where nt thus indicates, in present value, what the typical individual of each generation can expect, net, to pay now and in the future to the government. Having estimated the government’s debt position, the present value of net payments by current generations, and the projected present value of the government’s spending programme, the total amount future generations will have to pay can be calculated as a residual from (B7.1). The standard assumption adopted in the literature is that the average life-time tax payment of successive generations is spread smoothly across all future generations. The time path of generational is adjusted with the exogenous growth factor. Hence the generational accounts of all future generations are equal, except for the growth adjustment. Since the generational accounts of current newborns and future generations take into account the net tax payments over these generation’s entire life-times, the life-time net tax payments of future generations are directly comparable only with those of current newborns.
Notes 1. 2.
3. 4.
5.
We wish to thank Peder Andersen, Lans Bovenberg and Niels K. Frederiksen for very helpful comments. The usual disclaimer applies. Financial support from the Danish National Research Foundation is gratefully acknowledged. Discussions of public debt are often confused by the fact that several alternative definitions are offered (see Ministry of Finance, 1993). When we write ‘almost 70 per cent of GDP’ we have in mind the so-called ‘EMU debt ratio’ – used by the EU for inter-country comparisons – which in 1996 was 69.9 per cent. It should be mentioned that if some specific Danish institutional features are allowed for – including a state-owned pension fund’s holdings of nongovernment bonds and the state’s deposits in Danmarks Nationalbank – the debt ratio would actually fall below 60 per cent. Gross government debt is a measure for which data are available for the entire period 1960–95. For reasons already pointed out, this measure tends to exaggerate the indebtedness of the public sector. Persson (1995) offers a political-economy model to explain why the Scandinavian countries with so little wage dispersion nevertheless have so large public sectors; a result opposite to what would be predicted by an optimal taxation argument. The indicator of sustainability used here is forward-looking, following the approach developed by Blanchard et al. (1991). See also Auerbach (1994, 1995). Other indicators make predictions on the basis of past behaviour (see Hamilton and Flavin, 1986; Trehan and Walsh, 1991). The latter approach looks at the time-series properties of government debt in the past and tests, basically, whether debt grows faster than the discount rate. See also Rudin and Smith (1994) for a simple indicator of sustainability based on past behaviour. The problem with indicators based on past behaviour of the debt is that they will not capture, by definition, the projected changes in the age structure of the population, a phenomenon widely perceived as the major threat to the sustainability problems faced by many countries.
Hansen, Jensen and Junge
218 6. 7.
8.
9.
10.
11.
Jensen and Motzfeldt (1994) contains a more detailed treatment of the corrections needed to establish full stock–flow consistence between the ‘DAU’ balance and the ‘registered government debt’. Ministry of Finance (1996) and OECD (1996) report output gaps slightly below zero. Notice, however, that the calculation of output gaps involves much uncertainty. Also, estimates may differ due to different methods being used (Giorni et al., 1995). In Jensen and Nielsen (1995) we discuss the importance of making cyclical correction of the current budget deficit for Denmark. Generational accounting is now used by several governments and international organizations. A more detailed methodological description can be found in Auerbach, Gokhale and Kotlikoff (1991) and Auerbach and Kotlikoff (1997). A lengthy and informal review of generational accounting is offered by Kotlikoff (1992). For a critical assessment of generational accounting, see Bohn (1992), Haveman (1994), Buiter (1995) and Fehr and Kotlikoff (1997). Details of the data used in the generational accounts for Denmark can be found in Junge et al. (1997). Since an empirical study cannot possibly comprise all future generations, generational accounts may not strictly obey the intertemporal budget constraint. Behind this ‘average’ account are hidden some, by international comparison, substantial gender-specific differences. We have elsewhere (Jensen and Raffelhüschen, 1997a), discussed the special gender-specific aspects of the Danish tax and transfer system. It should be noted that we have counted costs on education as transfer payments. If, instead, these costs are counted as government consumption the generational accounts of current newborns and future generations do, of course, increase. In this case, we find that the net tax burdens of future generations will be about 50 per cent higher than for current generations (see Jensen and Raffelhüschen, 1997b). Although the quantitative robustness is relatively low, the qualitative finding of an imbalance in favour of current generations is very robust. The EPRU model is documented in Jensen et al. (1996). Applications include Jensen et al. (1994), Jensen (1997), and, more recently, Lange, Pedersen and Sørensen (1997). Other studies of debt policies based on CGE models include James (1994) and Macklem, Rose and Tetlow (1994).
References Alesina, A. and R. Perotti (1997) ‘Fiscal adjustments in OECD countries: composition and macroeconomic effects’, NBER Working Paper, 5730. Auerbach, A. (1994) ‘The US fiscal problem: where we are, how we got here, and where we’re going’, in S. Fischer, and J. Rotemberg (eds), NBER Macroeconomics Annual (Cambridge, Mass.: MIT Press). Auerbach, A. (1995) ‘Solutions for developed countries’, in T. Hoenig (ed.), Budget Deficits and Debt: Issues and Options (Federal Reserve Bank of Kansas City). Auerbach, A. and L. Kotlikoff, ‘The methodology of generational accounting’, in A. Auerbach, L. Kotlikoff and W. Leibfritz (eds), Generational Accounting around the World (Chicago: University of Chicago Press). Auerbach, A., J. Gokhale and L. Kotlikoff (1991) ‘Generational accounts – a meaningful alternative to deficit accounting’, in D. Bradford (ed.), Tax Policy and the Economy, (Cambridge, Mass.: MIT Press).
Solvency, Social Security and Debt Reduction
219
Auerbach, A., J. Gokhale and L. Kotlikoff (1994) ‘Generational accounting – a meaningful way to evaluate fiscal policy’, Journal of Economic Perspectives, 8, pp. 73–94. Barro, R. (1979) ‘On the determination of public debt’, Journal of Political Economy, 87, pp. 940–71. Blanchard, O., J. Chouraqui, R. Hageman and N. Sartor (1991) ‘The sustainability of fiscal policy: new answers to an old question’, OECD Economic Studies, 15, pp. 7–36. Bohn, H. (1992) ‘Budget deficits and government accounting’, Carnegie–Rochester Conference Series on Public Policy, 37, pp. 1–83. Bovenberg, L. and C. Petersen (1992) ‘Public debt and pension policy’, Fiscal Studies, 13, pp. 1–14. Buiter, W. (1995) ‘Generational accounts, aggregate saving and intergenerational distribution’, NBER Working Paper, 5087. Danmarks Nationalbank (1996) Statens låntagning og gæld. Danmarks Statistik (various years) Statistical Yearbook. Fehr, H. and L. Kotlikoff (1997) ‘Generational accounting in general equilibrium’, Finanzarchiv, 52, pp. 1–27. Giavazzi, F. and M. Pagano (1990) ‘Can severe fiscal contractions be expansionary? Tales of two small countries, in O. Blanchard and S. Fischer (eds), NBER Macroeconomics Annual 1990 (Cambridge, Mass.: MIT Press), pp. 75–123. Giorno, C., P. Richardson, D. Roseveare and P. Noord (1995) ‘Potential output, output gaps and structural budget balances’, OECD Economic Studies, 24, pp. 167–209. Hagen, K., E. Norrman, P.B. Sørensen and T. Teir (1997) ‘Financing the Nordic welfare states in an integrating Europe’, in P.B. Sørensen, (ed.), Tax Policy in the Nordic Countries (London: Macmillan). Hamilton, J. and M. Flavin (1986) ‘On the limitations of government borrowing: a framework for empirical testing’, American Economic Review, 76, pp. 808–19. Haveman, R. (1994) ‘Should generational accounts replace public budgets and deficits?’, Journal of Economic Perspectives, 8, pp. 95–111. James, S. (1994) ‘Debt reduction with distorting taxes and incomplete Ricardianism: A computable dynamic general equilibrium analysis’, in W. Robson, and W. Scarth (eds), Deficit Reduction – What Pain, What Gain? (Ontario: C.D. Howe Institute). Jensen, S.H. (1997) ‘Debt reduction, wage formation and intergenerational welfare’, in D.P. Broer and J. Lassila (eds), Pension Policies and Public Debt in Dynamic CGE Models (Heidelberg: Physica-Verlag). Jensen, S.H. and C. Motzfeldt (1994) ‘Offentlig gæld og budgetunderskud i Danmark: er finanspolitikken holdbar?’, Nationaløkonomisk Tidsskrift, 132, pp. 318–34. Jensen, S.H. and S.B. Nielsen (1995) ‘Population ageing, public debt and sustainable fiscal policy’, Fiscal Studies, 16, pp. 1–20. Jensen, S.H. and B. Raffelhüschen (1997a) ‘Generational and gender-specific aspects of the tax and transfer system in Denmark’, Empirical Economics, 22, pp. 615–36. Jensen, S.H. and B. Raffelhüschen (1997b) ‘Public debt, welfare reforms, and intergenerational distribution of tax burdens in Denmark’, in A. Auerbach, L. Kotlikoff and W. Leibfritz (eds), Generational Accounting Around the World (Chicago: University of Chicago Press). Jensen, S.H., S.B. Nielsen, L.H. Pedersen and P.B. Sørensen (1994) ‘Labour tax reform, employment and intergenerational distribution’, Scandinavian Journal of Economics, 96, pp. 381–401.
220
Hansen, Jensen and Junge
Jensen, S.H., S.B. Nielsen, L.H. Pedersen and P.B. Sørensen (1996) ‘Tax policy, housing, and the labour market: an interemporal simulation approach’, Economic Modelling, 13, pp. 355–82. Junge, M., P. Jacobsen, S.H. Jensen and B. Raffelhüschen (1997) ‘Dokumentation af datamaterialet i det danske generationsregnskab’, EPRU, mimeo. Kotlikoff, L. (1992) Generational Accounting – Knowing who Pays, and When, for What we Spend (New York: Free Press). Kotlikoff, L. (1993) ‘From deficit delusion to the fiscal balance rule: looking for an economically meaningful way to assess fiscal policy’, Journal of Economics, 7, pp. 17–41. Lange, K., L.H. Pedersen and P.B. Sørensen (1997) ‘The Danish Tax Reform Act of 1993: effects on the macroeconomy and on intergenerational welfare’, EPRU, mimeo, see chapter 5. Lykketoft, M. (1995) ‘Den danske pensionsdebat’, Nordisk Forsikringstidsskrift, 76, pp. 28–30. Macklem, T., D. Rose and R. Tetlow (1994) ‘Government debt and deficits in Canada: a macro simulation study’, in W. Robson and W. Scarth (eds), Deficit Reduction – What Pain, What Gain?, (Ontario: C.D. Howe Institute). Masson, P. and M. Mussa (1995) ‘Long-term tendencies in budget deficits and debt’, in T. Hoenig (ed.), Budget Deficits and Debt: Issues and Options (Federal Reserve Bank of Kansas City). Ministry of Economic Affairs (1997) Økonomisk Oversigt. Ministry of Finance (1993, 1994, 1996) Finansredegørelse (Copenhagen). OECD (1996) Economic Surveys – Denmark (Paris: OECD). Ølgaard, A. (1995) ‘Borgerløn ad bagdøren’, University of Copenhagen, mimeo. Persson, M. (1995) ‘Why are taxes so high in egalitarian societies?’, Scandinavian Journal of Economics, 97, pp. 569–80. Rudin, J. and G. Smith (1994) ‘Government deficits: measuring solvency and sustainability’, in W. Robson and W. Scarth (eds), Deficit Reduction – What Pain, What Gain?, (Ontario: C.D. Howe Institute). Socialkommissionen (1993) De ældre – en belysning af ældregenerationens forsørgelse (Copenhagen: DKB Bogdistribution). Trehan, B. and C. Walsh (1991) ‘Testing intertemporal budget constraints: theory and applications to US federal budget and current account deficits’, Journal of Money, Credit, and Banking, 23, pp. 206–23.
Comment A. Lans Bovenberg 1
INTRODUCTION
Chapter 7 contains a rich menu for those interested in Danish fiscal policy. I will discuss the various components of this menu in turn. First, the description of the recent Danish experience is put in perspective by comparing it with recent Dutch performance. Subsequently, I discuss the methodology for measuring fiscal sustainability before turning to the results that the methodology yields for Danish fiscal policy. After exploring the limitations of generational accounting and discussing the general equilibrium results, I put fiscal consolidation in the broader perspective of a strategy addressing the ageing of the population.
2
THE DANISH AND DUTCH ECONOMIES COMPARED
Section 7.2 provides a brief overview of the recent performance of the Danish economy in general and Danish fiscal policy in particular. For an outside observer from the Netherlands, the similarities between the Dutch and Danish experiences are quite striking. On the positive side, the rising trends in public debt, fiscal deficits, and unemployment have been stemmed in recent years. Indeed, as far as economic and employment growth is concerned, the small open economies of Denmark and the Netherlands are at the moment outperforming the larger economies on the European continent. On the negative side, however, the stocks of public debt and the levels of inactivity remain high in both countries. A large part of the working-age population lacks formal employment and has become hidden unemployed. Those with few marketable skills especially do not have easy access to the labour market. Moreover, many people retire early, stimulated by generous early retirement and social insurance schemes, which were instituted about 15 years ago to alleviate the adverse impact of high unemployment rates on the employment prospects of young workers. As a result of a low effective retirement age and a high rate of hidden unemployment (especially among the low skilled), public transfer payments are still high.1 The policy objectives in the two countries are also quite similar. Major policy objectives are reducing the stock of public debt, cutting the share of government spending (and hence taxation) in GDP, and raising labour force participation and employment. On these scores, Denmark currently features a higher rate of labour force participation, whereas the Netherlands levies a lower tax burden.
3
FISCAL SUSTAINABILITY
Section 7.3 investigates the sustainability of fiscal policy in constructing so-called ‘permanent tax ratios’. These tax ratios indicate which constant (‘permanent’) tax level
221
222
Bovenberg
would be required to arrive at a certain pre-determined level for the debt–GDP ratio in the terminal year. These indicators for fiscal sustainability feature two major strengths. First, in contrast to traditional measures of fiscal sustainability (such as the fiscal deficit), they are forward-looking and thus explicitly account for the impact of future ageing on the public finances. Second, these tax ratios incorporate the expected developments of all categories of public spending – that is, not only public pensions but also spending on health care and other spending that may or may not be sensitive to demographic developments. Section 7.4 employs generational accounting. This methodology also provides a measure for fiscal sustainability. This measure is the difference between the accounts of the so-called ‘current newborn’ and the accounts of the generation that is born a year later. The current newborn are the youngest of the generations that are alive in the base year 1995. For these current generations, the generational accounts are based on the extrapolation of current fiscal policies. In extrapolating current policies, one does not take account of the government’s intertemporal budget constraint. In arriving at the accounts for the generations that are not yet born in the base year (the so-called ‘future’ generations), in contrast, the accounts are determined by the intertemporal budget constraint facing the public sector. In particular, these accounts are derived from the room that the intertemporal budget constraint leaves for future generations after subtracting the net benefits that accrue to present generations under current fiscal policies. Only if the accounts of the youngest of the present generations (the current newborn) coincide with the accounts of the oldest of the future generations (the generation born immediately after the base year) is fiscal policy sustainable; in that case, current fiscal policy (reflected in the accounts of the current newborn) coincides with the room allowed for by the intertemporal budget constraint so that present fiscal policy meets the budget constraint. Fiscal sustainability can thus be judged on the basis of the calculations in both section 7.3 and section 7.4. The major difference between the two sections is that section 7.3 employs a finite time horizon (2005, 2030 or 2070), whereas section 7.4 uses an infinite time horizon.
4
DANISH FISCAL POLICY
The results indicate that current fiscal policy in Denmark is not sustainable – despite the fact that public debt is declining at the moment. An immediate and permanent rise in the tax–GDP ratio of about 2 percentage points is required to avoid a longterm explosion of public debt. Compared to other countries, however, this adjustment is relatively minor. A similar exercise for the Netherlands, for example, revealed that an adjustment of about 5 per cent of GDP would be required.2 It would be interesting to know why the Danish public finances are less unsustainable than public finances in other countries. Is it because Danish public spending is less sensitive to demographic changes than is the public spending in other countries? Indeed, the required increase in the tax ratio that originates in demographic changes amounts to only half a percentage point, which is rather small. Another important result is that the public finances would become almost sustainable if the structural rate of unemployment declined by 3.4 percentage points over the years 1996–9. This result illustrates that a well functioning economy is an important precondition for sound public finances. A generational accounting exercise for the Netherlands provided another illustration of this important point.3 It showed that the sustainability of Dutch public finances depends on a race between ageing and an increasing rate of labour force participation. If labour force participation continued to
Solvency, Social Security and Debt Reduction
223
rise rapidly during the next two decades, Dutch public finances would actually be sustainable. This exercise revealed also that the implicit tax claim on the assets of private pension schemes can make an important contribution to sound public finances. Chapter 7 does not address the consequences of sustainable public finances for the development of the fiscal deficit and public debt over time. However, in setting objectives for fiscal policy, the government may want to know what sustainable public finances would imply for the future path of the public deficit. The exercise for the Netherlands indicated that sustainable public finances would require the current fiscal deficit of about 2 per cent of GDP to turn into a fiscal surplus after the turn of century. This surplus might have to reach a level of as high as about 4 per cent of GDP by 2020. After that date, the surplus would start to shrink as a consequence of the ageing of the population.
5
LIMITATIONS OF GENERATIONAL ACCOUNTING
Although generational accounting is a powerful tool, it suffers from several important limitations. In particular, generational accounting considers only intergenerational transfers that occur through the government budget. Accordingly, it abstracts from generational transfers within the family (for example, bequests), and ignores external effects across generations that are not recorded on the government budget. These external effects can be negative. Environmental damage is an important example of such adverse externalities. These external effects may also be positive, however. The intergenerational transfers of knowledge and human capital are important examples of these positive externalities. Indeed, future generations are likely to be richer than current ones on account of continued economic growth, which originates in a higher stock of knowledge. Accordingly, one has to be careful in drawing strong normative conclusions from generational accounting.
6
GENERAL EQUILIBRIUM ANALYSIS
Another important drawback of generational accounting is that it abstracts from behavioural changes. Section 7.5 addresses this drawback by using a dynamic applied general equilibrium model to explore how behavioral changes induced by changes in fiscal policy affect the general equilibrium of the Danish economy. An important question in this connection is whether the general equilibrium effects due to behavioural changes substantially change the results from generational accounting, which abstract from these general equilibrium effects. Unfortunately, the results in section 7.5 do not shed much light on this question, because the general equilibrium results are not compared with those from generational accounting. Another important issue involving the exercises in section 7.5 is the benchmark that is used for the model simulations. In particular, fiscal policy is assumed to be already sustainable in the benchmark. Some more information on the nature of the benchmark would have been welcome. Since tax rates appear to be constant in the benchmark, it is not immediately obvious why a short-term tightening of fiscal policy would be desirable. Tax smoothing considerations, for example, would counsel against first increasing but later decreasing tax rates. Indeed, the benchmark seems to abstract from the ageing of the population, which would appear to be the main reason for tightening fiscal policy at this time. Moreover, the benchmark assumes that current fiscal policy meets the intertemporal budget constraint and is thus sustainable. This is at variance with the
Bovenberg
224
generational accounting exercise. Indeed, given this assumption behind the model simulations, it is not surprising that tightening of fiscal policy is contractionary rather than expansionary. Indeed, a tighter fiscal policy is more likely to be expansionary if current fiscal policies are considered to be unsustainable.
7
DEALING WITH AGEING
Chapter 7 stresses the importance of raising public saving to address the ageing trend. This is certainly an important element of a broad-based strategy that deals with the economic challenges associated with ageing. However, such a broad-based strategy should include other elements as well, especially in view of the uncertainty about future economic conditions. In particular, raising financial saving appears to be an attractive strategy if the growth-adjusted real interest rate is high. A high value for this rate indicates that financial saving is attractive compared to saving in the form of human capital. However, ageing may well reduce not only the growth rate of the population but also the rate of return on capital (as ageing makes capital less scarce compared to labour). Consequently, investing in human rather than financial capital may become more attractive. In view of the uncertainty about the relative returns on financial and human capital, society may want to diversify over various types of investments. Indeed, in order to spread risks, policy makers should take action on several fronts. Important elements of such a strategy are a higher effective retirement age (by making early retirement schemes more actuarially fair), increasing the efficiency of labour and commodity markets, broadening the tax system by reducing tax privileges to the retired population, tightening the requirements for social security benefits, encouraging life-long learning, and raising labour force participation.4
Notes 1 2 3 4
For the Dutch experience, see Bovenberg (1997). See ter Rele (1997). See ter Rele (1997). For the fundamental uncertainties affecting pension schemes and investment strategies, see Bovenberg and van der Linden (1997). This paper explores also in more detail the various elements of a policy strategy aimed at addressing the ageing trend.
References Bovenberg, A.L. (1997) ‘Dutch employment growth: an analysis’, CPB Report, 1997/2, pp. 16–24. Bovenberg, A.L. and A.S.M. van der Linden (1997) ‘Can we afford to grow old? Adjusting pension policies to a more aged society’, in ‘Family Market and Community, Equity and Efficiency in Social Policy’, OECD Social Policy Studies, 21 (1997) ter Rele, H.J.M., ‘Ageing and the Dutch public sector, applying and extending generational accounting’, CPB Report, 1997/3, pp. 17–21.
8 Economic Expansions and Fiscal Contractions: International Evidence and the 1982 Danish Stabilization1 U. Michael Bergman and Michael M. Hutchison
8.1
INTRODUCTION
Consolidation of government fiscal positions is on the policy agenda of many countries. The necessity of improved fiscal positions has been evident for a number of years because of a pattern of large public sector deficits and a corresponding rise in public indebtedness. For a number of European countries, the desire to meet the convergence criteria of the Maastricht Treaty to qualify for the final stage of EMU sets national budgetary targets (public deficits not to exceed 3 per cent of GDP and public debt not to exceed 60 per cent of GDP) that entail substantial contractions in fiscal policy. Concerns over the negative output and employment consequences of fiscal consolidation, at least in the short term, have reinforced political opposition to specific adjustment measures in many cases. The call for early elections in France in May and June 1997, for example, were motivated in part by the government’s desire to receive a mandate before proceeding further with fiscal austerity measures, widely anticipated to adversely affect output and employment, in order to meet the converge criteria of the Maastricht Treaty. These fears appear well founded. Conventional economic analysis predicts that a fiscal contraction reduces aggregate demand directly – by reducing government expenditures – and indirectly – by reducing current income through cutbacks in transfer payments and increased taxes. Aggregate demand falls as government expenditures are reduced and consumption declines in tandem with the drop in household income. The extent of wage 225
226
Bergman and Hutchison
and price rigidities and other costs of adjustment determine how long the adverse output and employment effects of a fiscal contraction will last. Standard large-scale macroeconomic models predict substantial short-run output costs associated with a fiscal contraction (for example, Bartolini, Razin and Symansky, 1995). But are concerns over the short-term negative economic consequences of multi-year fiscal contractions justified? In principle, it is possible that a fiscal contraction which leads to expectations of significantly higher ‘permanent income’ – that is, higher average expected future disposable income – could lead to a boom in private consumption. Permanent income could rise, for example, if the government announces a policy, which is expected to reduce the future size of the government sector and associated tax burden on households. The jump in private consumption, if sufficiently large, could potentially offset or even reverse the initial contraction in aggregate demand arising from the restrictive fiscal policy stance. This theoretical possibility, opposite of the standard Keynesian prediction, has been termed ‘expansionary fiscal contractions’ (see Giavazzi and Pagano, 1990), and often characterized as the ‘German view’ in light of statements by the German Council of Economic Experts in the early 1980s that fiscal retrenchment should set the foundation for an economic expansion. Theoretical models emphasizing expectations about future government policy and its effects on current consumption, investment, output and other macroeconomic aggregates are standard in the literature. There are a number of conditions which need to be met, however, if an immediate fiscal contraction is to lead to an economic expansion. And these conditions are even stronger if, (as emphasized by Giavazzi and Pagano, 1990, 1996; and Bertola and Drazen, 1993), the economic expansion is to be led by a consumption boom. These include sufficiently high wage and price flexibility, a strong supply-side response, well functioning consumer credit markets, and a high degree of confidence that announced government policy will lead to significantly lower future tax burdens.2 In other words, the announced policy must credibly ‘signal’ lower future tax rates and consumers be in a position to act on this new information. If these conditions are not met, a fiscal contraction may easily lead to declining consumption and output in the short run. Given these theoretical ambiguities over the short-term effects of fiscal policy, empirical evidence must be brought to bear on the issue. This chapter evaluates the merit of the ‘expansionary fiscal contraction’ (EFC) view by considering international evidence on fiscal contractions and, in some detail, the specifics of the Danish 1982 fiscal reform. The Danish reform has attracted considerable international attention (for example Alesina and Perotti, 1996a, 1996b; Barry and Devereux, 1994, 1995; Bertola
Economic Expansions and Fiscal Contractions
227
and Drazen, 1993; Giavazzi and Pagano, 1990, 1996; OECD, 1996) for two reasons. First, the Danish economic expansion following implementation of the stabilization package was very strong – real GDP growth averaged over 3.5 per cent and consumption growth over 4 per cent during 1983–6. Second, the Danish case has been put forward, together with the (second) Irish stabilization a few years later, as a fiscal contraction episode providing strong support for the expansionary contraction view (see Giavazzi and Pagano, 1990; Bertola and Drazen, 1993) – that is, an episode where the announced fiscal contraction generated expectations of such high (net) future income levels that current consumption, aggregate demand and real GDP boomed as well. Our empirical investigation takes several perspectives. First, we consider the 15 significant fiscal contractions in the OECD countries during the past 20 years, identifying episodes which are associated with consumption booms and strong economic expansions. We investigate whether these episodes fit the theoretical pattern suggested by the EFC hypothesis, both in terms of assumptions and predictions of the theory. Second, we examine the 1982 Danish fiscal reform and its particulars, including the magnitude, timing and credibility of the announced policy changes. We explore whether the Danish reform could have set up such strong expectational effects in signalling higher personal disposable incomes as to generate the subsequent consumption boom and economic recovery. The EFC has specific predictions regarding the timing and composition of aggregate demand following announcement of a fiscal reform, which may be checked against the actual observations. In this context, we investigate the economic upswing that followed implementation of the fiscal reform, the initial conditions of the economy and other economic circumstances including changes in income policies, monetary policy, exchange rate policy and changes in the external environment. Third, we report the results from a formal econometric time-series model, which is able to distinguish between alternative explanations for the sharp rise in Danish private consumption during 1983–6. Section 8.2 provides the theoretical foundation for the expansionary fiscal contraction hypothesis. Section 8.3 considers some basic summary statistics over the major fiscal consolidations of the past 20 years and evaluates the state of the empirical evidence on the topic. Section 8.4 considers the Danish fiscal reform in some detail and the official projections of its anticipated impact published at the time. Section 8.5 evaluates whether the expansionary fiscal contraction hypothesis appears to fit the general features and statistical profile of the Danish fiscal reform and subsequent consumption boom. Section 8.6 considers more formal empirical tests of the EFC model using Danish data. Section 8.7 concludes the chapter.
228
Bergman and Hutchison
8.2
EXPANSIONARY FISCAL CONTRACTION HYPOTHESIS
Expectations Approach to Fiscal Policy What are the factors that distinguish fiscal consolidations in terms of their effects on economic performance? The expansionary fiscal contraction (EFC) hypothesis attempts to provide an explanation for the exceptional cases where consolidations are marked by consumption booms, strong economic activity and falling unemployment. As discussed in section 8.1, the basic argument is that fiscal consolidation leads to a revision in expectations about future tax burdens. Future taxes are expected to be lower, which in turn increases average current and expected future income net-of-tax (‘permanent’ disposable income) and increases current consumption. If this expectational effect is sufficiently large, thereby increasing (expected) ‘wealth’ significantly, a consumption-boom-led rise in aggregate demand and output could follow. The main operational mechanism is a wealth effect on consumption. To illustrate this case, consider a standard representative agent model with infinitely lived households who smooth expected consumption over time (for example, Blanchard and Fischer, 1989; Obstfeld and Rogoff, 1996). Assume that there is only one composite good, the supply of which is exogenously given outside the model for simplicity. The household’s intertemporal budget constraint is defined as s− t s− t ∞ 1 ∞ 1 − = + + Et ∑s= t C E ( r ) B ( Y T ) 1 ∑ s t t s s s= t 1 + r 1 + r
(8.1)
where Cs is consumption at time s, Bt is initial assets at time t, Ys is income at time s, and Ts is taxes (lump-sum, non-distortionary). This budget constraint states that the present discounted value of (private) consumption must be equal to the present discounted value of net income plus the initial assets. Given preferences (a utility function) the household solves the optimization problem – that is, solves for the path of consumption over time subject to the intertemporal budget constraint. Assume now that households are maximizing utility subject to the intertemporal budget constraint above, the utility function is quadratic and the individual’s time preference is equal to the constant interest rate. In this case, we obtain the following consumption function s− t r ∞ 1 − + + Ct = r B E Y T ( 1 ) ( ) ∑ t t s s s= t 1 + r 1 + r
(8.2)
Economic Expansions and Fiscal Contractions
229
where r is the constant real interest rate.3 We observe that revisions of current or expected future income and taxes immediately change private consumption. For example, a permanent decrease in current (and expected future) taxes increases permanent disposable income and private consumption one-for-one. Expected future tax changes, though not yet in place, may also have a large effect on both current and future private consumption if they are expected to be permanent or long-lasting. On the other hand, a transitory change in taxes will have only a marginal effect on permanent disposable income, leaving current and future private consumption little affected. The government’s intertemporal budget constraint, assuming no initial debt level and the No Ponzi Game condition, is simply s− t s− t ∞ 1 ∞ 1 Et ∑s= t Ts Gs = Et ∑s= t 1 + r 1 + r
(8.3)
where Gs is real government consumption at time s. According to this budget constraint, the present value of current and future government consumption must be equal to the present value of current and future taxes. This implies that the time path of taxes must adjust to a given level of current and future government consumption. If current taxes are cut, for example, future taxes have to be raised in order to restore government solvency (assuming unchanged government consumption). The standard neoclassical model also assumes that households incorporate the government’s intertemporal budget constraint into their own expected wealth – that is, Ricardian equivalence. Substituting the present value of government consumption for the present value of taxes in the household’s intertemporal budget constraint gives the following consumption function r ∞ 1 Ct = (1 + r )Bt + ∑s= t 1 + r 1 + r
s− t
Et (Ys − Gs )
(8.4)
where private consumption is a function of current and future expected income and government consumption. This relationship suggests that changes in private consumption are based on revisions in expectations over future income net-of-government consumption (net-of-tax). This implies a central role for expected future fiscal policy. Consider the effects on private consumption from large changes in current or expected future government consumption which are expected to be long-lasting – that is, the type of fiscal reform identified by Giavazzi and
230
Bergman and Hutchison
Pagano (1990) as likely to create a consumption boom. From the government’s intertemporal budget constraint given in (8.3) we note that expected future taxes will be affected since a lower present value of government consumption implies lower future taxes. Expectations of lower government spending must therefore imply lower expected tax burdens, higher expected net-of-tax income and, thus, higher current and future private consumption. A key issue here is that the fiscal reform programme is credible – that is, that government consumption will be significantly reduced for a sustained period. A credible fiscal reform is likely to give an expansionary effect on private consumption. On the other hand, if it is not credible, consumption will be little affected. This expectations effect is, of course, standard in permanent income models which emphasize wealth (present discounted value of expected labour and non-labour income) as the central determinant of consumption behaviour. This model predicts that a current government expenditure reduction which is expected to be permanent will increase current consumption by the same magnitude and leave aggregate demand (and output) unaffected. This case may be characterized as a consumption boom but not an expansionary economic response to the fiscal contraction. The ‘German view’ or the EFC hypothesis, however, suggests that both consumption and output should expand significantly. In an open economy, consumption and aggregate demand may rise more than the (absolute value of the) fall in government expenditure if it ‘signals’ additional future government expenditure cuts. But for output to expand together with aggregate demand, one would need the fiscal reform to create a positive supply-side response. Of course, government consumption is not the only component affecting private disposable income. Other factors which may lead to a rise in current and future disposable income (and hence permanent income) include productivity increases, more efficient labour market structures, factors enhancing cost competitiveness in the export sector, natural resource discoveries and increases in the terms of trade. Indeed, these other developments may at times have coincided with government expenditure shifts, making it difficult to identify the true source of private consumption expansions.
Extensions The neoclassical permanent income model of consumption reviewed above constitute the basis for Giavazzi and Pagano’s (1990, 1996) argument that large fiscal consolidations, if they signal slower future growth in government expenditures and transfers, may generate a current consumption boom. In the extreme case, an economic expansion may result but this would entail supply-side effects which are not normally considered part of the standard EFC model (see, for example, Barry and Devereux, 1995). Blanchard (1990)
Economic Expansions and Fiscal Contractions
231
also points out the possibility that even a fiscal consolidation with substantial tax increases may have an expansionary effect on consumption. Blanchard’s argument is that a current tax hike may preclude larger and more disruptive future tax increases. If taxes have distortionary effects on the economy, an unexpected current tax hike – if it sets up expectations of smaller future distortionary tax increases – could have a positive wealth effect and generate a rise in consumption. Barry and Devereux (1994) examine the validity of the EFC hypothesis within an overlapping-generations model, thus relaxing the assumption of Ricardian equivalence. They find that, in accordance with the EFC hypothesis, permanent reductions lead to higher consumption, employment, and output. In addition, their analysis suggests that fiscal adjustments leading to lower debt–GDP ratio give even larger output effects. An interesting result is that the effects of transitory reductions in government spending exceed those of permanent reductions, leading the authors to question the credibility hypothesis suggested by other studies. However, the consumption boom is still driven by expectations as is the case in the permanent income model above. Bertola and Drazen (1993) consider a model, similar in spirit to targetzone models of exchange rates, where government spending follows a random walk with a positive drift and national income is assumed constant. Fiscal consolidations or ‘interventions’ have a discrete character, reflecting the political obstacles to reducing government expenditures, but eventually must occur (with probability 1) since the government sector is gradually claiming an increasingly large share of national resources. Large government spending cuts are expected to occur only when spending reaches some upper bound of national resources. The implication is that at low levels of government spending, consumption falls less than one-for-one with current increases in government spending. (If no consolidations were expected to occur, then permanent income and consumption would move inversely onefor-one with government expenditure). This occurs since a current increase in government spending implies a rise in the probability of a future discrete spending cut and hence a lower increase in future expected taxes than in the no-expected-intervention (fiscal consolidation) situation. At high levels of government spending, there is a large probability of it increasing further to the upper-bound trigger point. When it occurs, and fiscal consolidation in the form of a large decline in government spending is undertaken, consumption jumps upward. The discrete fall in government spending increases household permanent disposable income, reflecting the discrete fall in the present discounted value of future taxes. However, consumption increases by less than the absolute value of the fall in government expenditure since the high probability of the fiscal consolidation had already been incorporated into expectations (and hence expected permanent income). A consumption boom is predicted by the Bertola and Drazen
Bergman and Hutchison
232
(1993) model, but not enough to fully offset the fall in government expenditure and generate a rise in overall aggregate demand. The more a fiscal consolidation is anticipated, the less effect it is predicted to have on permanent income and consumption. Also emphasizing the ‘expectations approach to fiscal policy’, Alesina and Perotti (1996b) suggest that the expectational effects of a fiscal consolidation will vary depending upon its composition. In particular, they argue that fiscal consolidations focused on cuts in social expenditure and the wage component of government consumption (as opposed to labour tax increases and cuts on public infrastructure capital) are likely to have more ‘persistent’ or long-lasting effects. Persistent fiscal consolidations, in turn, imply a substantial decline in the present value of future taxation and are therefore more likely to be associated with large increases in consumption. Alesina and Perotti (1996a) suggest that employment is likely to be higher after a fiscal reform than before. This is encouraging for the European governments struggling with fiscal adjustments. However, the empirical evidence also suggests a less clear picture for economic growth. For half the sample of fiscal adjustments studied, they find lower growth after the fiscal reform. The models discussed above highlight cases where consumption booms may follow large fiscal contractions. Even in these models, however, it need not be the case that contractionary fiscal policy generates an economic expansion. Barry and Devereux (1995) survey a number of familiar general equilibrium macroeconomic models, including the frictionless neoclassical growth model with infinitely lived consumers, overlapping-generations models, and neoKeynesian models. They find that, in the context of the neoclassical model with Ricardian equivalence, there is little possibility for a consumption boom to more than offset a permanent fiscal contraction and lead to a economic expansion. Although the overlapping-generations model is capable of supporting the EFC hypothesis, they argue that the underlying assumptions are unrealistic and do not adequately reflect real-world constraints. Their own neo-Keynesian model, incorporating constraints on wage and price flexibility, also rejects the optimistic predictions of the EFC hypothesis. They suggest that other explanations must be used to explain, for example, the fall in unemployment in Denmark and Ireland following the fiscal reforms during the 1980s.
8.3
INTERNATIONAL EVIDENCE
Summary Statistics Whether or not a large fiscal contraction leads to an output expansion is ultimately an empirical question. Many countries have undergone significant fiscal consolidations in attempts to restore solvency to budgetary positions.
Economic Expansions and Fiscal Contractions
233
Table 8.1 shows the 15 episodes of ‘significant’ fiscal consolidations during the past 20 years among the OECD countries, where significant is defined as an improvement in the general government cyclically adjusted financial balance (structural balance) equivalent to at least 3 percentage points of GDP (OECD, 1996). Table 8.1 also shows macroeconomic developments during the fiscal consolidation episodes. Most countries undertaking significant fiscal consolidations experienced recessions, with nine episodes marked by weak output growth and rising unemployment. The remaining six episodes are potential candidates for expansionary fiscal contractions, considered here as real GDP growth exceeding 2 per cent per annum (average) and unemployment either rising modestly (1 percentage point or less) or falling. In this group, four consolidations took place amidst strong economies and falling rates of unemployment (Denmark, 1983–6; Australia, 1986–8; Ireland, 1986–9; Sweden, 1986–7), while two episodes were associated with strong output growth but a modest rise in unemployment (Italy, 1976–7; Japan, 1980–7). Clearly, fiscal contractions are not uniformly associated with recessions. But to what extent are the expansions identified in Table 8.1 due to consumptionled booms as predicted by the EFC hypothesis? During seven of the 15 fiscal consolidations, real consumption growth was moderate to strong – averaging 2 per cent per annum or greater. However, in only three episodes among those identified as potential candidates for EFC was consumption growth both strong and above real GDP growth during the fiscal consolidation period – Denmark, 1983–6; Ireland, 1986–9; Sweden, 1986–7. These three episodes, characterized by booming consumption growth, strong output growth and falling unemployment rates, are of most interest for our purposes since they seem to fit the pattern predicted by the EFC hypothesis. Can we identify structural aspects of these three fiscal consolidations (components of taxes and expenditures changed) that set them apart from the other cases? Giavazzi and Pagano (1996), for example, argue that a sharp decline in government expenditure is an important component of a successful EFC. Table 8.2 shows that fiscal consolidation was achieved in Ireland by reducing taxes and, to a much greater extent, decreasing government expenditures. However, Ireland is the exception. Denmark and Sweden undertook a combination of tax hikes and government expenditure reductions to achieve fiscal consolidation. Similarly, all of the other consolidation episodes were associated with substantial tax increases and the majority (10 of 15 episodes) with substantial government expenditure decreases. Alesina and Perotti (1996b) argue that EFC episodes are likely to be characterized by substantial cuts in government consumption and transfer payments. If we define ‘substantial cuts’ as reductions of the order of 3 per cent of GDP, two of the three expansionary episodes generally fitting the EFC group fall into this category. This is shown in Table 8.2. Denmark
234
Table 8.1
Fiscal consolidation and macroeconomic performance
Fiscal consolidation Actual deficitb
Period of fiscal consolidationa Australia Belgium Denmark Finland Germany Greece Ireland Italy
1980–2 1986–8 1982–7 1993–5 1983–6 1975–6 1980–5 1994–5 1982–4 1986–9 1976–7 1991–5
Year prior to consolidation –2.2 –2.8 –13.1 –7.1 –9.1 5.1 –2.6 –14.2 –12.8 –10.7 –12.9 –10.9
Macroeconomic development during consolidation
Structural deficitc
Last year of consolidation
Year prior of consolidation
–0.5 1.0 –7.6 –4.4 3.4 8.1 –1.2 –9.2 –9.3 –1.8 –8.6 –7.2
–2.6 –2.9 –13.8 –7.7 –8.1 3.7 –4.4 –13.3 –14.3 –9.3 –11.7 –11.6
Change in structural deficit during period of consolidation
Average real GDP growth rate per cent per annumd
Change in inflation rate Percentage pointsa
Change in umemployment rate Percentage pointsa
Change in current account balance/ GDP Perentage pointsa
Average real consumption growth rate per cent per annumd
Last year of consolidation 0.5 0.7 –5.8 –3.0 1.4 8.3 –0.2 –8.3 –7.8 –0.4 –8.4 –6.6
3.1 3.6 8.0 4.7 9.5 4.6 4.2 5.0 6.5 8.9 3.3 5.0
1.8 3.6 1.4 0.8 3.7 0.7 1.1 1.7 2.1 3.6 4.9 1.2
1.2 2.3 –2.4 –1.3 –6.0 –9.7 –1.7 –1.7 –11.0 0.2 2.2 –2.8
1.0 –1.0 1.2 2.6 –2.0 2.2 4.8 0.3 5.6 –2.4 1.0 2.9
–2.7 1.4 6.4 3.1 –1.5 1.4 3.4 –1.7 8.0 2.1 1.4 4.1
3.5 2.5 1.4 0.7 4.2 0.4 0.7 1.6 –1.4 4.0 4.2 0.9
Table 8.1
(Cont’d)
Fiscal consolidation Actual deficitb
Period of fiscal consolidationa Japan Sweden United Kingdom
Year prior to consolidation
Last year of consolidation
Macroeconomic development during consolidation
Structural deficitc
Year prior of consolidation
Change in structural deficit during period of consolidation
Average real GDP growth rate per cent per annumd
Change in inflation rate Percentage pointsa
Change in umemployment rate Percentage pointsa
Change in current account balance/ GDP Perentage pointsa
Average real consumption growth rate per cent per annumd
Last year of consolidation
1980–7 1986–7
–4.7 –3.8
0.5 4.2
–5.6 –5.0
1.9 1.8
7.5 6.8
3.3 2.7
–2.7 –1.8
0.8 –0.8
4.5 1.3
3.0 4.5
1979–2
–4.4
–2.5
–5.4
–0.1
5.3
0.2
–4.0
5.2
1.0
1.4
a Defined as change in general government structural balance of at least +3.0 percentage points of potential GDP. b As a percentage of GDP. c As a percentage of potential GDP. d During period of consolidation. Source: OECD (1996).
235
236
Table 8.2
Structure of fiscal consolidation and macroeconomic conditions prior to the first year of consolidation Structure of fiscal consolidation
Period of fiscal consolidationa
Change in total revenue as a percentage of GDPb
Of which
Change in total outlays as a percentage of GDPb Transfers
Australia Belgium Denmark Finland Germany Greece Ireland Italy
1980–2 1986–8 1982–7 1993–5 1983–6 1975–6 1980–5 1994–5 1982–4 1986–9 1976–7 1991–5
3.4 1.1 1.6 1.2 7.1 10.6 1.2 2.5 4.2 –2.7 2.0 2.0
1.5 –2.7 –4.0 –1.4 –5.5 7.7 –0.2 –2.5 0.7 –11.7 –2.3 –1.8
0.4 –1.8 –1.0 0.1 –2.3 3.9 –0.1 0.2 2.2 –2.3 –0.7 1.0
Consumption 1.1 –1.3 –2.3 0.2 –4.3 2.8 0.4 0.3 –0.8 –3.2 –0.4 –1.1
Investment
Debt Otherc interest paymentsd
–0.5 –0.7 –1.7 –0.1 –0.5 0.3 –1.1 –0.1 –1.7 –2.0 –0.2 –1.1
0.1 0.4 2.8 –1.5 2.8 0.1 1.3 0.1 1.8 –2.1 0.7 1.1
0.2 –0.3 –1.8 –0.3 –1.2 0.7 –0.5 –3.1 –1.0 –2.2 –1.7 1.4
Macroeconomic initial conditions Output gap percentage pointa
Inflation (GDP deflator) percentage per annum
Current account balance as percentage of GDPc
1.1 0.2 1.0 1.0 –1.7 4.0 3.3 –1.8 2.6 –2.5 –3.9 1.8
10.1 6.0 4.7 3.5 10.6 22.4 3.8 14.0 17.4 5.2 16.4 7.6
–2.0 –5.4 –4.4 3.0 –4.0 –5.1 –0.7 –0.8 –13.5 –3.4 –0.4 –1.6
Table 8.2
(cont’d)
Structure of fiscal consolidation Period of fiscal consolidationa
Change in total revenue as a percentage of GDPb
Of which gap
Change in total outlays as a percentage of GDPb Transfers
Japan Sweden United Kingdom
Macroeconomic initial conditions
1980–7 1986–7
6.2 2.6
1.0 –5.5
1.7 –0.1
1979–2
5.1
3.1
2.1
Consumption –0.3 –1.2 1.8
Output percentage pointa
Investment
Debt Otherc interest paymentsd
–1.3 –0.2
1.8 –1.9
–1.2
–0.9 –2.1 0.8
2.2 1.6 –0.4
2.2
Inflation (GDP deflator) percentage per annum
2.8 6.6 11.6
Current account balance as percentage of GDPc
–0.9 –1.4 0.7
a Defined as change in general government structural balance of at least +3.0 percentage points of potential GDP. b From year prior to consolidation to final year of consolidation. c Current transfers and social security items received, property and entrepreneurial income. d For some countries this item includes other property and entrepreneurial income paid. Source: OECD (1996).
237
238
Bergman and Hutchison
and Ireland (1986–9) reduced total government consumption and transfer payments by 6.5 and 5.5 per cent of GDP, respectively. Sweden cut these spending categories by much less (1.5 per cent of GDP). It is noteworthy that Australia (1986–8) and Belgium (1982–7) also significantly reduced consumption and transfer payments during fiscal consolidation episodes (on the order of 3.1 and 3.3 per cent of GDP, respectively). Australia experienced a strong economic expansion during this episode, but it was not led by a consumption boom. Belgium also does not fit the EFC hypothesis: real GDP and consumption growth rates were slow (1.4 per cent on average per annum) and unemployment rose by 1.2 basis points. It is also difficult to observe common elements in the initial macroeconomic situations (that is, prior to the fiscal contraction) of the three expansionary cases fitting the EFC model which help to distinguish them from the other episodes. This is also shown in Table 8.2. In two of the three episodes, albeit the most dramatic cases (Ireland and Denmark), the economies were operating at levels well below potential prior to consolidation – that is, output gaps were large in magnitude and negative. However, the other expansionary episode fitting the pattern predicted by the EFC hypothesis (Sweden) was associated with an economy operating at a level well above potential output prior to fiscal consolidation. Other than Denmark and Ireland, two other fiscal consolidations in the sample – Italy (1976–7) and Greece – were preceded by weak economies. While Italy had a robust expansion during the fiscal consolidation (despite a small rise in unemployment), Greece had a very weak economy. No clear patterns in other initial conditions are evident which clearly distinguish the ‘EFC episodes’ of Denmark, Ireland and Sweden from the other cases. Every country in the sample had moderate to high inflation rates and all but two had current account deficits. Denmark had relatively high inflation and a substantial current account deficit position prior to its fiscal consolidation. Both Ireland and Sweden had moderate inflation rates prior to consolidation, but only Ireland had a substantial current account deficit. In summary, these statistics suggest that common characteristics among the fiscal episodes are not easy to identify. Significant fiscal consolidations were generally associated with weak economies, but not uniformly. Only three economies experienced strong consumption-led output expansions during the consolidations, seeming to fit the prediction of the EFC hypothesis. Nor do economic conditions prior to the reform – state of the business cycle, inflation, or current account position – help predict the path of the economy during the fiscal consolidation stage. The composition of the fiscal consolidation may be important, as the two most dramatic examples thought to support the EFC view (Denmark and Ireland) were associated with substantial cuts in government consumption and transfers. However,
Economic Expansions and Fiscal Contractions
239
other countries significantly reducing these expenditure categories do not fit the EFC pattern, either, because there was not a strong economic expansion (Belgium, 1982–7) or because the economic expansion was not led by a consumption boom (Australia, 1986–8). Existing Empirical Research Despite this diversity of experience, Giavazzi and Pagano (1990, 1996), Bertola and Drazen (1993) and others have argued that severe fiscal contractions in Denmark and Ireland are primarily responsible for the consumption booms and economic expansions experienced at the time. Giavazzi and Pagano present empirical evidence, based primarily on unexpectedly high consumption (residuals from empirically estimated consumption functions) in these countries which appears correlated with unanticipated changes in permanent government consumption, leading them to conclude that ‘there are cases in which the German view has a serious claim to empirical relevance’ (Giavazzi and Pagano, 1990, p. 105). Alesina and Perotti (1996a, 1996b) emphasize the importance of composition to the success of a fiscal consolidation. They argue that fiscal contractions relying primarily on spending cuts (‘type 1’), especially when concentrated on government wages and transfer payments, are more likely to be permanent and therefore have a larger expansionary effect on consumption and output. By contrast, they suggest that fiscal contractions relying primarily on tax increases (‘type 2’), particularly on households, and decreases in public investment expenditures, tend to be reversed by further deterioration in the budget and have worse macroeconomic consequences. To investigate this hypothesis they employ a sample of 20 OECD countries for the period 1960–94, and observe that type 1 fiscal contractions tend to have a higher probability of success than type 2 fiscal contractions (see Alesina and Perotti, 1996b). Here they define ‘success’ as either a significant improvement (2 per cent of GDP) in the cyclically adjusted general government fiscal deficit or a significant reduction in the debt to GDP ratio (5 per cent of GDP). Alesina and Perotti (1996b) also present summary statistics, but no formal tests, indicating that average real output growth (unemployment) is higher (lower) during and after successful fiscal contractions than with unsuccessful fiscal contractions. Indirectly, this indicates that fiscal contractions identified as type 1 are more likely to be associated with economic expansions. They point out that it is the growth of private investment which is driving these empirical observations, as investment booms are associated with the successful fiscal contractions. The growth rate of consumption, by contrast, does not seem to vary with the two types of adjustments. This evidence suggests that a successful fiscal contraction may lead to an economic expansion because of an investment boom rather than a consumption boom.
240
Bergman and Hutchison
A number of individual case studies have been critical of the EFC hypothesis as posited by Giavazzi and Pagano (1990). Several hypotheses have been suggested to explain the Danish experience. Christensen (1987) finds that consumption in Denmark is very sensitive to changes in interest rates. The fall in interest rates at the time, partly attributable to the new hardcurrency policy adopted by the Danish authorities may explain a large part of the sharp rise in consumption during 1984–6. This point is emphasized by Andersen (1994), who notes that the Danish long-term interest differential against Germany fell sharply (1.6 percentage points) the first trading day following parliamentary approval of the fiscal reform in October 1982 and again in February 1983 (1.2 percentage points) immediately after an important labour market agreement. (It is noteworthy, however, that the most substantial decline in the long-term differential occurred gradually during the 3 years prior to announcement of the fiscal reform.) Andersen (1994) also points to the rapid expansion in consumer credit, presumably related to liberalization of financial markets, which supported the rise in consumption demand during the Danish fiscal consolidation. There is some ambiguity about the effect of interest rates on consumption at this time, however. Nominal interest rates – both long- and short-term – declined markedly between 1982 and 1986. The long-term (short-term) interest rate declined from 21.4 per cent (16.8 per cent) to 10.1 per cent (9.1 per cent) during this period. In addition, the ex post real interest rate, calculated before tax, also declined somewhat. The after-tax real interest rate increased, however, between 1982 and 1986. In particular, the ex post shortterm real interest rate net of taxes, using an estimated average marginal tax rate on interest income of 56 per cent, increased from –2.8 per cent in 1982 to 1.5 per cent in 1986.4 Similarly, the long-term real interest rate net of taxes also increased substantially. This result holds if the assumed tax rates are varied quite substantially. Other themes have also been emphasized. Andersen and Risager (1990) point to the importance of increased Danish cost competitiveness associated with labour market policies, undertaken in conjunction with the fiscal reform, which significantly moderated real wage growth. A number of authors have pointed to the good timing of the fiscal reform. Nielsen and Søndergaard (1991), for example, partly attribute the sharp rise in aggregate demand during 1983–6 to a delayed response of private investment and employment to the significantly improved competitiveness and profitability situation that was brought about by the effective devaluation prior to the fiscal reform. The complexity and interaction of these factors working together in creating the 1983–6 Danish economic expansion is emphasized by Jensen (1996). The Irish experience has also attracted considerable international attention. Alesina and Perotti (1996b) find Ireland’s second fiscal consolidation to
Economic Expansions and Fiscal Contractions
241
be most supportive of their theory since improvement in government finances was entirely attributable to sharp reductions in government expenditure, much of which was concentrated on wages and transfer payments. The Irish case is also one of the showcase examples of the ‘German view’ highlighted by Giavazzi and Pagano (1990). Focusing on the second fiscal stabilization, however, Barry (1991) and Barry and Devereux (1994, 1995) take issue with the EFC explanation of the Irish economic expansion. Barry (1991) argues that other factors were working to support a strong economy at the time – buoyant world demand, improvements in cost competitiveness and an inflow of foreign capital. He concludes that the EFC theory application to Ireland ‘is belied by the behavior of virtually every important macroeconomic aggregate’ (p. 39). Dornbusch (1989) also sharply disagrees with the general view that tight monetary and fiscal policies were virtually costless. Considering the combined effects of the two Irish stabilization programmes, he considers them a failure in at least two important respects (high level of unemployment and debt–GDP ratio in 1988 compared to 1980–2).
8.4
THE DANISH FISCAL REFORM
In Denmark, as elsewhere, the years immediately following the second oil price shock were characterized by deteriorating economic performance. Weak output growth in the 3 years prior to 1982 were accompanied by high and rising unemployment, a large gap between actual and potential output, strong inflationary pressures, a large current account deficit and a widening of the government budget deficit to one of the highest levels among the industrialized countries (see Tables 8.1 and 8.2). In late 1981 the minority Social Democratic government, failing to gain parliamentary approval for its economic stabilization programme, called for a new general election. A new Social Democratic minority government was formed in December, and the following June, its economic programme was approved by Parliament (the so-called ‘March package’). This programme focused mainly on job creation measures (including a job creation scheme subsidizing wages of new employment for young people, support to new apprentice jobs, increased support for reeducation of long-term unemployed and additional funds for municipalities to combat youth unemployment). The ‘March package’ also increased indirect taxes on a large range of luxury goods. Earlier in the year, the new government also had devalued the Danish Kroner by 3 per cent against other EMS currencies excepting Belgium (which devalued 8.5 per cent).5
242
Bergman and Hutchison
The economic situation continued to be difficult, however, and new measures were again proposed to Parliament in early Fall. This proposal was rejected by Parliament, however, and the new Social Democratic government resigned on 3 September 1982 after less than a year in office. A minority conservative government consisting of the Liberals, the Conservatives, the Centre Democrats and the Christian People’s Party was then formed, and its far-reaching economic reform measures approved by Parliament in October. The basic outlines of Danish fiscal adjustments following the October announcement (to 1989) are shown in Table 8.3. The policy objects of the new conservative government were broadly similar to the out going Social Democrats with the main emphasis placed on the need to improve international ‘competitiveness in order to reduce progressively the external constraint and restore the basis for renewed growth and higher employment in the longer run’ (OECD, 1983, p. 24). However, the mix of instruments were different from the previous government in that improved international competitiveness was to be achieved by moderation of domestic prices and costs rather than through devaluations, and a reduction in the budget deficit was assigned a key role. The basic premise of the new programme was the idea that the problems of the Danish economy – high unemployment, large public sector and current account deficits and high inflation – could be solved only by introducing a medium-term policy orientation designed to expand the private sector in an internationally competitive environment. The immediate objective focused on the current account, leaving responsibility for employment to the labour market, and the main policy tools were a fixed exchange rate policy and tight fiscal policy. The compromise on the new policy package reached in October was seen as a medium-term approach to policy actions – that is, an internally consistent fiscal and budgetary policy strategy which would be applied over a number of years (see OECD, 1983, 1984).
Income Policy A number of incomes and fiscal policy measures were central features of the new policy package. The incomes policy measures included: (1) suspension of all indexation of wages, salaries and transfer incomes until 1985 (excepting for old-age and disability pensions); (2) limitation of pay increases in the public sector to 4 per cent per year, and the government indicated that it was prepared to seek tax cuts if the private sector wage settlement was similarly restrained; (3) abolition of formal wage–wage links; (4) limitation on dividends and bonuses to the 1982 level; (5) a freeze on the maximum amount of unemployment and sickness benefits, as well as eliminating the payment of sickness benefits for the first day of sickness
Economic Expansions and Fiscal Contractions
243
Table 8.3
Danish fiscal adjustments following the October 1982 announcement
Date
Announcement
1982:3/9
The Social democratic Government resigns
1982:10/9
A new minority right-wing coalition government (the Conservative, the Liberals, the Centre Democrats and the Christian Party) is formed under the leadership of the conservative Poul Schlüter.
1982:16/10
Economic–political compromise:
Planned fiscal action
Income – political, including: – Suspension of index-regulated rises for public employees (with effect from the regulations that take place after October 1983) – Suspension of ‘high-price regulation’, indexation of wages, salary and transfer income (until February 1985). – Profits freeze of goods sold in the country (private sector) – A total wage stop (from October 1982 until end-February 1983) Fiscal-political, including: – Suspension of price indexation of some social payments (1983 and 1984) – The first day of sickness not covered by Social insurance (from April 1983) – The contribution to unemployment insurance raised from 3.75 to 6 times the benefit rate for wage salary earners and from 3.25 to 11.25 times for employers – Rules for supplementary unemployment insurance benefits tightened – Retirement pensions set according to income level – Price-indexation of income-taxes and personal allowances suspended – The income tax ceiling raised from 70 to 73% – Wealth tax deduction raised to 1.158.900 kr.
1983:9/3
New agreement (on the labour market) is accepted
Bergman and Hutchison
244
Table 8.3
(Cont’d)
Date
Announcement
Planned fiscal action
1983:1/6
Tax on pension funds, life insurance, etc. introduced; real return on bond holdings exceeding 3.5% taxed away (for the aggregate pension fund sector)
1984:3/11
General election: the right-wing minority government continues
1984:17/5
Economic–political compromise
Including: – Suspension of ‘high-price regulation’ continued until 1987 – ‘Sickness Fund’ contribution increased from 0.15% to 2% – Unemployment insurance contribution increased for wage salary earners
1985:31/3
Economic–political compromise: on the background of the breakdown of the general agreement negotiations
Among other things the working week reduced 1 hour (with wage compensation) and the wage of public employee raised by 1.75 and 1.25% during the next 2 years: – Corporate tax raised from 40 to 50% – Persons younger than 67 years with income above 150.000 crowns in 1984 are requested to make a ‘bounded saving’ for a limited period
1986:27/2
‘EF-pakken’ is adopted (the EC-package)
1986:18/3
Tax reform
Income divided into personal income and capital income Personal income taxed progressively whereas capital income is (almost) taxed proportionally: – Marginal tax rates reduced from 72 to 68 per cent for personal income and to approximately 50 for capital income
1986:21/3
Raised duties from April 1986
Duties are raised on energy and luxury goods
1986:4/6
Wage Indexation Law
‘High-price’ wage indexation removed entirely
1986:17/10
The Potato Diet:
– 20% tax on consumer credit interest payments imposed And among other things: – Tightening of regulation regarding credit purchases
Economic Expansions and Fiscal Contractions Table 8.3
245
(Cont’d)
Date
Announcement
Planned fiscal action
1986:17/10
The Potato Diet:
– Changes in law about mortgage loans, effectively reducing tax credits for interest deductions – Increased energy duties.
1987:2/3
New general agreement on the private labour market 4-years agreement Working week gradually reduced to 37 hours (wage compensated), agreement on wage rates is limited to two years (Economic Council estimates hourly wages will rise 6.5% and 5% in the two years)
1987:1/4
Agreement on the public labour market 4-years agreement, with wages renegotiated in 2 years – Working week reduced to 37 hours over 2-year period – Wage rise of 4% the first year – Regulation Parity arrangement with respect to wage increases in the private sector
1987:8/9
General election: the right-wing minority government continues
1987:8/12
Political compromise
1988:10/5
General election: (The NATO election)
1988:3/6
A new minority government is formed with the Social–Liberal Party, Liberals and Conservatives. Poul Schlüter continues as prime minister
1988:1/10
Liberalization of capital movements ‘EF’s kapitaliberaliseringsdirektiv’
1989:1/12
The tax on consumer credit interest payment eliminated
Reduction in employer responsibility for duration of employee sickness payments.
Source: Økonomisk-politisk kalender, Statistiks ti-års oversigt, Statistics Denmark
246
Bergman and Hutchison
(in order both to enhance work incentives and to reduce public spending); and (6) a temporary freeze on wages and profit margins until March 1983. Fiscal Measures The fiscal policy measures included a number of provisions to reduce public expenditures and increase revenues in order to improve the fiscal deficit. In addition, several of the incomes policy measures also reduced government outlays. Tax increases included: (1) a rise in social security contributions relating to the unemployment scheme; (2) a temporary tax on wealth of tax free pensions with the intent of replacing it in the following year by a 1 permanent taxation of yields exceeding 3 2 per cent in real terms; (3) abolition of the lagged indexation of tax schedules and deductions (but adjustments were made to take account of the expected slowdown in inflation in 1983); and (4) an increase in the maximum tax ceiling from 70 to 73 per cent. Measures to limit the growth of government expenditures included the incomes policies actions noted, having the effect and operating both on the public sector wage bill and on public payments for unemployment and illness, and a number other measures to limit discretionary expenditures. Official Estimates of the Impact of Fiscal Contraction Ministry of Economics’ estimates prior to the fiscal reform measures predicted continued deterioration in the central government financial position, with the current investment and lending accounts rising from about 1 12 per cent of GDP in 1982 to 15 2 per cent in 1983. These forecasts, issued in September 1982, were conditional upon no policy change and a weaker international environment (see OECD, 1983). It was widely anticipated that the new stabilization policy would have a short-run contractionary effect on the economy. GDP forecasts for 1983–5 issued by the Economic Council, for example, were revised downwards after announcement of the reform programme. Table 8.4 shows that forecasts issued in May 1982 forecast 3.2 per cent GDP growth in 1983 and 3.7 per cent average growth in 1984–5. In October, the Council forecast 2.0 per cent growth in 1983 and 3.0 per cent average growth in 1984–5. Although the fiscal contraction was expected to slow the economy, and deteriorate the cyclical part of the budget, it was nonetheless anticipated that the overall fiscal position would improve. The direct effects of the fiscal stabilization programme were estimated to improve central government 1 finances by about Kr. 21 2 billion in 1983 (Kr. 10 billion from direct expenditure cuts, Kr. 5 billion due to tax increases and most of the Kr. 1 6 2 billion remainder due to lower wages and price increases in the public sector and to reduced real transfers). Given the anticipated cyclical
Economic Expansions and Fiscal Contractions Table 8.4
247
Forecasts published by Economic Council, 1982–7
Economic prospects in Denmark: Economic forecast, report publication date
Expectations (year)
1982:5
1982 1983 1984 1985
9.7 9.5 8.8 8.2
3.3 3.3 4.0 4.6
1.5 1.9 –2.2 4.2 3.2 –4.7 3.7 3.7 10.3 (av 1984–5) (av 1984–5) (av 1984–5)
1982: 10
1982 1983 1984 1985
9.8 10.2 10.4 10.2
5.0 5.0 4.7 4.8
2.0 1.5 2.9 1.6 2.0 1.1 2.4 3.0 5.5 (av 1984–5) (av 1984–5) (av 1984–5)
1983: 5
1982 1983 1984
9.9 10.5 11.9
5.1 3.2 2.3
2.5 1.2 1.8
2.2 2.1 1.3
12.0 5.4 1.2
1984: 11
1982 1983 1984 1985
10.5 11.2 11.0 10.9
9.9 10.5 10.4 10.3
4.7 2.5 3.6 2.8
2.2 2.8 2.4
2.9 4.5 3.0
1.0 10.7 12.9
1984 1985 1986 1987
10.7 9.5 9.3 9.3
10.0 8.9 8.7 8.7
3.5 4.4 3.2 2.3
2.5 3.5 3.0
2.0 3.4 2.7
15.7 6.4 1.1
1984 1985 1986 1987
10.7 9.7 9.1 9.1
10.0 9.1 8.5 8.6
3.6 5.4 4.2 2.0
3.5 4.0 2.3
3.1 2.2 1.5
21.5 6.2 –7.0
1985 1986 1987 1988
9.5 8.2 9.6 11.4
9.0 7.6 8.8 10.4
5.4 6.1 2.7 0.7
4.0 –0.8 –0.2
2.7 –0.5 –0.3
4.5 –9.4 –11.7
1985 1986 1987 1988
9.6 8.3 9.4 11.6
9.0 7.8 8.8 10.9
5.6 6.3 3.1 1.5
4.8 –1.2 –1.8
3.3 –0.6 –1.2
18.1 –9.0 –11.7
1986 1987 1988 1989
8.4 8.4 10.4 11.5
7.8 7.9 9.8 10.8
6.2 3.2 2.1 2.8
–2.2 –1.3 1.0
–1.2 –1.2 0.2
–6.3 –9.9 –0.7
1985: 12
1986: 5
1986: 11
1987: 5
1987: 12
Unemployment full time (%)
total (%)
Current account deficit percentage of GDP at factor cost
Private consumption (real growth) (%)
Source: Dansk Økonomi (Det Økonomiske Råd), various issues.
GDP (real growth) (%)
Private investment (business) (real growth) (%)
Bergman and Hutchison
248
deterioration in the budget balance, the official estimate put the net improvement in the investment and loan account balance at around Kr. 12 billion or 2.4 per cent of GDP (in comparison with the ‘no-policy-change budget’ from September 1982). Although an improvement, the official estimates still had forecast a substantial rise in the central government deficit from a year earlier (see OECD, 1983).
8.5
EVALUATION
Expectations of Higher Permanent Disposable Income? The announcement of the Danish fiscal reform in October 1982 had a number of elements which may have led to a significant upward revision in expectations over the level of future permanent disposable income. The reform programme was announced less than a month after the formation of a new Conservative Government and viewed as a substantial departure from previous policies – that is, it was ‘news’ warranting a revision in expectations. Moreover, a basic tenet of the programme was to encourage private sector expansion and limit public sector expenditures. Most of the direct effects on the stabilization package in improving the budget were anticipated to come from government expenditure reductions (about three-quarters of the total), a significant portion of which was because of reduced transfers and lower wage increases in the public sector. Permanent reductions in government consumption (especially the wage bill) and cutbacks in social expenditure in the form of transfer payments to households are frequently argued to be essential parts of a fiscal stabilization package if it is to have an expansionary effect on the economy (see Giavazzi and Pagano, 1996; Alesina and Perotti, 1996a, 1996b). However, substantial increases in taxes were also an integral part of the programme, most of which were viewed as permanent, including hikes in social security contributions, taxation of higher-yielding pensions (formerly tax-free) and an increase in the maximum income tax rates. Tax increases of this type would not normally lead households to revise upward their expectations of future disposable income. It is possible for a tax increase to have an expansionary effect on consumption if it generates expectations of less dramatic, less disruptive and less distortionary future tax increases. But this is not obvious. Nonetheless, Blanchard (1990) suggests that Danish (and Irish) consumers may have viewed the consolidation moves as an indication that even more punitive tax rates would be avoided in the future. In the event, the Danish budgetary position improved much faster than had been anticipated. The actual budget moved from a 9.1 per cent of GDP deficit in 1982 to a 3.4 per cent surplus in 1986. During the same period the
Economic Expansions and Fiscal Contractions
249
structural budget position (also shown in Table 8.1) moved from 8.1 per cent deficit to a 1.4 per cent surplus. Controlling for the business cycle, Denmark’s fiscal consolidation improved government finances by 9.5 per cent of GDP – the largest fiscal consolidation among the industrialized countries within the last 20 years. The improvement in government finances (not cyclically adjusted) was achieved by raising total government revenues by 7.1 per cent of GDP and decreasing government expenditures by 5.5 per cent of GDP. The lion’s share of the rise in revenue was attributable to direct tax increases (4.0 per cent of GDP) and most of the decline in expenditure was due to a reduction of government consumption (4.3 per cent of GDP). Government transfers also declined by 2.3 per cent. It is noteworthy that the sharp decline in government outlays came about despite a sharp rise in net interest payments on government debt (2.8 per cent of GDP). Surprisingly Strong Economic Performance Consumption, investment and GDP growth were much stronger than anticipated following the fiscal consolidation. The Economic Council, in its October 1982 forecast, predicted that average private consumption growth (GDP growth) would be around 2 per cent (2.6 per cent) during the 1983–5 period. Private consumption, in fact, averaged over 4 per cent and GDP growth averaged 3.7 per cent during 1983–6 (Table 8.2). Although the 2.6 per cent private consumption growth in 1983 was substantially better than the previous 3 years (consumption fell in 1980–1 and grew only 1.4 per cent in 1982), it increased further during 1984–6 (peaking at 5.7 per cent in 1986). This is consistent with uncertainty over the fiscal reform announcement, which initially may have led to a relatively cautious consumption response. As the particulars of the programme were implemented, the fiscal reform may have gained credibility, causing consumption growth to increase further. Although consumption was very strong following fiscal consolidation, the growth of private fixed investment was even more surprising and impressive. The Economic Council, in its October 1982 forecast, anticipated average private business investment growth of around 1 per cent in 1983 and 1 5 2 per cent in 1984. In the event, business gross fixed investment (non1 residential) grew 2.6 per cent in 1983, 10 2 per cent in 1984 and averaged 18 per cent per annum during 1985–6. Alternative Explanations The short-term economic boom following the Danish fiscal expansion was impressive and largely unanticipated. Analysts had anticipated that fiscal consolidation would weaken the economy in the short term, but would lay the foundation for better medium- and long-term growth prospects. The
250
Bergman and Hutchison
Danish economy went through a surprisingly strong phase during 1983–6 as the result of a consumption and investment boom. The jump in consumption, however, was viewed with some alarm as it contributed to a widening current account deficit and indications of renewed price and cost prices were evident. A key objective of the 1982 fiscal reform – reducing the external deficit – was never satisfactorily met. A series of additional policy measures were adopted, including the October 1986 ‘potato diet’ measures, in order to reduce aggregate demand and slow consumption growth. The economy entered into recession in late 1986, which lasted until mid-1988. However, both the budget position and the current account improved sharply. In particular, Denmark reported surpluses in both the total budget (in 1987 and 1988) and in the current account (after 1990). 8.6 STRUCTURAL VAR-EVIDENCE ON THE 1983–6 DANISH EXPANSION This section presents empirical results concerning the EFC hypothesis and the Danish 1982 fiscal reform found in Bergman and Hutchison (1996). They apply a structural vector autoregression (VAR) analysis to study the effects of government consumption changes on private consumption controlling for other domestic and foreign influences.6 Model Specification Two broad categories of economic disturbances are identified, those having permanent effects (usually interpreted as supply-side shocks such as shifts in labour productivity or technology) and those having only transitory effects (usually interpreted as aggregate demand shocks). In particular, we assume that the joint behaviour of output, government consumption, private consumption and the terms of trade can be approximated with a VAR model. The behaviour of these four variables is governed by four so-called ‘structural shocks’. Three of these shocks can be characterized as permanent shifts in government consumption, income and in the terms of trade. These three shocks determine both the long-run and the short-run behaviour of the four variables. The fourth shock is a composite transitory disturbance reflecting short-run aggregate demand considerations and business-cycle effects. This shock will, however, affect the variables only in the short to medium run. Bergman and Hutchison (1996) estimate the model and obtain theoretically plausible results. An unexpected permanent increase in government consumption causes a sharp decline in private consumption; unexpected permanent terms of trade improvements increase private consumption; and
Economic Expansions and Fiscal Contractions
251
unexpected demand disturbances affect private consumption over a 3-year time horizon. Uncertainty and learning about the permanent or transitory nature of government consumption plans will affect both the initial effect and the dynamics. Explaining the 1983–6 Danish Consumption Boom The objective is to test whether the EFC hypothesis is the main explanation for the consumption boom in Denmark following the 1982 fiscal reform. There are two main questions: Was the consumption boom in Denmark following the fiscal reform entirely unexpected? Was the decline in government expenditure primarily responsible for the consumption boom? This experiment (historical decomposition) is performed in an unrestricted system where the only restrictions imposed are those needed to identify the shocks. The dashed line in Figure 8.1 shows the forecast value of consumption from the structural VAR model given all of the available information from 1982: Q2 (deterministic terms and lagged values of the four dependent variables in the system). The solid line is actual consumption and the difference between the two lines indicates ‘unexpected’ consumption – that is, consumption forecast errors. This figure shows that Danish consumption was unexpectedly low immediately following the announcement of the fiscal programme, but soon rebounded and was above the forecast level (based on information prior to the fiscal reform) by the latter part of 1983. This initial Figure 8.1 5.525
Actual and expected private consumption, 1982:3–1987:4 Actual Expected
5.500 5.475 5.450 5.425 5.400 5.375 5.350 5.325 1982
1983
1984
1985
1986
1987
Bergman and Hutchison
252
overprediction of consumption, which is inconsistent with the ‘German view’ of the Danish experience, is also noted by Andersen (1994). However, the consumption boom from that point on, for the most part ‘unexpected’, is clearly evident. Did the permanent fall in (expected) government consumption expenditure cause the consumption boom? Figure 8.2 shows the historical decomposition of consumption during the five years following the Danish contraction into the various disturbances. The solid line is the unexpected part of consumption (the difference between the actual and forecasted values shown in Figure 8.2) and the dashed line is the effect on consumption due to the structural shock noted. The immediate downturn may be attributable to several factors, but most important for our purposes is that the permanent government consumption expenditure shock had a negative effect on consumption immediately following announcement of the fiscal reform programme.
Figure 8.2 Historical decomposition of unexpected private consumption (solid line) attributable to structural shocks (dashed line), 1982:3–1987:4 0.112
permanent terms-of-trade shocks
0.112
0.086
0.086
0.080
0.080
0.064
0.064
0.048
0.048
0.032
0.032
0.016
0.016
0.000
0.000
–0.016
0.108
1982 1983
1984
1985
1986
1987
permanent income shocks
–0.016
0.125
0.090
permanent government consumption shocks
1982 1983
1984
1985
1986
1987
transitory income shocks
0.100
0.072
0.075
0.054 0.050 0.036 0.025
0.018
0.000
0.000 –0.018 1982 1983
1984
1985
1986
1987
–0.025
1982 1983
1984
1985
1986
1987
Economic Expansions and Fiscal Contractions
253
Although current and lagged effects arising from the current and anticipated permanent government consumption expenditure reduction played a role in the sharp upturn in consumption at the end of 1983, the effect was shortlived and soon dominated by other factors. Only a modest stimulative effect was realized despite several measures to consolidate fiscal finances on a longer-term basis, including changes in budgetary practice and greater control of local fiscal authorities, which should have enhanced the credibility of the policy. By contrast, a strong upward movement in the Danish terms of trade has been a strong positive force on consumption growth since the beginning of the fiscal reform programme and continuously through most of the 1980s. Falling world energy prices in real terms played an important role in the terms of trade gain. Cyclical factors such as the upturn in world growth, led by the strong US economy recovery, also played an important role in 1983 and 1984. Other permanent income shocks, by contrast, tended to dampen consumption through mid-1984, and then became the dominant force pushing up consumption for the next 2 years. These ‘other’ permanent income shocks may include wealth effects associated with the interest rate decline emphasized by Christensen (1987). Interest rate declines in turn may have been attributable to the credibility of the new ‘hard-currency’ policy adopted by the government despite the historical experience of several discrete devaluations (see Andersen and Risager, 1991; Andersen, 1994). Other aspects of the reform programme in Denmark which would tend to boost this part of expected permanent income would include measures designed to increase efficiency in labour markets (for example, suspension of indexation of wages and unemployment benefits) and financial markets (for example, financial market liberalization and deregulation).
8.7
CONCLUSIONS
This chapter reviews the theoretical underpinnings to the expansionary fiscal contraction hypothesis and the international evidence. We look more closely at the Danish 1982 fiscal reform, presenting evidence on its intended magnitude, timing and credibility by comparing the characteristics of the reform and the expectations issued by the Economic Council. We test the EFC hypothesis with Danish data using an econometric structural time-series model. Our analysis aims at examining whether fiscal reforms – and, in particular, the 1982 Danish fiscal reform – are likely to create such high expectations about future tax cuts so as to lead to a consumption boom.
254
Bergman and Hutchison
The international evidence does not provide much empirical support for the EFC hypothesis. Only three out of 15 significant fiscal reforms reviewed in this chapter are associated with macroeconomic outcomes consistent with EFC. And these cases have quite diverse experiences in terms of initial macroeconomic conditions and in the particulars of the fiscal reform programmes. The 1983–6 Danish experience was the largest fiscal adjustment among the OECD countries reviewed. When the programme was announced in October 1982, it was widely anticipated that the economy would slide back into recession as a result of fiscal tightening. Instead, a consumption and investmentled boom occurred. The fiscal reform was also designed to lay the foundation for sustainable long-run growth, with the cost anticipated to be a weak economy in the short run. In the event, the fiscal problem of large budget deficits was solved but the sharp rise in aggregate consumption led to continued current account deficits and ultimately another round of tight policy starting in late 1985 (including the so-called ‘potato diet’ implemented in October of the following year). To some extent policy-induced, hopes for long-run sustainable growth following the 1982 fiscal reform were dashed, and the Danish economy turned down sharply in 1987. More formal empirical work provides some support for the EFC hypothesis, using Danish data, in that a ‘pure’ government consumption decline is more than offset by a short- and long-run rise in private consumption. But upon closer examination, the EFC hypothesis does not explain the Danish consumption boom during 1983–6. The boom in private consumption is mainly explained by a cyclical upswing, favorable terms of trade developments and – most importantly – increases in permanent income associated with supply-side factors.
Notes 1.
2. 3. 4.
This chapter was originally written for the conference ‘Macroeconomic Perpectives on the Danish Economy’ (Hornbæk, 19–20 June 1997). We thank Ninette Pilegaard Hansen for research assistance and Frank Barry, Ole Risager and Anders Møller Christensen for valuable comments. This is discussed in detail by Barry and Devereux (1994, 1995). They develop a neo-Keynesian general equilibrium model in order to explore simultaneously the supply- and demand-side effects of a fiscal contraction. This consumption function is called the permanent-income consumption function. The assumed average personal marginal tax on interest income of 56 per cent is an estimate from Sørensen (1988). In 1986, the rate was reduced to 50 per cent (see Sørensen, 1988).
Economic Expansions and Fiscal Contractions 5.
6.
255
The EMS also had a realignment in June, whereby the intervention rates for the French franc and Italian lira were devalued and the German D-mark and Dutch gilder were revalued, but the Danish Kroner central rate was not changed. For a detailed description of the econometric methodology behind the results presented here, see Bergman and Hutchison (1996). For a general discussion of identification in cointegrated VAR models, see King et al. (1991), Quah (1994) and Warne (1993).
References Alesina, A. and R. Perotti (1996a) ‘Reducing budget deficits’, Swedish Economic Policy Review, 3, pp. 113–34, with comments by Torsten Persson. Alesina, A. and R. Perotti (1996b) ‘Fiscal adjustments in OECD countries: composition and macroeconomic effects’, NBER Working Paper, 5730. Andersen, T.M. (1994) ‘Disinflationary stabilization policy: Denmark in the 1980s’, in J. Åkerholm and A. Giovannini (eds), Exchange Rate Policies in the Nordic Countries (Cambridge: Cambridge University Press for the Centre for Economic Policy Research). Andersen, T.M. and O. Risager (1990) ‘Wage formation in Denmark’, in L. Calmfors (ed.), Wage Formation and Macroeconomic Policies in the Nordic Countries (Oxford: Oxford University Press). Andersen, T.M. and O. Risager (1991) ‘The role of credibility for the effects of a change in the exchange-rate policy’, Oxford Economic Papers, 43, pp. 85–98. Barry, F.G. (1991) ‘The Irish recovery 1987–90: an economic miracle?’, The Irish Banking Review (Winter), pp. 23–40. Barry, F.G. and M.B. Devereux (1994) ‘The macroeconomics of government budget cuts: can fiscal contractions be expansionary?’, in W. Scarth and W. Robson (eds), Deficit Reduction: What Pain, What Gain? (Ontario: C.D. Howe Institute). Barry, F.G. and M.B. Devereux (1995) ‘The “expansionary fiscal contraction” hypothesis: a neo-Keynesian analysis’, Oxford Economic Papers, 47, pp. 249–64. Bartolini, L., A. Razin and S. Symansky (1995) ‘G-7 fiscal restructuring in the 1990s: macroeconomic effects’, Economic Policy, 20, pp. 125–60. Bergman, U.M. and M.M. Hutchison (1996) ‘The “German view”, fiscal consolidation and consumption booms: empirical evidence from Denmark’, EPRU Working Paper, 96–10. Bertola, G. and A. Drazen (1993) ‘Trigger points and budget cuts: explaining the effects of fiscal austerity’, American Economic Review, 83, pp. 11–26. Blanchard, O.J. (1990) ‘Comments on Giavazzi and Pagano’, in O.J. Blanchard and S. Fischer (eds), NBER Macroeconomics Annual 1990 (Cambridge, Mass.: MIT Press). Blanchard, O.J. and S. Fischer (1989) Lectures on Macroeconomics (Cambridge, Mass.: MIT Press). Christensen, A.M. (1987) ‘Indkomst, formue og privatforbrug’ [‘Income, wealth and private consumption’], Working Paper, Danmarks Nationalbank. Dornbusch, R. (1989) ‘Ireland’s disinflation’, Economic Policy, 8, pp. 173–210. Giavazzi, F. and M. Pagano (1990) ‘Can severe fiscal contractions be expansionary? Tales of two small European Countries’, in O.J. Blanchard and S. Fischer (eds), NBER Macroeconomics Annual 1990 (Cambridge, Mass.: MIT Press), pp. 75–123.
256
Bergman and Hutchison
Giavazzi, F. and M. Pagano (1996) ‘Non-Keynesian effects of fiscal policy changes: international evidence and the Swedish experience’, Swedish Economic Policy Review, 3, pp. 67–103, with comments by Jonas Agell. Jensen, S.E.H. (1996) ‘Recent developments in the Danish economy: problems, policies and prospects’, EPRU Analyses, 13, Economic Policy Research Unit, Copenhagen. King, R.G., C.I. Plosser, J.H. Stock and M.W. Watson (1991) ‘Stochastic trends and economic fluctuations’, American Economic Review, 81, pp. 819–40. Nielsen, S.B. and J. Søndergaard (1991) ‘Macroeconomic policy and the external constraint: the Danish experience’, in G. Alogoskoufis, L. Papademos and R. Portes (eds), External Constraints on Macroeconomic Policy: The European Experience (Cambridge: Cambridge University Press). Obstfeld, M. and K. Rogoff (1996) Foundations of International Macroeconomics (Cambridge, Mass.: MIT Press). OECD (1982, 1984, 1996) Economic Outlook (Paris: OECD). Quah, D. (1994) ‘Identifying vector autoregressions’, in V. Bergström and A. Vredin (eds), Measuring and Interpreting Business Cycles (Oxford: Clarendon Press). Sørensen, P.B. (1988) ‘Skatteincitamenter til opsparing og investering: en vurdering af firsernes skattereformer’, Nationaløkonomisk Tidskrift, 3, pp. 303–22. Warne, A. (1993) ‘A common trends model: identification, estimation and asymptotics’, Working Paper, 555, Institute for International Economic Studies, University of Stockholm.
Comment Frank Barry Giavazzi and Pagano (1990) propose that the beneficial expectational effects of fiscal contraction on private consumption can, under certain circumstances, dominate the standard demand-reducing effects. They term the resulting domestic-demand-driven boom an episode of ‘expansionary fiscal contraction’ (EFC). These circumstances, they argue, prevailed at the time of the Danish stabilization of 1982 and the Irish stabilization of 1987. That this can happen in theory is undeniable. More important is the question of whether it has happened in practice. Bergman and Hutchison find that conventional factors such as terms of trade and permanent income improvements played a more important role in the Danish consumption boom of the early to mid1980s than did the beneficial expectational effects occasionally associated with fiscal contraction. By ignoring these other factors, Giavazzi and Pagano (1990), they find, overestimated the role played by fiscal contraction in the consumption boom. This is exactly the conclusion I reached in my 1991 analysis of the Irish boom that coincided with Ireland’s fiscal contraction. While their results appear sensible, then (by which I mean of course that they confirm my prejudices!), I nevertheless feel they leave themselves open to two criticisms; one coming from those who believe in the EFC hypothesis, and the other from those who do not. The first weakness appears when we recognize that the authors’ conclusions are unlikely to convince those who believe in EFC. This is because the chapter ignores the non-linearities that recent proponents of the EFC hypothesis have focused upon: they will argue that the impulse response functions upon which the present authors’ analysis is based are different under normal and under crisis conditions. Several models have been advanced that generate this prediction, all of them based on the idea expressed by Blanchard (1987) that fiscal contraction in a crisis situation reduces the likelihood of ‘bad news such as [debt] repudiation, monetization or civil war, in which capital suffers large losses’. In Miller, Skidelsky and Weller (1990) private spending is crowded out through interest rate hikes as a critical level of public debt is approached, since beyond this level government raises taxes on bond holders. Bertola and Drazen (1993) model government spending as a stochastic process with positive drift, so that it becomes unsustainable unless stabilized. A non-linear relationship arises since the public believes that government spending will fall sharply when specific trigger points are reached. The role of expectations over the intergenerational distribution of taxes is emphasized by Sutherland (1997). At low levels of debt the next stabilization is remote for current generations, so fiscal policy has conventional Keynesian effects. At high debt levels, however, an increase in the deficit brings the expected stabilization closer, with strongly adverse effects for current consumers. The essential results from these models is then that fiscal spending will have different effects on expected future taxes when the economy is in crisis. EFC proponents point out that Ireland and Denmark were both clearly in crisis when the fiscal stabilization occurred. In Ireland, the parliamentary opposition resisted the temptation to make political capital out of the cutbacks that the fiscal stabilization entailed; anybody who knows anything of Irish politics will realize how deep the crisis must have been to induce such responsible behaviour! In Denmark, a ‘credit watch’ had been added to the rating of Danish foreign debt immediately before the fiscal adjustment. Giavazzi
257
258
Barry
and Pagano (1990) argue furthermore that the pegging of the Danish and Irish currencies to the D-mark in the wake of the devaluations preceding the stabilization programmes ‘heightened the sense of urgency about the need for a fiscal correction’. While it is admittedly unfair to criticize Bergman and Hutchison for failing to deal with the non-linearities associated with crisis situations, since they set themselves the somewhat narrower task of further exploring the Giavazzi and Pagano analysis, I nevertheless suspect, since the literature has gone in this direction, that proponents of expansionary fiscal contraction may not be convinced by their results. Next we come to the criticism that would be levelled by those hostile to the notion that fiscal contraction can raise aggregate demand. Although fiscal contraction is found not to be the dominant factor behind the consumption boom, the analysis nevertheless finds that ‘a “pure” government consumption decline is more than offset by a short- and long-run rise in private consumption’; thus fiscal contraction in Chapter 8’s model, while not the main cause of the boom, is nevertheless expansionary. It becomes imperative in this case to identify the conditions prevailing in Denmark which may have allowed this to occur.1 That there was an unusual combination of factors in place in Denmark at the time is demonstrated by the results of Whelan’s (1996) examination of the cross-country relationship between public and private consumption. That study reports a robust relationship such that both the debt–GDP ratio and the size of fiscal adjustment have an impact on the effect of public expenditure on private consumption. For ‘large’ fiscal adjustments, as defined by Alesina and Perotti (1995), the effect becomes negative for debt–GDP ratios above 33 per cent; for smaller adjustments the equivalent trigger level is 68 per cent. Stimulating a large adjustment even for Italy, however, still yields a consumption increase smaller than the public expenditure cut, so that aggregate demand falls, in contrast to the EFC position. So what else might have been going on to account for the negative relationship between government consumption and aggregate income found in the Danish case? The fact that the consumption and investment booms coincided with fiscal contraction, while the current account deficit expanded, would at first glance appear to support the notion of a demand-driven boom.2 The Commission of the European Communities (1991) shows that this phase coincided, however, with a leftward movement of the Phillips curve, suggesting a positive supply-side shock rather than the positive demand-side shock posited by Giavazzi and Pagano (1990). Beneficial supplyside changes reduce unemployment, and can easily stimulate sufficient investment to bring about a current account deterioration, as happened in the Danish case. Furthermore, the latter will correct itself over time as new business investment comes on stream. Did such a positive supply shock occur during the Danish stabilization? Note first that the real exchange rate had depreciated considerably since 1979, putting the economy in a good position to benefit when world demand picked up after 1982 (see Nielsen and Søndergaard, 1991; Andersen, 1994). These competitiveness gains would arguably have been inflated away were it not for the fiscal contraction of 1982. The contraction, indeed, alongside the abolition of wage indexation mechanisms and the temporary wage freeze, appears to have resulted in further competitiveness gains between 1982 and 1986 (see Andersen and Risager, 1990). The output consequences of cost-competitiveness gains of this type would show up, as Bergman and Hutchison recognize, as non-government-related ‘other factors’ raising permanent income, and these of course they find to be the main driving force behind consumption in the 1984–6 period. Cost competitiveness also improved in Ireland as fiscal contraction proceeded, suggesting that this may indeed be the missing link between fiscal contraction and output
Comment
259
and employment expansions. this is the line of research that Devereux and I are currently pursuing. In Barry and Devereux (1997) we assume that a monopoly union sets two-period wages taking future taxes into account. The effects of fiscal contraction we find depend on the associated tax implications. ‘Small’ counter-cyclical policies have no tax implications and so employment effects go in the normal (‘Keynesian’) direction. Overly large or pro-cyclical policies (what we might call ‘abnormal policies’) do have tax implications, however, and these reverse the direction of the employment effects. When we embed this view in the type of macro models used in our 1995 paper (Barry and Devereux, 1995) we find that ‘abnormal’ fiscal contractions raise consumption, investment, employment and the current account deficit. The demand-oriented EFC hypothesis then errs in its failure to recognize that improvements in cost competitiveness are necessary if the economy is to expand in the wake of a fiscal contraction. In this I find myself re-emphasising the point made by Bean in his comments on Vastrup (1989), that ‘the key feature in an understanding of the Danish experience … is the response of the labour market’. Some of the ‘important lessons to be learned here’, that Bean alluded to, we can glean from the present stimulating chapter.
Notes 1.
2.
Denmark is more interesting than Ireland in this regard since, as I showed in my 1991 paper, the Irish recovery was much less robust than the picture painted by Giavazzi and Pagano (1990) when presented against the backdrop of world market conditions. During the Danish fiscal contraction, however, output, employment and investment growth were all stronger there than elsewhere in Europe (see OECD, 1986; and Andersen and Risager, 1988). A number of commentators however point out that the simultaneous capital market liberalization may have exerted an independent stimulus to consumption and investment.
References Alesina, A. and R. Perotti (1995) ‘Fiscal expansions and fiscal adjustments’, Economic Policy, pp. 205–48. Andersen, T.M. (1994) ‘Disinflationary Stabilization policy: Denmark in the 1980s’, in J. Åkerholm and A. Giovannini (eds), Exchange Rate Policies in the Nordic Countries (Cambridge: Cambridge University Press for the Centre for Economic Policy Research). Andersen, T.M. and O. Risager (1988) ‘Stabilization policies, credibility, and interest rate determination in a small open economy’, European Economic Review, 32, pp. 669–79. Andersen, T.M. and O. Risager (1990) ‘Wage formation in Denmark’, in L. Calmfors (ed.), Wage Formation and Macroeconomic Policies in the Nordic Countries (Oxford: Oxford University Press). Barry, F.G. (1991) ‘The Irish recovery 1987–90: an economic miracle’, The Irish Banking Review (Winter), pp. 23–40. Barry, F.G. and M. Devereux (1997) ‘Fiscal contraction and the possibility of supply-side expansion’, Centre for Economic Research Working Paper, University College Dublin. Barry, F.G. and M. Devereux (1995) ‘The “expansionary fiscal contraction” hypothesis: a Neo-Keynesian analysis’, Oxford Economic Papers, 47, pp. 249–64.
260
Barry
Bertola, G. and A. Drazin (1993) ‘Trigger points and budget cuts: explaining the effects of fiscal austerity’, American Economic Review, 83, pp. 11–26. Blanchard, O. (1987) ‘Germany and the world economy: a US view’, Economic Policy, 1(4), pp. 195–200. Commission of the European Communities (1991) Country Study: Denmark (Brussels: European Commission). Giavazzi, F. and M. Pagano (1990) ‘Can severe fiscal expansions be expansionary? Tale of two small European countries’, in O.J. Blanchard and S. Fischer (eds), NBER Macroeconomics Annual 1990 (Cambridge, Mass.: MIT Press), pp. 75–123. Miller, M., R. Skidelsky and P. Weller (1990) ‘Fear of deficit financing: is it rational?’, in R. Dornbusch and M. Draghi (eds), Public Debt Management: Theory and History (Cambridge: Cambridge University Press). Nielsen, S.B. and J. Søndergaard (1991) ‘Macroeconomic policy and the external constraint–the Danish experience’ in G. Alogoskontis, L. Papademos and R. Portes (eds), External Constraints on Macroeconomic Policy: The European Experience (Cambridge: Cambridge University Press). OECD (1986) Country Study: Denmark (Paris: OECD). Sutherland, A. (1997) ‘Fiscal crises and aggregate demand: can high public debt reverse the effects of fiscal policy’, Journal of Public Economics, 65(20), pp. 147–62. Vastrup, C. (1989) ‘Economic policy and adjustment in Denmark’, in M. Monti (ed.), Fiscal Policy, Economic Adjustment and Financial Markets (Washington, DC: IMF). Whelan, K. (1996) ‘Expansionary fiscal contractions?’, Massachusetts Institute of Technology, unpublished m.s.
Comment Anders Møller Christensen Bergman and Hutchisons’ Chapter 8 is well balanced. The overall conclusion of the chapter is that on close inspection the Danish experience in the period from 1982 to 1986 does not provide empirical support for the EFC hypothesis – the possibility that a tightening of fiscal policy turns out to have an expansionary impact on the economy via a boom in private consumption. This Comment will offer an interpretation of the boom in which the fiscal contraction is one of three roughly equally important elements of the economic policy pursued. The other core elements of the policy were a fixed exchange rate policy and an incomes policy. Accordingly, from a purely theoretical perspective the Danish boom starting in 1983 is interesting, as any of the above-mentioned elements of economic policy normally will be expected to have a contractionary or maybe neutral impact on economic developments. The Danish boom has therefore been of considerable interest to economists in as well as outside the country, but as witnessed by Bergman and Hutchison the evidence does not support EFC hypothesis as such. In the following an attempt will be made to explain in relatively broad terms what was the background for the changes in policy in late 1982, and argue that all elements of policy were of importance for boosting confidence and thus bringing about the massive decline in nominal interest rates that triggered the boom. Relying just on the fiscal policy pursued is a too narrow point of view to capture the evidence of the period. In 1982 the Danish economy performed rather badly by almost any standard. Unemployment was around 10 per cent and rising, the general government deficit was around 9 per cent of GDP, inflation was running in excess of 10 per cent, and in spite of the depressed state of the economy the current account of the balance of payments was showing a deficit of around 4 per cent of GDP. The position was virtually unsustainable, as demonstrated by a long-term interest rate of nearly 22 per cent. This unfortunate position was the result of more than a decade with an obviously poor performance. Since 1970 growth had been slower than in most other European OECD countries while inflation had been higher in spite of a fixed exchange rate policy within first the snake and from 1979 onwards within the European Monetary System (EMS). After the hike in oil prices in 1973, unemployment began to surface and government finances started to show a deficit. The crucial point is that even if the same trends were present in most other European countries the Danish performance was extraordinary poor by international standards. In fact, a discussion of whether or not the Danish economy was approaching the abyss went on, initiated in 1979 by the former Minister of Finance, Knud Heinesen. It might deserve to be spelled out that Mr Heinesen at that point in time still was a leading member of Parliament for the Social Democrats who were in government as well. The sense of unsustainability showed up in an increasing interest rate differential against Germany. In early 1982 Belgium called for a realignment of exchange rates. Denmark joined, calling for a depreciation of the Danish Kroner by 7 per cent. In the end the Danish government only got a depreciation of 3 per cent but it had become known that they were aiming at higher numbers. This led to a further increase in interest rates and therefore a further discussion of the burdens associated with servicing government
261
262
Christensen
debt with real interest rates around 12 per cent. At one point in time during the discussions the chairman of Parliament’s budgetary committee suggested that it might be necessary to change the terms of outstanding government debt. In the period from 1970 to 1982 most parties in Parliament had government responsibilities at some period. Accordingly it was not obvious that a change in government would mean a genuine shift in economic policy. In September 1982 a change in government took place. The Social Democrats stepped down, and a new minority government was formed by the conservatives and the liberals without an election, pace to a recent description of the experiences by Alesina and Perotti (1997), who erroneously state that it was after a convincing electoral victory. Before the new government had been formed rumours started to circulate that a strategic devaluation of the Kroner was part of the incoming government’s policy. This led the Prime Minister and the Minister of Finance to announce that ‘a devaluation is not part nor will be part of the economic policy that will be put forward by the new government’ (see Schlüter and Christophersen, 1993). Besides maintaining a stable exchange rate the new government proposed legislation against wage indexation along with several fiscal measures – in the end increasing taxes and stabilising government consumption. Thus the basic fact is that the new government was formed on a package composed of three types of economic policy – fiscal policy, incomes policy and a fixed exchange rate policy. When assessing the results, all three legs should be borne in mind, not just fiscal policy. In fact, it was the exchange rate policy that came to be tested immediately and thus the most likely part of the package to which to attribute to the initial rapid decline in interest rates. In early October 1982 the Swedish government announced a depreciation of the Swedish Krona by no less than 16 per cent. As Sweden is Denmark’s second biggest trading partner the Swedish policy was critical for the Danish government. To the surprise of, among others, the Swedish authorities, the Danish government maintained its objectives and gained credibility on the exchange rate issue immediately. The importance of this episode for the developments in the following years can hardly be overestimated even if it is rarely reported in catalogues of government actions as no action was taken from a formal point of view. The exchange rate policy was confirmed in March 1983 when the French franc was devalued in the EMS. The incomes policy was of great importance in making the exchange rate policy more credible. Wage indexation was abolished to break the inflation dynamics and inflation declined at a rather rapid pace. After an initial decline, real wages started to increase steadily and approximately in line with productivity in a low-inflation environment in the second half of the 1980s. From 1982 to 1986 government finances improved from a deficit of around 9 per cent of GDP to a surplus of around 4 per cent. This improvement shows the combined effects of the significant fiscal tightening and the favourable cyclical developments. The cyclical developments can basically be explained as the consequence of the decline in interest rates, a development that probably would not have happened unless the whole economic package including the tightening of fiscal policy had been implemented. However, many measures of fiscal policy tend to exaggerate the fiscal rigour, as a significant part was a tax on the interest income of pension funds and thus a tax that did not directly influence the disposable income of households. To summarize: the economic performance in Denmark in the period from 1983 to 1986 was impressive but has to be seen against a very depressed and unbalanced economy at the outset. The fiscal tightening, along with the fixed exchange rate policy
Comment
263
and the incomes policy, triggered a rapid decline in interest rates that in turn triggered the economic expansion. All elements of the policy package are considered of importance, not just fiscal policy as suggested by the EFC hypothesis.
References Alesina, A. and R. Perotti (1997) ‘Fiscal adjustments in OECD countries: composition and macroeconomic effects’, IMF Staff Papers 44, pp. 210–48. Schlüter, P. and H. Christophersen (1993) ‘Statement’ (9 September 1982), quoted and translated from Erik Hoffmeyer ‘Pengepolitiske problemstillinger 1965–1990’, Dansk Pengehistorie, Bind 5, Danmarks Nationalbank, p. 94.
9 The Macroeconomics of the Welfare state1 Torben M. Andersen and Torben D. Schmidt
9.1
INTRODUCTION
The structure and size of the public sector is hotly debated, and it is often questioned whether the public sector has become too large or whether its structure has become inappropriate. The overall performance of the economy is also often related to the size and structure of the public sector. Accordingly the role and future of the welfare state is high on the political agenda. The aim of this chapter is to use recent theoretical results as a guide for an attempt to make an empirical assessment of some central macroeconomic implications of the public sector. Is it at all possible to say something on the importance of the size and structure of the public sector for overall macroeconomic performance? If so, this may yield insights on the cost and benefits of public sector activities which are needed for a political prioritization of resource allocation. By the ‘welfare state’ we understand (see Sandmo, 1995), that economic and social policies are strongly geared towards equality and individual protection against social hazards. According to this broad definition, macroeconomic policy is also part of the welfare state policy to the extent that it focuses on stabilizing economic performance and lowering unemployment. The macroeconomics of the welfare state thus raises two central questions – namely, what are the overall macroeconomic consequences of welfare state policies, and to what extent can macro policy contribute to achieving the goals of the welfare state? A central characteristic of the welfare state is a large public sector which is responsible for the provision of services such as health care, education and infrastructure as well as for transfers to old, sick, disabled and unemployed people.2 The financing of these activities requires taxation which mainly have distortionary effects.3 From this arises the traditional viewpoint that the larger the public sector, the larger the tax burden and therefore the distortionary costs. If this was the only effect it would be simple to conclude on the importance of a large public sector. Matters are complicated by the fact that many public sector activities are directed at improving economic performance by exploiting economies of scale, correcting market imperfections and 264
The Macroeconomics of the Welfare State
265
compensating for externalities. Moreover, the provision of both services and transfers can be interpreted as social insurance which can be justified both on equity grounds but also in many cases on efficiency grounds. There is obviously no easy route by which to arrive at conclusions on the macroeconomic importance of the welfare state. Obviously it is difficult if not impossible to capture all these effects in a single model which can be subjected to empirical testing. Therefore, we follow the alternative approach of using cross-country comparisons as a way to see if there is any obvious and strong connection between the size and structure of the public sector and macroeconomic performance. This approach holds some potential as there are substantial differences in the size and structure of the public sector even among Western European countries as revealed by the summary measures provided in Table 9.1. We use these differences as the base from which to infer something on the macroeconomic consequences of the welfare state. This approach also has some disadvantages. First it blends the size and structure of the public sector with other institutional arrangements which may make it difficult to infer the underlying causes for cross-country differences. Second the method is very coarse and therefore it may neglect consequences at the micro level which may add up to have macroconsequences. Consequently, Table 9.1
Government spending and revenues: international comparison, 1993 Of which
Total disbursement
USA Japan Germany France UK Canada Austria Belgium Denmark Finland Greece Norway Switzerland Australia Source: OECD (1995).
35 36 49 54 42 48 51 55 61 59 47 48 47 38
Of which
Public consumption relative to GDP
Transfers relative to GDP
Receipts relative to GDP
Taxes relative to GDP
16 10 20 19 21 19 19 15 25 22 14 21 14 17
14 13 18 23 14 15 15 22 22 24 16 16 18 13
31 32 46 49 38 42 47 51 59 53 38 50 47 35
31 28 43 45 34 36 44 46 52 46 34 42 39 31
Debt relative to GDP
64 81 62 60 60 100 69 134 80 63 112 40 23 44
266
Andersen and Schmidt
the reported empirical findings should be seen as a first attempt to identify major consequences of public sector activities on macroeconomic performance which can be used as a guide for future work aiming at a more precise identification of the mechanisms at stake. It is important to stress from the outset that this chapter does not deal with the ‘microeconomic structure’ of the public sector, and thus leaves out questions on how to organize public institutions and the extent to which privatization can provide efficiency gains without impairing with other objectives of the welfare state. Macroeconomic policy can, as noted, be seen as an essential part of the welfare state to the extent that it focuses on stabilizing aggregate activity and thereby provides social insurance. It is therefore natural to start by considering the role of the public sector for macroeconomic stability. This is done in section 9.2. Next we consider in section 9.3 whether any effects on the overall activity level and its growth rate can be observed. Finally, we address in section 9.4 the question of whether economies with large public sectors tend to be more afflicted by a systematic deficit bias and therefore debt accumulation. Section 9.5 summarizes conclusions which can be drawn from the empirical findings reported in the chapter.
9.2
MACROECONOMIC VOLATILITY
A central aspect of many public sector activities is that they depend on the business-cycle situation. This means that when the economy is in an upturn, certain types of government expenditures such as unemployment benefits fall while the revenue generated from taxation increases (also relative to income due to non-linear elements in taxation) and vice versa in a recession. From a microeconomic perspective this implies that there is an element of social insurance in the public budget in the sense of diversification across states of nature. In the macroeconomic literature this is discussed under the heading of automatic stabilizers as one of the main consequences of this is an automatic adjustment of expenditures and revenues to the business cycle situation as well as a stabilization of the overall income level and thus possibly aggregate demand. Table 9.2 provides evidence on the empirical importance of the automatic stabilizers by considering the cyclical sensitivity of budget revenues and expenditures for EU countries. The tax revenue is seen to move procyclically while expenditures move counter-cyclically in all EU countries. One also finds that budgetary receipts are numerically relatively more cyclically dependent than expenditures. A consequence of this cyclical dependency in receipts and expenditures is that the public budget is highly cyclically dependent and moves pro-cyclically.
The Macroeconomics of the Welfare State Table 9.2
267
Cyclical sensitivity of budget revenues and expenditures in the EC, 1995a · · · · · · · · · · · · · · · · · · · 1% change in GDP · · · · · · · · · · · · · · · · · Impact on the revenue to GDP-ratio
B DK D EL E F IRL I L NL A P FIN S UK EUR
0.5 0.5 0.4 0.3 0.5 0.5 0.4 0.3 0.4 0.5 0.4 0.4 0.5 0.6 0.5 0.4
Impact on the expenditure GDP-ratio –0.1 –0.3 –0.1 –0.1 –0.2 –0.1 –0.2 –0.1 –0.2 –0.2 –0.1 –0.1 –0.2 –0.2 –0.2 –0.1
Impact on the budget to GDP-ratio 0.6 0.7 0.5 0.4 0.6 0.5 0.5 0.5 0.6 0.8 0.5 0.5 0.6 0.9 0.6 0.5
a
Figures are rounded up. Source: Commission of the European Communities (1997).
It is to be expected that the cyclical sensitivity of revenues and expenditures is dependent on the size of the public sector. Figure 9.1 brings this out very clearly,4 the larger the public sector the more sensitive are both revenue and expenditures. Expanding the public sector is difficult without also increasing its cyclical sensitivity. As it is a natural consequence that the cyclical sensitivity is increasing in the size of the public sector, it becomes of interest to analyze the extent to which this affects the overall macroeconomic performance. Is it the case that the public sector actually stabilizes economic performance such that more stability in the sense of less business cycle volatility is observed in countries with a large public sector? If one looks at the relationship between the size of government proxied by tax revenue relative to GDP and macroeconomic volatility proxied by the standard deviation of detrended (log) GDP, one finds a weak negative relationship (see Figure 9.2).5 Although it is not possible to make strong conclusions, the figure does indicate that the larger the size of the public sector,
268
Andersen and Schmidt
Figure 9.1 Budget sensitivity and public sector size measured by revenue to GDP ratio, EU countries, 1995
Note: Budget sensitivity defined as in Table 9.2; public sector size measured by revenue to GDP ratio for 1995 OECD, Economic Outlook, 1999).
Figure 9.2 Public sector size and standard deviation of cyclical measure for total activity, 1971–90
Note: Standard deviation of cyclical measure of total activity, own calculation available on request; see appendix; Public sector size measured by the average revenue to GDP ratio (–) (see Table 9.3).
The Macroeconomics of the Welfare State
269
the less is macroeconomic volatility (see also Gali, 1994). This suggests that the public sector succeeds in stabilizing macroeconomic performance. To look more carefully at this issue, let us make the following simple decomposition of total GDP (y) y = yp + yo
(9.1)
where yp is GDP in the private sector and yo is GDP in the public sector. It follows that var( y) = var( y p ) + var( yo ) + 2 cov( y p , yo )
(9.2)
We can now distinguish between three possible channels through which the public sector can reduce macroeconomic volatility:6 (1)
(2)
(3)
Composition: The activities of the public sector are to a lesser extent driven by market forces and therefore the shocks inducing business cycles in the private sector. Consequently, activity in the public sector is less volatile than in the private sector (var(yo) < var(yp)), and a relative large public sector implies low macroeconomic volatility. Countercyclicality: The public sector activities may be designed so as to be a buffer to variations in private sector activities – that is, when private sector activity goes down public sector activities (employment) are expanded and vice versa (cov(yo,yp) < 0).The more public sector activities are adjusted in a counter-cyclical way, the more they contribute to macroeconomic stability.7 Stabilization: The public sector activities may also be designed so as to stabilize activity in the private sector via stabilization of aggregate demand primarily through taxes and transfers 8 (var(yp) is lowered due to the way in which taxes and transfers vary over the business cycle).
The following aims at an empirical assessment of the importance of these factors for the observed negative relationship between macroeconomic volatility and the size of the public sector. To address these questions, we need a measure for private and public sector activity. We use data for 14 OECD countries and the sample period is 1971–90. To evaluate the cyclical properties of private and public sector activities, we need a method to decompose the trend from the cyclical component. We do this by using the Hodrick–Prescott filter (HP filter) as well as the more simple first difference of the logarithmic value of the variable in question (LD filter) (see, for example, Bergström and Vredin, 1994).9 Table 9.3 reports some summary statistics on the cyclical properties of taxes and transfers. It is interesting to note that the standard deviation of tax rates is rather high and although there is a tendency for tax rates to move
Andersen and Schmidt
270 Table 9.3 Country USA Japan Germany France UK Canada Austria Belgium Denmark Finland Greece Norway Switzerland Australia
Cyclical properties of tax rates and social security transfers
–
(∆)
(∆, ∆y)
e–
(∆ess)
(∆ess, ∆y)
28.3 25.5 40.9 39.8 35.8 32.6 41.1 43.8 45.8 39.2 29.0 43.8 30.2 28.7
1.39 2.04 1.04 1.17 2.71 1.99 1.38 1.38 1.91 3.80 3.43 1.49 1.79 2.27
0.44 –0.14 0.12 –0.42 –0.68 –0.30 0.02 –0.78 0.27 –0.53 –0.26 0.39 –0.60 –0.19
9.61 9.33 14.2 17.3 10.0 10.8 14.8 20.8 15.3 10.6 11.1 13.9 12.5 8.41
4.33 3.74 3.61 1.80 3.11 4.00 2.07 2.29 3.67 7.07 4.47 2.70 3.53 4.71
–0.95 –0.75 –0.79 –0.75 –0.71 –0.78 –0.62 –0.71 –0.89 –0.81 –0.42 –0.31 –0.60 –0.64
Source: Own calculation (see Appendix, available upon request), ∆ defines HP-filtered data.
pro-cyclically the effect is not as strong as might have been expected given the revenue sensitivity reported in Table 9.2. An important difference between the two measures is that the latter includes only the automatic effects built into the taxation system while the measure used here also includes discretionary tax changes. Social security transfer expenses also display substantial volatility and move counter-cyclically. Turning to the implications of the public sector for macroeconomic stability, it can be seen from Table 9.4 that the composition effect is rather strong. The variance of public sector activity is less than 10 per cent of the variance of private sector activity for all countries included and in some case much below this number. This applies for both filtering methods. 10 Lowering the relative size of the private sector is therefore bound to lower the variability of total activity. Considering the cyclical properties of public sector activities relative to the private sector we find for most countries a negative correlation – that is, public sector activity moves in an opposite direction to private sector activity over the business cycle. However, there are some exceptions to this (for example, the United States and Belgium), reflecting both differences in institutions and in the use of fiscal policy. It can, however, be concluded that it is possible to design the public sector such that it moves counter-cyclically, and therefore act as a buffer to variations in private sector activity.
The Macroeconomics of the Welfare State Table 9.4
Variance decomposition ∆y
Variance of public sector activity, percentage of the variance of private sector activity USA Japan Germany France UK Canada Austria Belgium Denmark Finland Greece Norway Switzerland Australia
271
0.4 0.6 0.9 1.5 3.8 1.4 1.1 3.9 3.3 1.5 5.3 5.6 1.0 3.3
Correlation coefficient between private and public sector activities
0.48 –0.35 0.06 –0.28 –0.50 –0.56 –0.40 0.21 –0.72 –0.49 –0.03 0.10 0.05 –0.60
Source: Own calculation (see Appendix, available upon request).
Whereas it is easy to address the composition and countercylicality issues directly by considering the cyclical properties of private and public sector activity, it is more difficult to deal with the stabilization effect. Is the private sector business cycle affected by the public sector? This is a controversial question which has been much debated in the theoretical literature. In particular, the variations induced via the tax system have been subject to much research. This is natural as we have seen that this accounts for most of the observed cyclical sensitivity of public budgets independently of the relative size of the public sector. According to standard macro models an income tax would tend to stabilize disposable income and thereby consumption and aggregate demand. The basic mechanism is simply that the multipliers become smaller and therefore the volatility of aggregate activity to different types of shocks is reduced (see Christiano, 1984, for an overview of the earlier debate and for references11). The more recent literature has stressed that the effects of cyclical variation in tax revenues depend on the temporal structure of distortions and the implicit insurance provided (see, for example, Andersen and Dogonowski, 1997). The variance of disposable income is lower, the higher the tax rate – the Domar–Musgrave (1944) effect – and this may have an independent effect
272
Andersen and Schmidt
on behaviour via changes in risk. This may be reinforced by progressive elements in taxation. These arguments support the hypotheses that automatic variations in tax revenues, not only improve welfare, but also tend to stabilize aggregate activity.12 What is the empirical evidence? Model simulations indicate that automatic stabilizers are actually stabilizing and can be of substantial quantitative importance. This finding is supported both in models with a Keynesian Structure (see Det Økonomiske Råd, 1996, and Finansministeriet, 1997), and models with a real business-cycle structure (for example, King, Plosser and Rebelo, 1988). Obviously, this conclusion is based on the assumed theoretical structure which is biased in favour of stabilizing effects. We have tried to shed some light on this question by relating the variability of private sector activity to the level and cyclical properties of taxes and transfers (see Table 9.3). The potential stabilizing effects induced by taxation and social security transfer payments arise primarily from the fact that variation in income (y) are not reflected one-to-one in disposable income (y(1 – )). Of course variability in taxes and transfers may be an independent source of variability in disposable income, but this may be counteracted to the extent that taxes move pro-cyclically and transfers counter-cyclically. The result of various regressions in which the volatility of private sector activity is related to properties of taxes and social security transfers is shown in Table 9.5. There is indication that taxes and social security transfers tend to stabilize private sector activity as the net effect of the average tax and transfer rate is negative on private sector volatility. Surprisingly, volatility in taxes does not contribute to volatility in the private sector, while volatility in transfers does. There is also indication that pro-cyclical taxes tend to stabilize the private sector, while counter-cyclical transfers do not contribute to stability in the private sector. Of course, care should be taken in interpreting these results as they do not reflect estimation of a structural model and multicollinearity problems are present. Overall, we conclude that the evidence supports the hypothesis that the public sector can stabilize the overall activity level in the economy. There is also some evidence in support of activity in the private sector being stabilized.
9.3
THE PUBLIC SECTOR AND GROWTH
Having considered the consequences of the public sector for business-cycle fluctuations, it is natural to turn to the implications for growth. Assessing the importance of the public sector for growth is difficult, as the direction of causality may run both ways.
Table 9.5
Regressions of volatility in private sector activity on various measures of taxes and social security transfers
– (1) (2) (3) (4) (5) (6) (7) (8)
( )
(, ∆yp)
–0.007 (0.533) –
–
–
–
–
0.022 (0.162) –0.007 (0.567) –
–
–
0.024 (0.127) 0.017 (0.280) 0.016 (0.099)
0.001 (0.926) – –0.0006 (0.685) – –0.0012 (0.199)
–0.344 (0.540) – – 0.473 (0.279) –0.522 (0.083)
e–ss – –0.041 (0.057) –0.076 (0.027) – –0.055 (0.005) –0.097 (0.14) –0.058 (0.100) –0.081 (0.004)
(ess)
(ess, ∆yp)
R2
F 0.413 (0.533) 4.424 (0.057) 3.566 (0.064) 0.292 (0.830) 8.393 (0.004) 2.534 (0.114) 2.465 (0.120) 8.227 (0.007)
–
–
0.033
–
–
0.269
–
–
0.393
–
–
0.081
0.003 (0.006) 0.002 (0.333) – 0.005 (0.015)
–2.089 (0.006) –
0.716
–1.018 (0.203) –2.154 (0.004)
0.523
0.530
0.876
Note: All regressions contain an intercept, although not reported. They enter all regressions significantly (10%). The only exception is regression (8) where it is almost significant.
273
274
Andersen and Schmidt
A positive relation between growth and the size of the public sector may reflect the fact that when societies become more affluent, they want to spend more on health care, education, cultural activities, etc. – that is, the demand for public goods has a high income elasticity. This also applies for the insurance function of the public sector which has expanded alongside the increase in ‘traditional’ public sector activities. There is a tendency that countries which have a high public spending ratio also have a high ratio of transfer to GDP. The increase in the relative size of the public sector is reinforced by the fact that public sector activities tend to be labour-intensive. Such activities inevitably become relatively more expensive over time because of their generally constant labour productivity (Baumol’s Law). A negative relation between growth and the size of the public sector may arise because public consumption has to be financed by distortionary taxes with consequences for activity in the private sector. Therefore, the cost of an additional unit of public consumption would in general be larger than 1 as on top of the direct resource use there is a cost in terms of lower activity in the private sector caused by distortionary taxes. From this channel one may expect a negative relationship between the size of the public sector and growth, as a larger share of public consumption squeezes the private sector from which growth is going to originate.13 The theoretical literature confirms that it is not possible to give a clear-cut conclusion on the relation between growth and the size of the public sector. On the one hand, public spending can be viewed as collective spending which has to be financed by distortionary taxes lowering, among other things, incentives to save and work and therefore possibly economic growth. This may be reinforced to the extent the political process leads to excessive public spending and there are inefficiencies in public organizations. On the other hand, the rationale for many public activities arises from a need to supply public goods, exploit economies of scale and correct market imperfections and externalities. Expenditures on such items can be viewed as investments which improve the performance of the economy and therefore they have positive spillover effects on private sector activity and thereby possibly growth. From a theoretical perspective, it is thus not possible to make general statements on the implications of the size of the public sector for economic growth, it depends on the specific expenditures and the mode of financing them. There has been considerable interest in trying to establish an empirical relationship between economic growth and the size of the public sector measured by spending or taxes. Such a relationship would provide information on whether the positive or negative effects tend to dominate in practice. Figure 9.3 gives a simple cross-plot of average growth rates and size of the public sector for 14 countries, and it displays a weak negative relation between the size of the public sector and growth.14 Although the effect may
The Macroeconomics of the Welfare State Figure 9.3
275
Growth and public sector size, 1971–90
Note: Average growth rate over the sample period, (see OECD Economic Outlook, 1997); public size measured by average taxes to GDP ratio ( –) see Table 9.3.
not seem strong, it should be noted that the effect cumulates, even a small growth difference as 0.2 per cent per annuum would accumulate over 16 years to a 3.2 per cent lower level of GDP. However, figures of this type are very sensitive to the sample period and the definition of variables, and it is therefore very difficult to settle the question of whether the public sector impedes or spurs growth based on the simple correlation between growth and the public sector. Evaluating whether the positive or negative effects tend to dominate is moreover made difficult by problems of measurement. Growth in the public sector tends to imply that activities previously taken care of privately – like, for example, care taking of children and elderly people – become organized in the public sector. Hence, even if these activities substitute each other one for one, the registered production would increase, and therefore the growth rate would be higher in the transition phase. This development should be seen alongside the increase in the labour force participation rate for women which in itself had a positive growth effect. A blurred picture is also to be expected given that there has been made no attempt to control for other variables and to distinguish between demand and supply effects. This ambiguity is also found in other empirical analysis based on cross-country comparisons as some find a clear positive relationship
276
Andersen and Schmidt
between growth and the size of the public sector (see, for example, Ram, 1986), while others find a negative relationship (see, for example, Engen and Skinner, 1992; Barro, 1991). Obviously, there are substantial problems with data and the test strategy in these investigations.15 A more disaggregated approach may be more informative by distinguishing between different spending and taxation items in the public budget. Hansson and Henrekson (1994) consider the effect of the size of public expenditure on total factor productivity in 14 OECD countries. They find a significant negative effect on productivity growth from total public expenditures. Disaggregating, they find that public consumption exerts a significant negative effect, the effect of public investments is also found to be negative but insignificant while expenditures on education are found to have a significant positive effect. Considering the disaggregated approach in more detail, it is useful to distinguish between the effects of the mode of financing (taxes) and different categories of expenditures. In a recent and careful survey on the effects of taxes for economic growth based on both micro and macro studies, it is concluded by Engen and Skinner (1996) that tax policy does affect economic growth. Both a higher tax burden and more progression do have a negative effect on growth. Though the effects are significant, they are not enormous and therefore the effects will not materialize as abrupt changes to tax reforms. However, the effects cumulate over time. Turning to expenditures, there are two important types of public consumption for which there are positive spillover effects for the private sector – namely, education and infrastructure. A well developed infrastructure benefits the private sector by reducing, for example, transportation costs. Although not a pure public good – only a given number of cars can pass a bridge in a given period of time – it has a public goods character, and organizing the supply and maintenance of infrastructure in the public sector has a number of advantages. Obviously, infrastructure also has a consumption value by making it easier/less costly to get to vacation resorts, etc. Aschauer (1989) considers the role of public investments and argues that investments in, for example, infrastructure increase productivity of private investments and thereby work to boost growth. Hence, the direct crowding out effect of such investments arising from the resources absorbed should be weighted relative to the indirect benefits of higher productivity of private investment. Considering US data for 1953 to 1986, Aschauer (1989) finds strong support for positive spillovers from infrastructure investments, and a simulation exercise shows that the costs appear fairly quickly while the benefits cumulate over time and dominate in the long run. Aschauer (1989) goes as far as to suggest that the recent fall in productivity can be due to lower public investments. Country studies for Denmark (Economic Council, 1996) and Sweden (Berndt and Hansson, 1992) confirm that public infrastructure
The Macroeconomics of the Welfare State
277
investments have a positive spillover effect to the private sector. Englander and Gurney (1994) include in a cross-country study the role of public investment in infrastructure and physical capital, but they do not find support for such investments to exert a significant effect on growth. It is thus not possible to make general statements on whether there has been an overinvestment in public infrastructure capital or whether there are still gains to be reaped by increasing investment in infrastructure. Education may also have positive spillover effects since expansion of knowledge through experience and research may be easier, the larger the knowledge base in society. More indirectly, public provision of education may also help overcome capital market imperfections and other barriers to education and thereby ensure that the intellectual potential is better utilized. Empirical evidence confirms that education is important for economic growth. This is reflected both in casual empiricism for the rapidly growing East Asian countries which over the past three decades have improved both the quantity and the quality of schooling and other forms of training (see World Bank, 1993), but also econometric analyses of comparative growth performance confirm that education has a siginificant effect on productivity and thus growth (see, for example, Barro, 1991; Mankiw, Romer and Weil, 1992; Levine and Renelt, 1992).16 Clearly such results should be interpreted with care as it is difficult to compare educational data across countries and the studies have to rely on measures of input like enrolment data due to lack of reliable figures on the output produced by the education sector. Also reverse causality could be at stake as richer countries tend to spend more on education. However, it seems safe to conclude that it is well supported by empirical evidence that education contributes to increasing productivity and thus living standards. The increase in average OECD enrolment rates in secondary schooling from 70 per cent in 1960 to 95 per cent in 1985 is thus associated with about 0.6 percentage point per year faster productivity growth, Englander and Gurney (1994). As already noted, there has been substantial growth in income transfers since the early 1970s. The question is whether this is only a matter of redistribution paid by distortionary taxes or whether transfer payments in some ways directly affect economic efficiency. The growth in the transfer payments made by the public sector reflects that the insurance mechanism of the welfare state was brought into operation after the first oil price shock. Rules are designed to ensure that unlucky circumstances or adverse developments in, for example, the labour market should not lead to social misery. Obviously, this also applies to some public goods – as, for example, hospitals where treatment is available to all free of charge. In general, all redistributional schemes and taxes serve a role as providing social insurance (see, for example, Sandmo, 1991; Sinn, 1995; Varian, 1980).
278
Andersen and Schmidt
There are surely efficiency gains from this as private insurance markets would not be able to deal efficiently with all needs for insurance. 17 Availability of insurance may induce people to engage in risky activities which they otherwise would avoid due to risk aversion. This may have positive spillover effects on the allocation of resources and eventually growth. However, there is also a cost as insurance may affect the incentive structure (moral hazard problems) so as to imply that people supply too little effort or become inflexible – if you have a guaranteed income why struggle hard? There is a risk that the insurance part of the public sector is expanded to a level where the disincentive effects come to dominate. There is thus a practical difficulty in designing social insurance so as to reap the potential benefits of insurance without creating important disincentive effects. It is difficult to test these effects, but some indirect evidence is found from the observed labour market inflexibility and persistent unemployment. Egalitarian objectives are a central part of the welfare state. Recent theoretical work (see, for example, Agell and Lommerud, 1993; Cahuc and Michel, 1996; Ravn and Sørensen, 1997), has pointed out that an egalitarian wage structure may overcome market failures and be conducive to growth if it accelerates the speed by which resources are moved from declining to expanding sectors or if it induces people to take more effort in education to attain the productivity level needed to qualify for a job. It is important to stress that these growth-enhancing effects all presume that becoming unemployed is a threat which provides incentives to take action. Based on political-economy models, it has also been suggested that income inequality may be harmful for growth as the political pressure for redistribution with an initial unequal distribution of income may block for growth-promoting policies (see Bénabou, 1996, for a survey of the theoretical and empirical literature). Indeed cross-country evidence seems to support the idea that equality is conducive for growth. There is no clear-cut conclusion on the relationship between growth and the size of public sector, and it is accordingly difficult to make strong conclusions. Although there is support that taxes (average and marginal) depress growth, the overall effect depends critically on the type of expenditure the taxes are going to. Certain types of expenditures like infrastructure, education and even transfer payments can potentially have positive spillovers to the private sector.
9.4 DEFICITS AND DEBT – SOCIAL INSURANCE OR POLITICAL BIAS? Over the last couple of decades there has been a tendency for governments to run systematic budget deficits and therefore public debt has been
The Macroeconomics of the Welfare State
279
accumulating (see Table 9.1). Over time for a given country there is a positive correlation between government size and the level of public debt. The cross-country correlation between government size measured by the expenditure–GDP ratio and the debt–GDP ratio is also positive and equal to 0.2 in 1990. Different explanations have been proposed why public debt has been accumulating. One simple explanation is political myopia inducing policy makers to transfer benefits to the present and defer the cost to the future. This process may have been reinforced by economists paying little attention to the cost of accumulating public debt level in the initial phase of debt accumulation.18 However, myopia seems an unlikely candidate to explain debt accumulation running over several decades. Recent political-economy literature has suggested different channels through which a systematic deficit bias may arise even though the consequences are fully perceived (see Alesina and Perotti, 1995, for an introduction and references). These explanations point to different forms of fragmentations as the cause of debt accumulation. Political fragmentation may arise from different views on the type of public spending and its geographical and temporal location. The incumbent government may use debt accumulation strategically so as to tie the hands of future governments. There may also be procedural fragmentation within the fiscal decision making process which can affect the overall fiscal decisions. One dimension of this is whether a ‘bottom-up’ or ‘top-down’ approach is used in arriving at the overall budget and its composition. Finally, social fragmentation may play a role to the extent that distributional conflicts exert a dominant influence on the political decision process. As a consequence, fiscal consolidations may be delayed due to a battle between different groups over who is going to bear the burden. The empirical evidence on these explanations is somewhat mixed and suffers from the problem of how to measure political variables. Moreover it is not quite clear why fragmentation leading to a growing public sector also should imply a deficit bias (see Perotti, 1997). One important distinction for the current debate is whether a systematic deficit bias and thus debt accumulation can be related to the overall size of the public sector or specific functions performed by the public sector. As noted, casual empiricism seems to support the former hypothesis, but this may conceal important structural issues. One important aspect in this respect is that one of the primary aims of the welfare state is to provide social insurance. As noted earlier many items on the expenditure and the revenue side provide some form of insurance. Budget deficits may serve an insurance function as they reflect the fact that the public sector uses international capital markets in an endeavour to diversify aggregate shocks – that is, more expenditures and in particular less revenue in bad states of nature and the opposite in good states of nature. To
Andersen and Schmidt
280
the extent that there are capital market imperfections, there is a welfare case for such an implicit insurance via the public budget (see Andersen and Dogonowski, 1997), and this may be seen as one of the primary objectives of the welfare state. A business cycle downturn may thus induce a budget deficit and debt accumulation, and in accordance with basic diversification principles it will not be optimal to run down the accumulated debt immediately but rather to smooth the ‘burden’ over time. This is thus a competing explanation to the predominant explanation that budget deficits arise due to growth of the public sector. Already casual evidence suggests that the insurance aspect may be important as the tendency for the public debt level to increase did not arise in the period where the traditional public sector activities were expanding, but not until the crises in the 1970s was there a call for an active use of this insurance mechanism. More direct evidence on this question can be obtained by testing whether government deficits tend to be driven by public spending or total transfers by regressing the debt–GDP ratio on the transfer–GDP ratio and the public consumption–GDP ratio. To have a sample period including significant changes in both ratios, we used the period 1960–89, implying that the selection of countries is reduced significantly relative to the number of countries Table 9.6
Public debt, transfers and public consumption
Country
Coefficient estimates for transfer, percentage of GDP
Coefficient estimates for public consumption, percentage of GDP
USA1
4,789* (0,000) 3,111* (0,027) 5,530* (0,001) 1,511* (0,000) 19,065* (0,000) 4,201 (0,197) 4,400 (0,060)
1,641* (0,036) 3,029 (0,087) 0,449 (0,734) 1,031* (0,002) –11,231* (0,000) –1,846 (0,325) –1,412 (0,203)
Germany1 Canada1 Belgium1 Finland2 Switzerland2 Australia2
R2
0,803 0,628 0,681 0,895 0,665 0,088 0,195
F
34,043* (0,000) 14,046* (0,000) 17,755* (0,000) 71,097* (0,000) 17,175** (0,000) 0.834 (0,488) 2,093 (0,126)
DW
1,353 1,439 1,862 1,878 2,020 1,537 1,281
Notes: 1 Regressions contain both intercept and trend, although neither are reported in the table; both deterministic components enter significantly. 2 Regressions contain a trend, but no intercept, the trend has not been reported in the table, although it enters the regression significantly.
The Macroeconomics of the Welfare State
281
included in the previous estimations. The results of this exercise are shown in Table 9.6. The findings clearly support the hypothesis that it is the insurance function of the welfare state more than public consumption as such which is driving debt accumulation. A deficit bias does not seem to arise from the problem of financing increasing public consumption, but more from the problem of how to finance increasing transfers (redistribution). The fact that there has been a systematic tendency to run budget deficits (permanent shocks cannot be diversified away) indicates that the process of recognizing the permanency of the problems was slow and that the political process has inertia. It is, undoubtedly, part of the story that the distortionary costs were underestimated as indicated by the fact that policy makers as well as economists drew little attention to this problem at the time. It is hard to avoid the conclusion that there is a political barrier preventing the implementation of the needed measures. This may reflect a basic weakness of the welfare state – namely, that the built-in insurance mechanisms work fairly well relative to minor shocks (usual business cycles), but when exposed to large shocks, the system may have too little adaptability and persistency arises. This may reflect both the fact that insurance can hamper flexibility and that social fragmentation can delay adjustment, precisely the distributional conflict stressed by Alesina and Drazen (1991). The real challenge for the welfare state is thus how to avoid the increase of total transfers.
9.5
CONCLUDING REMARKS
The size and structure of the public sector is a political choice. Therefore, the question of whether the public sector has become too large or whether its structure is an impediment for economic progress cannot be answered without knowing the political preference ordering. However, for the political decision process it is important to know not only the direct consequences of changes in public spending and taxation but also the implied consequences for the macroeconomic performance of the economy. In this chapter we have considered three important aspects. First we found that the public sector can in a significant way influence the business-cycle behaviour of the economy. In general it is possible to design the public sector so as to moderate business-cycle fluctuations. Secondly, the implications for economic growth are more blurred. Overall it is hard to detect any effect due to the problems of distinguishing causes and effects. There is, however, evidence that taxes lower growth, and that certain type of government activity can contribute to economic growth, most notably education and infrastructure investments. Even transfer payments may have a positive effect, although the empirical evidence on this point is very soft. Finally, the increase in transfers related to the insurance function of the welfare state seems to drive the systematic deficit bias. Although there is a strong case for
282
Andersen and Schmidt
diversification of (temporary) shocks via the budget, the permanency of the deficit bias seems to indicate that there is political inertia possibly related to social fragmentation. The adaptability of the welfare state to large (average) shocks may thus be questioned. This latter aspect relates to one important aspect which has not been addressed in the chapter – namely, whether the welfare state impairs the flexibility of the system and thereby introduces hysteresis mechanisms such that even temporary shocks can get permanent effects. This problem relates primarily to the labour market. It is outside the scope for this chapter to address this issue in detail (see Søndergaard, 1997; Hansen and Tranæs, 1997). Notes 1.
4.
Comments and suggestions from Niels Kleis Frederiksen and the two discussants Per Callesen and Douglas Hibbs are gratefully acknowledged. Rules and regulations may also be affected. With the important exception of Pigovian taxes directed at market imperfections. The OLS estimates are: budget sensitivity = 0.06 + 0.01·public sector size
5.
R2 = 0.433, = 0.102. The estimated relation is: cyclical measure = 2.02 – 0.015·public sector size
2. 3.
(0.709) (0.008)
(0.00) (0.165)
6. 7. 8. 9. 10. 11. 12. 13.
14.
R2 = 0.154, = 0.247. It is also possible that the design of the public sector influences Var(yo). We leave this out as a proper study of this channel would require detailed data on the public sector. In real business-cycle models it is often asserted that variations in public sector activities may in themselves be a source of business-cycle fluctuations (see, for example, Jonsson and Klein, 1995). There is a voluminous theoretical literature exploring this issue (see, for example, Andersen and Holden, 1997, for a recent analysis and for references). For details on the data and the filtering methods we refer to an Appendix available upon request. For LD-filtered data the ratio is somewhat higher. There are also examples where the presence of automatic stabilizers may affect the dynamic structure of the system such that it becomes unstable. See Röell and Sussman (1997) for an example where social insurance tends to increase macroeconomic volatility. Of course, there can be productivity improvements in the public sector owing to new techniques, but the accounting practice of imputing the value of public production from the input side means that registered growth must originate in the private sector. This implies that growth figures may be biased especially in countries with a large public sector. A simple regression yields: growth = 3.686 – 0.025·public sector size (0.709) (0.008)
15.
R2 = 0.065, = 0.666. A particular problem is that public expenditures and taxes are treated as independent variables paying no attention to their interrelation via the public budget.
The Macroeconomics of the Welfare State 16.
17. 18.
283
These studies do not allow for externality effects from human capital accumulation. Herbertsson and Magnusson (1995) make an attempt to do so and find these to be significant and to account for a large share of growth in total factor productivity (TFP). In general, social insurance cannot overcome all failures in private insurance markets. This is primarily the crowding-out effect running via interest rates. With liberalized capital movements this mechanism works globally and estimates suggest that each time the debt–GDP rate increases by 1 percentage point the interest rate increases by some 7 to 25 basis points (see Tanzi and Fanizza, 1995; Ford and Laxton, 1995). The increase in the debt–GDP ratio from the beginning of the 1980s to the beginning of the 1990s has caused an increase in real interest rates by between 150 to 450 basis points.
References Agell, J. and K.E. Lommerud (1993) ‘Egalitarianism and growth’, Scandinavian Journal of Economics, 35, pp. 559–79. Alesina, A. and A. Drazen (1991) ‘Why are stabilizations delayed?’, American Economic Review, 8, pp. 1170–88. Alesina, A. and R. Perotti (1995) ‘The political economy of budget deficits’, IMF Staff Papers, 22, pp. 1–31. Andersen, T.M. and R.R. Dogonowski (1997) ‘Aggregated shocks, optimal taxation and social insurance via the public budget’, Working Paper, Department of Economics, University of Aarhus. Andersen, T.M. and S. Holden (1997) ‘Demand management in an open economy with structural unemployment’, Working Paper, Department of Economics, University of Aarhus. Aschauer, D.A. (1989) ‘Does public capital crowd out private capital?’, Journal of Monetary Economics, 24, pp. 171–88. Barro, R. (1991) ‘Economic growth in a cross section of countries’, Quarterly Journal of Economics, 106, pp. 407–43. Bénabou, R. (1996) ‘Inequality and growth’, NBER Macroeconomics Annual 1996. Bergström, V. and A. Vredin (eds) (1994) Measuring and Interpreting Business Cycles (Oxford: Clarendon Press). Berndt, E.R. and B. Hansson (1992) ‘Measuring the contribution of public infrastructure capital in Sweden’, Scandinavian Journal of Economics, 94, Supplement, pp. 151–68. Cahuc, P. and P. Michel (1996) ‘Minium wage, unemployment and growth’, European Economic Review, 40, pp. 1463–82. Christiano, L.J. (1984) ‘A reexamination of the theory of automatic stabilizers’, Carnegie–Rochester Conference Series on Public Policy, 20, pp. 147–206. Commission of the European Communities (1997). 1997 Annual Report – Growth, Employment and Convergence on the Road to EMU, COM (97) 27 (Brussels) Domar, E. and R.A. Musgrave (1944) ‘Proportional income taxing and risk sharing’, Quarterly Journal of Economics, 58, pp. 388–422. Economic Council (1996) The Danish Economy (Spring). Engen, E.M. and J. Skinner (1992) ‘Fiscal policy and economic growth’, NBER Working Paper, 4223.
284
Andersen and Schmidt
Engen, E.M. and J. Skinner (1996) ‘Taxation and economic growth’, NBER Working Paper, 5826. Englander, A.S. and A. Gurney (1994) ‘OECD productivity growth: medium term trends’, OECD Economic Studies, 22 (Spring), pp. 111–28. Finansministeriet (1997) Finansredegørelse. Ford, R. and D. Laxton (1995) ‘World public debt and real interest rates’, IMF Working Paper, 95–30. Gali, J. (1994) ‘Government size and macroeconomic stability’, European Economic Review, 38, pp. 117–32. Hansen, C.T. and T. Tranæs (1997) ‘Effort commitment in active labour market policy’, Chapter 11 in this volume. Hansson, P. and M. Henrekson (1994) ‘A new framework for testing the effect of government spending on growth and productivity’, Public Choice, 81, pp. 381–401. Herbertsson, T.T. and G. Magnussen (1995) ‘Accounting for growth in the five Nordic countries 1971–1992’, Iceland Economic Papers, 36, University of Iceland. Jonsson, G. and P. Klein (1995) ‘Stochastic fiscal policy and the Swedish business cycle’, IIES Seminar Paper, 592. King, R., C. Plosser and S. Rebelo (1988) ‘Production, growth and business cycles: II – new directions’, Journal of Monetary Economics, 21, pp. 309–41. Levine, R. and D. Renelt (1992) ‘A sensitivity analysis of cross country growth regressions’, American Economic Review, 82. Mankiw, G., D. Romer and D. Weil (1992) ‘A contribution to the empirics of economic growth’, Quarterly Journal of Economics, 107, pp. 407–37. OECD (1997), Economic Outlook (Paris: OECD). Pedersen, P.J. and J. Søndergaard (1996) ‘Velfærdstab og vækst’, Nationaløkonomisk Tidsskrift – Festskrift for Anders Ølgård, pp. 317–28. Perotti, R. (1997) ‘The political economy of fiscal consolidations’, Scandinavian Journal of Economics. Ram, R. (1986) ‘Government size and economic growth: a new framework and some evidence from cross-section and time-series data’, American Economic Review, 76, pp. 191–203. Ravn, M.O. and J.R. Sørensen (1997) ‘Minimum wages: curse or blessing?’, Research in Labor Economics (San Francisco: JAI Press). Röell, A. and O. Sussman (1997) ‘Stabilization’, European Economic Review, 41, pp. 279–93. Sandmo, A. (1991) ‘Economists and the welfare state’, European Economic Review, 35, pp. 213–239 Sandmo, A. (1995) ‘The welfare economics of the welfare state’, Scandinavian Journal of Economics, 97, pp. 1–8. Sinn, H.W. (1995) ‘A theory of the welfare state’, Scandinavian Journal of Economics, 97, pp. 495–526. J. Søndergaard (1997) ‘The welfare state and economic incentives’, Chapter 10 in this volume. Tanzi, V. and D. Fanizza (1995) ‘Fiscal deficits and public debt in industrial countries 1970–94’, IMF Working Paper, 95–49. Varian, H. (1980) ‘Redistributive taxation as social insurance’, Journal of Public Economics, 14, pp. 49–68. World Bank (1993) The East Asian Miracle: Economic Growth and Public Policy (New York: Oxford University Press for the World Bank).
Comment Per Callesen The authors of Chapter 9 look for conclusions on the importance of total public expenditures and taxes for the overall macroeconomic performance. The chapter deals with potential effects on the volatility of growth, the average rate of growth and the size of public debt. It includes theoretical discussions of the pros and cons for such aggregate effects and presents some correlations of cross-country effects.
1
EXPENDITURES, TAXES AND VOLATILITY
The chapter identifies three effects, by which public sector activities should potentially reduce fluctuations in the aggregate growth rate. These are: (a) the dead weight effect from the public sector (primarily the production of welfare services) being little or not at all subject to market forces, (b) potential counter-cyclical variations of these activities, most possibly by discretionary action and (c) the stabilizing effect on private sector activities by taxes and transfers raising and lowering net revenues according to high or low economic activity due to both automatic stabilizers and discretion. The authors conclude that evidence supports the public sector effects and to some extent the stabilizing effects on the private sector, the evidence for this, however, being less clear. In case of unchanged private sector fluctuations an increase in the share of cyclically invariant non-market government services from 20 to 30 per cent of the economy would mechanically reduce the standard deviations of local GDP growth of historically around 2–3 per cent by 0.2–0.3 percentage points. Such an effect should also be detected in cross-country comparisons. The authors even detect some counter-cyclical movements in public sector activities, which should add a further effect, but probably a small one. Stabilizing effects on private sector activities should in principle be rather strong, even in the absence of fiscal discretion. In recoveries this is due to the automatic stabilizer effects of rising tax revenues and lower expenditures on transfers not being recycled as the economy picks up but instead improving the public finances. The reverse happens in recessions. The authors test the cyclical properties of the average taxes–GDP ratio and find some stabilizing effects. A priori such an effect on the taxes-GDP ratio should however also be expected to be quite small1 as most income and consumption is subject to more or less proportional taxation. The absence of a pro-cyclical tax ratio (or a small effect) does not imply a lack of automatic stabilizer effects from taxes. On the contrary the absence of a strong contra-cyclical tax ratio in itself implies strong budget effects when the economy fluctuates. A very progressive tax system might create some automatic tax ratio effects as a result of fiscal drag.2 Also the revenues from company taxation and indirect taxes on cyclically sensitive consumer goods do historically (in Denmark) show an elasticity vis-à-vis GDP fluctuations higher than 1.0, thus implying a small contribution to a
285
286
Callesen
higher taxes-GDP ratio in booms. But such effects are somehow dominated by dead weight effects from capital income taxation being less cyclical and revenues from taxation of income transfers varying against the cycle.3 Based on econometric models both the Economic Council 4 and the Ministry of Finance5 provide evidence of stabilizing effects from both automatic stabilizers and discretionary policy changes in Denmark. The actual standard deviation of GDP growth 1978–97 has been 1.5 per cent. Excluding model-based first-year effects of discretionary fiscal policy changes the deviations would have been 1.9 per cent according to the Ministry of Finance. Excluding also the effects from automatic stabilizers the deviation would have been 2.4 per cent according to the study. These effects assume that private consumption is affected by changes in private sector disposable income and is not counterbalanced by changes in the savings ratio because of a Ricardian equivalence effect. Such an effect cannot be excluded, but in the case of the automatic stabilizers this requires a reinforcing effect from consumers first spending more, thus creating improvements of the budget, and second spending even more based on this improvement, even if rational consumers should know that the budget improvement is cyclical. This is not likely. To the extent countries with less automatic stabilizers rely heavier on monetary policy to stabilize the economy – i.e. through inflation targetting – cross-country comparisons may not find significant differences in the variance of GDP growth due to fiscal factors. Comparisons do, however, suggest such an effect. In the years 1978–97 standard deviations of GDP growth in the United States and United Kingdom was about 2.0 per cent, compared to about 1.5 per cent in the North-Western European countries excluding the United Kingdom, in which levels of taxes and spending are at a higher level.
2
EXPENDITURES, TAXES AND GROWTH RATES
The authors rightly discuss an extended list of counteracting structural forces affecting the link between the overall public spending, tax levels and economic growth. Also the direction of causality may pose problems for the attempt to sort out clear conclusions which support the expected result, that high expenditures and taxes might reduce economic growth. A few more factors might be added. First, public spending for care of children and the elderly may not only technically raise registered production, but the resulting increase in the labour force participation rate most probably also reflects a more efficient use of resources. Secondly, an extended social safety net may contribute to other structural policies being more flexible, examples being less pressure for employment protection, subsidies for sunset industries, overregulation of semi-public industries and even trade barriers. Thirdly, an even distribution of income may limit the necessary private and public resources used on the prevention and punishment of crime. Basically the composition of expenditures and taxes may potentially be much more important than the levels in per cent of GDP. A country devoting a large share of public expenditures on infrastructure, education and labour force-promoting activities should in terms of GDP be better off than a country with lower overall expenditures, but concentrating on transfers payments and other welfare services. A country may also limit expenditures by high degrees of means-testing against the income of the recipients, thus raising effective marginal tax rates to the same extent or more than in
Comment
287
a country with less means-testing but higher public expenditures. Also the composition of taxes is of great importance. A broad tax base with low rates may distort less than a narrow base with high rates, even if the country with high rates has a lower taxes-GDP level. To sort out these questions one needs a very detailed general equilibrium model covering all the issues involved. Supposedly studies based on such a model – if carefully built – would confirm that for economic policy it is very important to focus on the composition effects. To sort out the question of the general level one may in such a study create a benchmark which is a proportional shift in all taxes and expenditures. But in policy terms even this involves an assumption of a given fiscal package which may probably not be considered likely. Not having such a model it is of course of some interest if all the composition effects net out to such an extent, that cross-country comparisons do indicate the sign of the overall effects. Such comparisons however also suffer from severe measurement problems because of different institutions in the countries compared. In Denmark it would, for example, be quite easy to change some institutions which would as a result reduce measured levels of expenditures and taxes by large amounts while not affecting the well-being and incentives of individuals. The major thing to do would be to pay out all social transfers net of taxes, thus adjusting the level of present tax payments out of transfers (6 per cent of GDP). The cash subsidy for all children could be converted to a tax deduction (1 per cent). Indirect taxes could be lifted from the government’s procurement of goods and services (1 per cent). Further changes in this direction would be to implement means-testing to a larger share of public transfers and service, thus switching the marginal tax effects onto the expenditure side (expenditure-taxes).6 Of course changes in the other direction are indeed also possible, such as converting tax expenditures, the personal income tax allowance and the tax value of interest deductions into subsidies. The point is, that these institutions differ between countries to such an extent that correlations based on unadjusted expenditure and tax ratios may be severely affected. These differences are not necessarily correlated with ‘true’ expenditures and taxes, but the uncertainty in the author’s Figure 9.3 (p. 275) is big enough for such adjustments to potentially change the sign. Some attempt has been made to adjust for differences in institutions. According to a Danish study,7 one group of adjustments would reduce the difference between Denmark and Germany in expenditures and taxes as a percentage of GDP by no less than 10 per cent of GDP, the difference with the United Kingdom by 5–7 percentage points and make smaller adjustments vis-à-vis Sweden and the Netherlands. One structurally interesting comparison between countries could cover weighted effective marginal tax rates including all taxes, social security payments and the effects of means-testing. Of course the data requirements for such a comparison should not be underestimated. Looking at the authors’ regression it suggests that a 10-percentage point higher level of expenditures and taxes reduces annual economic growth by 0.2 per cent – a figure which compares to the 1991 Barro study, which included more variables, but was also based on reduced-form regressions. Interestingly, this result is exactly the mechanical effect one would expect to find due to the measurement technics of public sector growth used in most countries. In the national accounts public sector production is deflated by the price of inputs (mainly wages) thus by definition excluding any measured productivity growth except from minor composition changes.
288
Callesen
If private sector growth is, say, 2 per cent higher than public sector growth8 – then countries with identical growth rates in the private and public sector would differ in average growth rates by 0.2 per cent, if public consumption and taxes differ by 10 per cent of GDP. The point is not that increases in the effectiveness of public sector activities are slower (which they probably are as they are service activities, thus affected by ‘Baumol’s Law’). The point is, that such increases are by definition not recorded. In order to draw policy conclusions comparative studies should thus actually look for effects being bigger than 0.2–0.3 per cent per 10-percentage point difference in expenditures and taxes. In a recent survey the OECD9 highlights the fact that a 10-percentage point difference in tax ratios seems to explain growth differentials between countries of 0.5 percentage points in the period 1980–95, thus allowing for some effects in addition to the mechanical effect. To what extent this result is biased by the short period covered and the measurement problems discussed above is not clear. Yet another way to proceed could be to compare GDP levels and growth rates split up into three components: employment rates, hourly productivity and average working hours. A quick look at these would suggest no effects on employment. The highest employment figures in the OECD are found in the United States, Japan and the Nordic countries, which in different directions are the extremes on expenditure and tax levels. Productivity comparisons might prove to have a small effect, whereas one could not exclude a significant effect on working hours. The highest figures for working hours are found in the United States, the United Kingdom and Japan, where marginal tax rates are low. The lowest figures are found in Germany, France, the Netherlands and the Nordic countries, where marginal tax rates are high.
3 GOVERNMENT DEBT AND THE SIZE OF TAXES AND EXPENDITURES The authors discuss and test if high levels of expenditures and taxes are associated with high debt levels. This is cautiously confirmed. In the splitting up between transfers and public consumption an effect associated with transfers is detected while no effect from consumption is there. First, the result is plausible, because what in most countries created the big rise in public debt was deficits generated by the large increase in unemployment which took place especially in the early 1980s and which later proved to be the reflection of an underlying structural increase. Whereas high levels of transfers and taxes smooth economic fluctuations around a stable trend they also create larger structural deteriorations of the budget, when economic activity undergoes structural changes. It is thus no surprise, that countries with high transfer expenditures experienced the largest deficits and thus a stronger increase in debt and a more difficult consolidation period afterwards. Secondly, a suggestion would be that myopia is a minor explanation for government debt-building compared to the effects of recommendations which for public choice reasons (difficult to decide for more spending when there is a deficit) indirectly lead to permanent deficits on public finances. If one operates government finances under the assumption that surpluses are unhealthy, then it is clearly unlikely that debt will be reduced. Thirdly, the historical experience of the link between expenditures and debt need not be repeated in the future. On the contrary, the Nordic countries, with the high
Comment
289
expenditure and tax levels, are those with the most ambitious targets for government finances. The Nordic countries have officially targeted surpluses on average over the business cycle, whereas most countries target balance or small deficits.10 Hopefully also countries with large expenditures and taxes have learned from the historical experience which suggests that high expenditures and taxes need to be combined with a solid balance and deficit position in order to let automatic stabilizers and some contra-cyclical discretion work. Measurement problems in comparing public debt figures are even larger than the (important) problems with expenditure and tax levels. A look at OECD’s figures for net general government financial debt might solve some problems, but would also add others – i.e. the allowance for assets in public pension funds is not necessarily contributing to the right picture.
4
CONCLUSION
I welcome the chapter by Andersen and Schmidt. Reading it was stimulating and thought-provoking. The analysis is well balanced and the basic conclusions are shared.
Notes 1. 2. 3.
4. 5. 6. 7. 8.
9. 10.
What Table 9.2 reports is contributions to budget improvements through taxes and expenditures and not effects on taxes and expenditure ratios. Tax brackets being adjusted less than average income is growing. Estimates by the Ministry of Finance suggest a short-run elasticity between GDP and taxes in Denmark of about 1.2 (Finansredergørelse 96, Finansministeriet, 1996). A cyclical increase in GDP (about 1.000 billion DKK) of 1 per cent – about 10 billion DKK – would thus raise taxes from about 500 to 506 billion DKK. The tax ratio would increase from about 50 to 50.1 per cent. The 6 billion DKK improvement in tax revenues is 0.6 per cent of GDP, which is slightly higher than the EU estimate of 0.5 recorded in Table 9.2. The reason for the elasticity being above 1.0 – and thus allowing for a small pro-cyclical tax ratio effect – is mainly changes in the composition of GDP as the cyclical variations in the highly taxed car sales is much stronger than other components of GDP. Such an effect is probably smaller in most other countries. In that case the expected cyclical tax-ratio effects should be approximately zero. Dansk Økonomi, Forar (1996), Det Økonomiske Råd, (1996). Finansredegørelse 97 (Finansministeriet, 1997). Such changes would probably imply not only technical effects but also some distributional and incentive effects at the micro level. Budgetredegørelse 94 (Finansministeriet, 1994). In case of 2 per cent higher productivity increases and parallel growth in employment in the two sectors. In that case the share of public consumption would be approximately constant over time, provided that wage increases are parallel. Price increases in the private sector would, however, be 2 per cent lower than measured price increases in the public sector. ‘Taxation and Economic Performance’, Working Paper 176 (Paris: OECD, 1997). OECD, Economic Outlook, 61 (Paris: OECD) (June 1997)
Comment Douglas A. Hibbs, Jr 1
INTRODUCTION
Chapter 9 aims to investigate empirically, using data for EC and OECD countries, some macroeconomic consequences of the welfare state, where the ‘welfare state’ is characterized by the scale of public revenue and expenditure in relation to gross domestic output. The authors concede at the outset that the large tax burdens required by big welfare states (which in practice never rely significantly on the lumpsum taxes favoured by optimality demonstrations in high public finance theory) may well yield distortionary costs – that is, may induce inefficient economic activity driven more by the tax code than by maximizing behaviour in unfettered markets. Andersen and Schmidt argue, however, that equity considerations aside1 there may be macroeconomic upsides to the large-scale fiscal state which offset to some degree, and perhaps even completely, the distortionary effects commonly emphasized by economists, especially by microeconomists who frequently argue from theoretical first principles without much reference to systematic empirical evidence. Macroeconomic stabilization is the potential benefit of welfare state fiscal policy that Andersen and Schmidt supply the most new evidence on, and so I devote most of this comment to a critical review of their evidence and arguments pertaining to this issue. My remarks track sequentially the flow of their chapter.
2
THE PUBLIC SECTOR AND MACROECONOMIC VOLATILITY
Income-contingent public expenditures and revenues imply that fiscal systems automatically stabilize macroeconomic activity. Andersen and Schmidt present data showing that in every EC country aggregate revenues in relation to GDP are more cyclically sensitive than aggregate expenditures. They go on to investigate whether total budget sensitivity – measured as the output elasticity of revenues in proportion to GDP less the output elasticity of expenditures in proportion to GDP – becomes more pronounced as the size of the public sector (measured by revenues in per cent to GDP) rises as one looks across EC countries. They remark that a positive relation ‘is to be expected’ and is ‘a natural consequence’, but I see no reason in principle why the output elasticities of such fiscal ratios should be scale-dependent. The evidence on this issue graphed in Figure 9.1 is claimed to demonstrate the expected positive relation ‘very clearly’, but in my view this conclusion is not well founded. Reading a positive association into the data depicted in Figure 9.1 depends wholly on the high revenue – GDP ratios prevailing in cyclically sensitive Sweden and (to a lessor extent) Denmark, both small contributors to total EC output. At total budget sensitivity of 0.6, revenues in per cent to GDP range from around 37 (United Kingdom) to 53 (Finland), and at 0.5 total budget sensitivity the revenue – GDP per cent varies from around 35 (Ireland) to 50 (France). Since the relation of budget sensitivity to public sector size implied by the data for the big EC players (the
290
Hibbs
291
United Kingdom, Italy, Germany and France) is nil or negative, if weighted by the GDPs of EC member countries, there is no sign at all that the larger is the public sector, the more cyclically sensitive (and, hence, the more macroeconomically stabilizing) is the fiscal system. My interpretation of the evidence is reinforced by the data graphed in Figure 9.2 on the variability of output growth (standard deviations of detrended log output) in relation to public sector size – now measured by average tax rates – which the authors themselves describe as ‘weak’. (It was even weaker in an earlier version of the chapter that used the same revenue – GDP ratio used in Figure 9.1 to measure public sector size). Andersen and Schmidt pursue further the issue of whether the public sector stabilizes the macroeconomy by introducing the identity (9.1) y (Total output) = yp (Private output) + yo (Public output) and noting that total output variance may be decomposed
var( y) = var( y p ) + var( yo ) + 2 cov( y p , yo ) They point out that other things equal macroeconomic volatility is damped by the size of public output if var(y0) < var(yp), if cov (yp, yo) < 0 and if taxes and transfers are designed to stabilize directly var(yp). Although this decomposition is intended to motivate the analyses reported in Tables 9.3–9.5, the connection is tenuous because Andersen and Schmidt use filtered output series (which are roughly akin to proportional growth rates of output) in their empirical work. Moreover, the fiscal variables investigated (tax and social security transfer rates) correspond neither to public consumption nor to taxes net of transfers, which are the only public sector variables that map onto yo in the output identity and its decomposition developed in the chapter and written out again above. Nonetheless, the data in Tables 9.3–9.5 are certainly relevant to the issue of the public sector’s contribution to macroeconomic stability, analyzed here for a sample of OECD rather than EC countries. Table 9.3 (last column) implies some stabilizing effect of social security transfers, in that filtered transfers (roughly, transfers net of trend growth) rise when filtered output falls. Yet this pattern says nothing about the effect of public sector size on macroeconomic volatility (the correlation statistics used are scale-free). Indeed, the highest correlation reported is for the United States (among the least developed welfare states) and the lowest correlation is for Norway (among the biggest welfare states). And unlike the revenue measure reported in Table 9.2, the correlation between filtered aggregate tax rates and filtered output exhibits a positive, ‘stabilizing’ sign in only five of the 14 OECD countries shown, though one of them is the United States, a low-spending welfare state that produces nearly one-half of total OECD output. The composition data in Table 9.4 are more encouraging to the hypothesis that big public sectors are stabilizing, since in all countries analyzed the variance of public consumption is only a small fraction of the variance of private output. Moreover, in 10 of the 14 OECD countries public and private output covary negatively. However, the correlation is positive or nil for two of the three largest OECD economies (the United States and Germany), and so from the point of view of aggregate OECD output the implication would be that stabilization by the public sector is weak or even perverse via this channel. It is of course true, as Andersen and Schmidt note, that public sector activity could be designed to stabilize private activity, as the correlations they report imply is especially true in, for example, Denmark and Australia. Results for regressions of private sector macroeconomic volatility on various fiscal variables undertaken with cross-national data appear in Table 9.5. The last regression
292
Hibbs
– row (8) – is probably more informative than the other ‘partial’ regression experiments, and it subsumes some of the earlier analyses in systematic fashion. Using a generous standard of statistical significance (parenthesized ‘p-values’ of 0.10 or smaller) yields the following inferences about conditional fiscal influences on output volatility: (1) high average tax rates, taken alone, are destabilizing, (2) fiscal systems with positive correlation between filtered movements in private output and the aggregate tax rate are stabilizing, (3) high average social security transfer rates, taken alone, are stabilizing and (4) fiscal systems with negative correlation (which probably includes all the countries studied, given the correlations reported in Table 9.3) between filtered movements in private output and the social security spending rate are destabilizing. On the whole these results seem to me to have pessimistic implications for the idea that large public sectors (‘welfare states’) may yield significant benefit in the form of macroeconomic stabilization. The average tax and spending rates are the only variables in Table 9.5 germane to assessing the statistical influence of public sector size. Since high spending requires high taxation, effects from (1) and (3) tend to offset each other.2 Similarly, according to Table 9.5 countries with pro-cyclical average rates of taxation and counter-cyclical average rates of social security spending – the combination one would think most likely to reveal a pattern of fiscal stabilization – evidently also experience offsetting fiscal effects on output volatility.
3 THE PUBLIC SECTOR AND GROWTH AND DEBT ACCUMULATION Section 9.3 and 9.4 of Andersen and Schmidt’s chapter supplies informed reviews, but less empirical value-added, concerning the influence of welfare state fiscal policy on economic growth and the accumulation of public debt. The former topic, in particular, is the object of a voluminous literature which has delivered few, if any, conclusions that command consensus. The simple bivariate plot in Figure 9.3 of average growth rates against average tax rates (aggregate taxes in per cent to GDP) illustrates this point. The relationship is vaguely negative, yet I find the very modest growth rate ‘response’ to (international) variation in aggregate tax burdens amounting in magnitude to 20 percentage points of GDP to be quite startling. In the absence of obvious and sizable depression of growth effects from big international differences in OECD tax rates, Andersen and Schmidt properly focus on the composition of fiscal policy as the place to learn about the upsides and downsides of welfare state fiscal policy. Their reading of the evidence leads to endorsement of the sensible ideas that while taxation probably depresses economic growth, public expenditure on physical infrastructure may enhance growth, and expenditure on human capital formation most probably does enhance growth of productivity and living standards.3 Income security and redistributive transfer spending, on the other hand, appears to be the main fiscal variable driving the accumulation of public debt (Table 9.7), yielding what Andersen and Schmidt describe as ‘a systematic deficit bias’ in the highly developed welfare states. And research noted by Andersen and Schmidt indicates that deficit accumulation induces increases in the real interest rate, an important factor affecting the rate of investment – and, hence, presumably the rate of economic growth. This sequence of effects further implies that whatever contribution income contingent-transfer spending may make to stabilization of output volatility may levy costs on the secular rate of output growth. Trade-offs of this kind between the time-conditional mean and variance of output, however, are not developed in Andersen and Schmidt’s chapter.
Comment
293
Notes 1. 2.
3.
Of course for some, including me, the contribution of fiscal policy to equality is by itself a strong justification of the welfare state. Andersen and Schmidt write that the ‘net’ effect of taxes and transfers is stabilizing, perhaps because the sum of regression coefficients for the average tax and spending variables in their regressions are negative. But because the tax rate variable is global and the spending-rate variable is particular, one cannot add coefficients to obtain a meaningful reading of the net effect of fiscal policy averages. Here, as elsewhere, however, issues of reverse causation arise. The research Andersen and Schmidt report indicating that the rise in OECD secondary school enrolment rates from 70 per cent to 95 per cent over 1960 to 1985 was associated with a 0.6 percentage point per year rise in productivity growth, illustrates the point. The foregoing is cited as evidence that ‘education contributes to increasing productivity’ (p. 277). Yet a secular increase in the proportion of young workers receiving training in formal schooling, as opposed to low-productivity, in-the-labour-force, on-the-job training (apprenticeship schemes and the like), would create observed increases in productivity growth that were a measurement by-product rather than a causal response to formal (and publicly financed) human capital formation.
10 The Welfare State and Economic Incentives Jørgen Søndergaard
10.1
INTRODUCTION
The Danish welfare state like other welfare states is largely a post-war achievement although some of its elements were founded towards the end of the nineteenth century. Gradually social rights and entitlements have been added to the civil and political rights of citizenship. The essential idea behind the welfare state is to break the direct link between uncertain outcomes of the market economy and individual standards of living. Thus, it is no surprise that economists tend to see the welfare state as an intervention in the economic system aiming at a more equal distribution of incomes and opportunities in society than would appear in a laissez-faire market economy. Introducing distortionary redistributive measures in an otherwise distortion-free economy bears a cost in terms of efficiency. This big trade-off assumption that has been with us for a long time was reinforced by the elegant metaphor of the leaky bucket in Okun (1975). A similar perspective can be found also in current policy debates both in Denmark and elsewhere about the viability of the existing social security system and the welfare state in its present form.1 Political controversies focus on two sets of interrelated issues: (1)
(2)
possible disincentive effects embedded in the social security system that may cause labour market rigidities that enhance dependence on social security difficulties of financing the welfare system because expenditures in particular on social security are growing faster than real incomes.
The rising share of social security expenditures is common for many European countries (see Ploug and Kvist, 1997). However, the causes behind this development have not been clearly identified, although it is widely held among economists that European welfare systems carry with them some disincentive effects and so bear some of the blame. On its premises – in particular, the assumption of other distortions being absent – the perception of costly redistributions is, of course, indisputable. 294
The Welfare State and Economic Incentives
295
However, the direct inference to imply that a Scandinavian-type welfare state distorts incentives and creates allocative inefficiencies and poorer macroeconomic performance is questionable on at least three grounds: (1) (2) (3)
the Scandinavian welfare state is more than a redistributive mechanism there are market failures and inefficiencies also in the laissez-faire economy distortions of incentives are not always very influential on economic behaviour and therefore often not tractable in indicators of macroeconomic performance.
If the broad definition of the welfare state in Sandmo (1995) is accepted, one may distinguish between four characteristics or economic dimensions of the Scandinavian-type welfare model (see Pedersen and Søndergaard, 1996). These dimensions are not completely separate, as any particular element in the welfare system may share several characteristics. However, the distinction emphasizes the fact that different elements of the welfare state may rest on a broader justification from welfare economics than just income redistribution. The redistributive dimension is, of course, an important characteristic. In the Danish system this is most obviously inherent in the progressive income tax system and those income transfers that are not part of the safety net in case of involuntary loss of earned income, notably the old-age pension system and family allowances. At an earlier stage the public old-age pension system could also be seen as a remedy for imperfections in capital and insurance markets. However, as this part of the market economy has gradually become more efficient, the primary function of public pensions has become income redistribution. The welfare state also contains an insurance dimension. By providing a safety net to compensate for involuntary losses of earned income due to unemployment, sickness or disability, the welfare state is at least partly a substitute for private insurance. Some justification for public insurance rather than private could be found in inefficiency problems of private insurance markets due to asymmetrical information. Problems of adverse selection in private insurance markets lead to shortage of supply of coverage against certain risks. This problem can be avoided by compulsory (public) schemes (see Sandmo, 1995). To the extent that public authorities have better access to information or monitoring than private insurance companies, public insurance may also be more efficient in dealing with moral hazard problems. An additional feature of the welfare state particularly important in Scandinavia is a care dimension – that is, public provision of care for children and elderly people. In a non-distorted economy this could hardly be justified in economic terms, since cash transfers then are always superior to transfers in kind. However, as pointed out by Blomquist and Christiansen (1995), for
296
Søndergaard
example, in a distorted economy this may look different as transfers in kind may be welfare-improving by alleviating incentive compatibility constraints on other tax and transfer instruments. It is therefore not public provision as such that is the most important aspect, it is rather the built-in subsidy to close substitutes for household provision of care that may have an impact on market behaviour of households. The welfare state also has an important human resource dimension in providing education and training (or subsidies to education and training), health care and rehabilitation of disabled persons. Also, in these areas state provision can be justified at least to some extent in terms of alleviation of market failures. If the welfare state overcomes some deficiencies of markets for health and education of a laissez-faire economy, it may improve economic performance rather than hurt it. The existence of market failures can make redistributive policies that would have been harmful under Arrow–Debreu assumptions beneficial in efficiency terms. An example is the possibility of improved efficiency in human capital investments because social inheritance and hence income and wealth of the parents are important determinants of their children’s education (see Hoff and Lyon, 1995). Another efficiency argument related to redistribution is presented by Sinn (1995), who points out how the welfare state may enhance efficiency by stimulating private risk taking. With this broad definition of a welfare state an a priori assessment of its costs in terms of efficiency losses is impossible. In theory there need not even be a trade-off between equality and efficiency, since we are faced with a comparison of two economic systems each of which contains several inefficiencies. Ultimately, therefore, whether there is a cost of the welfare state in terms of lower average welfare than in the laissez-faire market economy is an empirical issue, and a balanced assessment must include both the disincentive effects from the welfare system and the alleviation of market failures. This should be borne in mind as the rest of the chapter will address incentive problems only. Assessing disincentive effects raises two interrelated difficulties. It is often implicit in economic analyses of welfare systems that the point of reference is a laissez-faire economy with no welfare provisions at all. Although what society would look like without government provision of welfare is hard to imagine, this assumption tends to overstate the efficiency loss. All four dimensions of the welfare state meet actual demands of individuals and households and some level of private provision would no doubt be in place if state provision were absent. While we may have an idea of how, for example, pension savings and disability insurance might look like in a situation without public schemes, to imagine what unemployment insurance or voluntary redistribution through private charity would look like is harder. Obviously
The Welfare State and Economic Incentives
297
not only public unemployment insurance, but also any kind of private unemployment insurance would raise incentive problems, and ideally an assessment of disincentive effects from the public system should be based on a comparison with this unknown situation. Because of these unknown features of the reference situation it is not at all clear how much the welfare state actually changes incentives. Furthermore, even if incentives are distorted the impact on actual behaviour is as mentioned an empirical question. We can easily think of cases where the resulting effect on behaviour and therefore on aggregate economic variables is small. Yet this may hide important distortionary effects in terms of loss of individual welfare owing to substitution effects, as has often been found in estimates of tax and benefit effects on labour supply (uncompensated wage elasticities close to zero and much smaller than compensated elasticities). The purpose of this chapter is to give a description of the incentive structures of the Danish welfare system and review the evidence on their behavioural impact. I shall concentrate on labour market incentives, while incentive problems related to savings, investment, and lack of competition in public provision will not be discussed. Section 10.2 provides a very brief review of the current Danish welfare system. Section 10.3 looks at the aggregate size of the welfare system. Section 10.4 addresses different types of labour market incentives. Section 10.5 summarizes the main findings.
10.2
THE DANISH WELFARE SYSTEM
The Danish welfare state belongs to the ‘Scandinavian type’, the main principles being universalism in coverage and general tax financing. However, in reality some elements are borrowed from other models – for example dependence on labour market participation or on contributions also occur occasionally in Denmark. A special feature of the Danish model is the widespread use of flat rate benefits in social security. Thus, there is less dependence on previous income in Denmark than in other Nordic countries. Finance by general taxation is also more pronounced in Denmark than in the other Scandinavian countries, where employers’ social contributions are much more important. Social Security The present system contains a large number of different benefits (see Table 10.1). There is a separate law for each type of benefit. Benefit levels vary to some extent across the system, as do durations and conditions of entitlement. The most recent elements are parental and educational leave schemes. An addition to the early retirement scheme (post-employment
Characteristics of the Danish system of income transfers
Coverage age
Eligibility criteria
Source of finance
Max benefit pr. month (1995)
Replacement Max rate (single duration APW; 1995) gross net of tax
(a) Means test (b) Dependence on other criteria
Old age pension
Age 67–
None
Taxes
7.560 DKK
44
No limit
(a) Fully on earned income, partly on other income. (b) On number of years of residence
Disability pension
Age 18–67
Health or social disability
Taxes
Until age 67
(a) Yes (b) On severity of disability
Early retirement benefits years
Age 60–67
Unempl. insurance for 20
See 11.072 DKK unemployment insurance
57
66
Until age 67
(a) Yes if more than 200 hours of work (b) Age; max 90 per cent of previous wage
Unemployment Age benefits 18–67 Voluntary
1-year insurance and employment
Contribution + 11.072 DKK general tax on earned income
57
66
5 years; longer for age 50–60
(a) No (b) Max. 90 per cent of wage; participation in active measures after 2 years
Social assistance
Age 18–
None
Taxes
(11.072 DKK) 34a
44a
No limit
(a) Fully on all income and wealth with a low threshold (b) Participation in active measures, no. of children, rent.
Leave schemes
Age 18–67
Different Taxes for educational and parental leave
11.072 DKK
66
1 year
(a) No (b) Type of leave scheme; max 90 per cent of wage
Child allowances
Parents
Children below 18
.800 DKK
18 years
(a) No (b) Age of child
Taxes
57
56
298
Table 10.1
Table 10.1
(Cont’d)
Coverage age
Eligibility criteria
Source of finance
Max benefit pr. month (1995)
Replacement Max rate (single duration APW; 1995) gross net of tax
(a) Means test (b) Dependence on other criteria
Maternity leave
As sick pay
As sick pay
Taxes
11.072 DKK
57
66
30 weeks
(a) No (b) Max 100 per cent of wage
Sick pay
Age 18–
1/4 year Taxes employment
11.072 DKK
57
66
1 year
(a) No (b) Max 100 per cent of wage
Rehabilitationsupport.
Age 18–
Disability
Taxes
11.072 DKK
57
66
No limit
(a) No (b) Visit required
Housing benefits
Age 18–
Rented apartmentb
Taxes
–
No limit
(a) Yes (b) Pensioner/non-pensioner, children, rent and flat size
Educational grants
Age 18–
Enrolled in education
Taxes
–
Normal length of study
(a) Yes (b) Age; independent housing
a b
Social assistance is supplemented by housing allowances not included in the replacement rate. For pensioners, all types of dwellings are eligible for housing benefits.
299
300
Søndergaard
wage for 60–66-year-olds) was introduced in 1992 allowing retirement for unemployed above 55 years of age, extended in 1994 to cover the age bracket 50–55 years. This addendum to the system was only in effect for a short period and is now closed for new entrants. Most benefits are flat rate benefits. With pensions as a main exception there is a limit at 90 or 100 per cent of the previous wage. Unemployment benefits, child allowances, maternity and sick pay, leave scheme benefits and rehabilitation support are not subject to any kind of means test. The remaining kinds of transfer are fully or partly means-tested. Old-age pensions, for example, consist of a basic pension that is independent of income except labour income, and an additional pension depending on total income including income of the spouse. No part of the pension depends on wealth. Early retirement benefits are means-tested against earned income and private pension income. Social assistance is fully means-tested above a very small threshold (including income and wealth of the spouse). Housing benefits are means-tested against total household income. The pension system is under change. In 1964, a small compulsory contribution financed and funded labour market pension scheme (ATP) was introduced and gradually most wage contracts have come to include funded occupational labour market pension schemes. This will gradually make the public, pay-as-you-go taxfinanced old-age pension system relatively less important in the future. The structure of net replacement rates within the whole system is rather opaque, and incentives vary across types of benefits and depend on the circumstances of the receiver. The importance of this is enhanced by the difficulty of assigning individuals to a specific type of benefit in an environment of high and increasing dependence on public support. In general terms, the net replacement rates at the lower end of the income distribution are high by international standards, but rather moderate at (and above) the average level of income owing to the flat rate structure. Apart from small contributions to unemployment insurance the system is financed by general taxes.2 Administration, however, varies significantly. Unemployment benefits are administrated by private, union-controlled unemployment insurance funds that also collect a premium to cover administrative costs and a small fraction of expenditures on benefits. Most other benefits are administrated by the municipalities with full or partial reimbursement from central government. Reimbursement varies across types of transfers. Active labour market policies that accompany the unemployment benefit system are administrated at the regional level by local branches of the Ministry of Labour. Individual rights (and obligations) regarding active measures are somewhat unclear. The importance of active labour market policies has increased in recent years and the control of the availability for work condition for entitlement to unemployment benefits has been tightened. The
The Welfare State and Economic Incentives
301
maximum duration of (passive) unemployment benefit is now 2 years. After 2 years on benefits follows 3 years of compulsory public work where working hours are reduced so that the total wage income exactly equals unemployment benefits. Prior to 1994 such public work or job training qualified for renewal of entitlement to benefits, but this is no longer the case. Unemployed between 50 and 60 years of age can maintain eligibility for unemployment benefits if they will be eligible for early retirement benefits when 60. Unemployed who are not covered by unemployment insurance (UI) get social assistance and in this system activation in particular for young unemployed usually starts immediately. In 1996 the maximum duration of unemployment benefits from UI funds to unskilled young unemployed has been shortened. The maximum length of benefits is now 6 months after which unemployed below 25 with no formal education are offered education or training for 18 months. During this period unemployment benefits are replaced by a special educational grant of half the unemployment benefit rate. Education and Health The main principle characterizing this part of the welfare state is public provision free of charge. Only a few caveats are in place. Primary health care is largely based on private doctors but with a few exceptions – like, for example dentists – their service is provided free of charge to the customer. The bill passed on to the public sector is based on standard rates. In primary education private schools count for about 10 per cent of the pupils. Public schools are fully tax-financed, while private schools get a public subsidy per pupil covering 85 per cent of standized expenditures leaving only rather modest charges to the parents. From the 18th year students are entitled to educational grants. The size of the grant depends on age and on whether the student is living with the family household or independently. The grant is means-tested against other sources of income to the student (above a certain threshold), but not against the income of the parents. In all areas except hospitals, higher education and some types of apprenticeship training, capacity is normally adapted to increases in total demand. However, in hospitals and universities capacity has generally not been sufficient to meet demand. In higher education a rationing system based on average grades from secondary schools (with some specific modifications) has been in effect since the late 1970s, but effective rationing is now rapidly decreasing owing to both increased capacity and smaller youth cohorts. A somewhat discretionary queueing system is applied in the hospital sector. In recent years, however, patients have increasingly been allowed to choose among all hospitals in the country, making the individual patient less dependent on the waiting time at a particular hospital.
302
Søndergaard
Care There are two main areas of public provision of care: day care for children and care for old-aged and disabled persons. The latter is generally provided free of charge, but based on a visit by the social department of the municipality. Day care for children is provided at a price covering between a quarter and a third of total costs of provision. For parents with low incomes the price is reduced. The reduction, which is means-tested, may cover the full price and is particularly important for lone parents. In several municipalities capacity has been insufficient to meet total demand and so parents have faced a waiting period of sometimes several months to obtain a place. Overall waiting time is decreasing in these years because capacity is expanding. In 1996 two out of three children between 6 months and 2 years of age and six out of seven between 3 and 5 years of age used a public day care facility. Taxation The main components of the Danish tax system are a progressive personal income tax (45.6 per cent of total revenue in 1996), a flat rate tax on earned income (7.7 per cent), a flat rate VAT and substantial excises on alcohol, tobacco, energy and cars (31.9 per cent), a flat rate corporate income tax (5.0 per cent), a special real interest tax on pension funds (3.9 per cent) and property taxes (2.0 per cent). A small wealth tax is being phased out and employers’ social security contributions have always been almost insignificant in Denmark. Compared with other North European countries the main difference is a higher effective taxation of consumption, while effective average taxation on labour and capital income is in line with other countries (see Lassen and Nielsen, 1996). A recent comparison of seven North European countries based on 1994 data shows that disposable income as a fraction of total labour cost to the employer is slightly higher in Denmark than in France, Germany and Finland, and at the same level as in Sweden and the Netherlands (see Hansen et al., 1995). Among the seven countries, only the United Kingdom has a higher ratio of disposable income to total labour costs. However, from the perspective of incentives marginal tax rates are more important than average rates, and on average Denmark has higher marginal taxes on labour income (and on some kinds of capital income) than other countries. In 1996 taxation of labour income has four brackets with marginal tax rates varying from 7 per cent (up to 33,000 DKK of gross income) to 65.3 per cent for gross income above 250,000 DKK. Average tax rates vary from 7 per cent at the bottom, to almost 60 per cent for the highest 1 per cent.
The Welfare State and Economic Incentives
303
From an incentive point of view it is furthermore the combined effect of marginal taxes and reduced benefits owing to means-testing that matters. A calculation of such combined effective marginal tax rates is shown in Table 10.2. Table 10.2 Share with a combined marginal rate in the range
Combined effective marginal tax rates, 1996 Among pensioners, per cent
Among non-pensioners, per cent
–50 50–60 60–70 70–80 80–
44.2 31.2 19.0 4.6 1.0
30.2 45.5 22.6 1.3 0.3
Total
100.0
100.0
Share of tax payers
0.21
0.79
Source: Ministry of Economic Affairs, Economic Survey (May 1996).
Approximately a quarter of the population has effective combined marginal tax rates higher than 60 per cent. Since means-tested benefits are more important among old-age and disability pensioners the frequency of very high combined marginal tax rates (above 70 per cent) is somewhat higher for pensioners despite their lower average incomes. Among non-pensioners lone parents especially may be exposed to high combined effective marginal tax rates.
10.3
THE SIZE OF THE WELFARE STATE
As in other Northern European countries public expenditures have been rapidly increasing during most of the postwar period (see Table 10.3). In the 1960s and 1970s expenditures on care and health grew particularly fast. However, during the last 15 years social security spending has shown a much faster growth than other elements of the welfare state. To some extent this may reflect a weakness of political control of social security expenditures in a system based on individual rights and entitlements. Right to education, care or hospital treatment is not in the same sense an automatic instant individual right. So political attempts to control government expenditures have focused more on the latter than on social security, causing queues and
Søndergaard
304 Table 10.3
Welfare state expenditures, 1960, 1980 and 1995
Per cent of GDP
1960
1980
1995
Transfers to households – pensions Public consumption – education – care for children and elderly – health
6.4 3.7 13.5 2.9 0.3 1.9
16.6 7.9 26.7 7.8 5.0 5.8
21.9 9.8 25.3 7.2 6.1 5.9
Total
19.9
43.3
47.2
Source: Statistics Denmark, 50-year review (1995); Statistics Denmark, 10-year review (various years); Statistics Denmark, The Statistical Yearbook (1961 and 1962).
unfulfilled demands for hospitals, higher education and child care. It is only recently that this unbalanced development of the two parts of the welfare state has entered policy debates. Since 1960 the number of adults dependent on social security at any given instant – temporarily or permanently – has increased from some 800,000 persons (31 per cent of the adult population) to approximately 1.7 million (42 per cent of the adult population). Over the single year more than half of the adult population collects income-replacing benefits. Part of the increase in provision of public support is due to ageing of the population, but the most significant change has been the increase in the share of the working-age population depending on social security. The latter is caused by at least four changes: a rising level of unemployment, a changed retirement pattern, a liberalization of the criteria of eligibility for several types of benefits and the introduction of new types of benefits with mild eligibility conditions. The consequence has been a similar decline in the share of the working-age population depending on the family or on unearned income (private support in Figure 10.1). As is clearly seen in Figure 10.2 the upward trend in dependence on social security is common to almost all kinds of transfers. The early retirement scheme (post-employment wage) was introduced in 1979 as a measure to redistribute employment voluntarily in favour of the young. The switch towards active measures as part of the unemployment policy began in the early 1980s and has been gradually intensified since then. Dependence on income transfers varies significantly with educational background and age. Among the unskilled between 18 and 66 years of age
The Welfare State and Economic Incentives Figure 10.1 1960–1996
305
Dependence on public support among the 18–66-year-olds,
Source: Statistics Denmark (various publications). Note: Based on annual data from 1990. For the period 1960–1990 based on data for 1960, 1970, 1980 and 1990.
44 per cent collect income-replacing benefits. This compares with only 16 per cent among those with further and higher education. Two-thirds of total benefits are received by the unskilled who count for only 46 per cent of the working-age population. A strong age-dependence in the occurrence of disability (in particular among unskilled) and the access to voluntary early retirement are the two main causes of a higher dependency ratio among the 50–66-year-olds. The rapid increase in dependence on social security and the fact that very few are dependent on private support has been seen by some as an argument in favour of moving from the present system of conditional income replacement to a system of citizens’ basic income ( an unconditional benefit to all citizens possibly subject to a means test). If such a system were operated at current benefit levels the costs and the effective marginal tax rates would become excessive. As will be seen in section 10.4, Denmark has gained some experience when it comes to participation effects of benefits that are not subject to conditions of involuntary loss of earned income.
Søndergaard
306 Figure 10.2 1960–96
18–66-year-olds dependent on different types of social security,
Note: See Figure 10.1. Source: See Figure 10.1.
10.4
INCENTIVES AND THE LABOUR MARKET
The welfare state may affect individual behaviour in the labour market in several ways: participation decision, the desired number of hours, work efforts and human capital investments and the intensity and range of job searches. Furthermore, it may influence firm behaviour and the collective bargaining behaviour of trade unions and employers’ organizations. This makes any attempt to assess the overall impact of welfare state incentives rather complicated.
Labour Supply Econometrics A recent survey of empirical labour supply functions for Denmark is Pedersen (1993), who also discusses some shortcomings of the methods, specifications and data. Only a few studies are available and they reach fairly similar results – numerically very small net wage and income elasticities implying rather small behavioural and welfare effects from taxation.
The Welfare State and Economic Incentives
307
Pedersen and Smith (1995a) have published a new analysis where they use combined survey and register data. The results are basically in line with previous work in that wage elasticities are in the range of –0.04 to –0.12 and income elasticities are zero. The only exception to this general picture is another study by Graversen (1996) who uses a double-hurdle model to take account of both involuntary unemployment and fixed costs of work. Using this approach with register data he finds rather large income and substitution elasticities, while uncompensated wage elasticities are more in line with previous estimates. In addition he has calculated the full distribution of elasticities, showing a large dispersion in the estimated elasticities across individuals. He also simulates the effects of tax reforms, in contrast to the results of similar calculations in other models he finds large positive labour supply effects from tax reforms that reduce progression. One surprising result from the econometric analyses is the lack of any impact on labour supply from children. Apart from a small positive effect on married men and a small negative effect on married women in Graversen (1996) all other studies have found no impact at all, something that is usually explained by the access to subsidized child care as part of the welfare system, but this has not yet been formally tested. It might also be due to difficulties in modelling joint household decisions. Labour supply effects from taxation are only telling part of the story we are looking for. The expenditure side of the welfare state is equally important. Suppose that tax revenue from a general wage tax, t, is spent on income transfers, I, on subsidies to reduce the price of certain services like day care for children, D, from p to p(1 – s) and on care free of charge but subject to individual rationing, G. If the price of consumer goods is set to 1 and the gross wage rate is w, which we take as exogenous, the labour supply function is given as: H = H (1, p(1 − s), w(1 − t), G, I )
(10.1)
and the balanced budget impact on labour supply from an expansion of the welfare state becomes: dH = dG∂H/ ∂G − pds ∂H/ ∂p(1 − s) − wdt∂H/ ∂w(1 − t) + dI∂H / ∂I
(10.2)
where dG + Dpds + dI = wHdt + twdH from the public budget constraint. If either d or G is a substitute for family housework, one of the first two terms is positive, but no econometric studies for Denmark have included such variables. The third term is theoretically indeterminate, but all Danish econometric studies except one have found very small net wage effects. The last term is negative under the usual assumption that leisure is a normal good. However, all econometric studies except one have found very small income elasticities. Thus whether a tax-financed expansion of the welfare state will
308
Søndergaard
reduce or improve incentives to work depends on the composition of welfare expenditures and on the magnitudes of elasticities of labour supply with respect to both taxes and expenditures. When collective bargaining is as comprehensive as in Denmark, econometric studies of individual labour supply may be biased. Since collective bargaining usually also covers the number of weekly working hours some individuals may be unable to choose their preferred amount of work. It is well known that weekly working hours are clustered around the number of hours of a normal work week in collective wage agreements. Under these circumstances an aggregate time-series approach may capture the possible effects of the welfare system on the supply of hours better than a cross-section of individual panel data. An alternative approach has been followed by Pedersen and Smith (1995a). Their survey data contain responses to a hypothetical question about how the respondent would like to change weekly or annual working hours in case of a change in tax rates. The results are consistent with the very low average elasticities found in individual supply studies, but the dispersion of the responses suggests that the small average elasticity may be the net result of larger but opposite reactions from different groups in the population, similar to what Graversen found using register data. As shown in Figure 10.3, average annual working hours in manufacturing have declined from some 2100 hours in 1960 to around 1600 hours in the early 1990s. This is largely but not exclusively due to negotiated reductions of the normal working week and prolonged holidays. Yet during the same period part-time work first increased from 10 per cent of the labour force in 1960 to 25 per cent around 1980, followed by a decline to around 18 per cent in the early 1990s. If this development in working time were explained by wage and income effects in standard economic labour supply models uncompensated wage elasticities should be negative. However, the decline in individual working time may also be related to the changing family structure during the post-war period. If the transition from one-breadwinner households to two breadwinners has reduced the preferred amount of work per individual, the decline in average working hours over time can be consistent even with positive wage elasticities. This is in complete accordance with the general observation that evidence on Danish labour supply elasticities is scattered and hardly lending support to conclusions beyond the predominance of numerically very small elasticities. Participation Incentive effects on participation may be more important than incentive effects on hours, primarily because the welfare state interferes rather directly with the allocation of time between the family and the market. From 1950 to 1970 the participation rate in Denmark was almost constant at 70 per cent of
The Welfare State and Economic Incentives
309
Figure 10.3 Annual working hours (manufacturing) and part-time frequency (all sectors) 1948–94
Source: Statistics Denmark, 50-year review 1995.
the 15–69-year-olds. After 1970 the participation rate has been steadily increasing from 70 to 76 per cent. However, two important changes in the composition have occurred during the same period. The gender difference is gradually closing as male participation shows a slight decrease while female participation is increasing. Furthermore, participation has become more concentrated around the middle of the life-cycle. The possible impact of the welfare state on participation is complex, however. Let us assume that participation decisions over the life-time are divided into three types of decisions: if and when to enter, when to retire, and when and for how long to leave the labour force temporarily. The Danish welfare state has two opposite incentive effects on decisions to enter. Subsidized care, maternity leave, income security in case of sickness and unemployment and the possibility of temporary paid leave give incentives to accelerate participation, while free education and educational grants have the opposite effect. The effect of taxation on participation is indeterminate both in theory and empirically, while non-work-related benefits like housing subsidies and family allowances should be expected to decrease participation.
Søndergaard
310
Retirement in Denmark is probably accelerated by the welfare state through provision of fairly generous opportunities for voluntary retirement from the age of 60 while retirement at a younger age requires qualification for a disability pension. The latter is mainly a question of health conditions for persons below the age of 60 while the conditions are broader and weaker for the 60–66-year-olds. A recent Danish study indicates that the correlation between health problems/disability and disability pension is far from perfect. Bengtson (1997) has conducted a survey where individuals were asked a large number of questions concerning their abilities and chronic deceases. The data indicates that either moral hazard or imperfect targeting occurs, since the joint incidence of reported disability and receipt of disability pension in the sample accounted for less than two-thirds of the actual number of disability pensioners. However, some reservations are needed because of the possible errors in reporting in a survey like this. Since rights to early retirement depend on previous participation, this may have a – probably rather small – positive effect on participation in the form of reduced temporary leave. Sick pay and unemployment insurance reduce incentives to leave the labour force temporarily because the benefits are conditioned on participation and because to some extent moral hazard may be a substitute for temporary leaves. From a policy perspective three issues related to participation are of particular interest. One is the incidence of moral hazard and cheating related to social security.3 The more widespread such behaviour is, the larger the
Table 10.4
Expected welfare state incentive effects on labour force participation
Participation effect from:
Entering
Temporary leaves
Retirement
Old-age pension Early retirement Disability pension Sick pay and unemployment benefits Maternity leave Leave schemes Subsidized care Education Housing benefits and child allowances Taxation
No No No Positive Positive Positive Positive Negative Negative ?
No Positive No Positive ? Negative Positive Positive Negative ?
Negative Negative Negative Positive No No (Positive) No Positive Negative ?
The Welfare State and Economic Incentives
311
measured labour force will be relative to actual labour supply. Lindbeck (1995) conjectures that moral hazard and cheating may develop asymmetrically over the business cycle – that is, that a generous welfare system owing to inherited social norms and habits may generate only moderate moral hazard and cheating problems as long as the macroeconomy is in good shape, while such problems may increase irreversibly if the economy is hit by a major macroeconomic shock. The gradual increase for more than three decades in dependence on public support displayed in Figure 10.2 does not necessarily contradict the shock hypothesis. The major set backs in employment in Denmark occurred in 1967–8, 1974–5, 1980–1 and 1988–92. Yet, an alternative hypothesis is that a political cycle is in operation, that is that the social security system is expanded during recessions both in the form of new types of benefits and in the form of weaker eligibility conditions. If moral hazard and cheating have become more widespread over time – and owing to lack of evidence that cannot be completely ruled out, of course – it may have been supported by changing social norms in the population at large which have been transformed through the policy process into weaker conditions of eligibility of public support rather than just a plunge in norms and habits of those living on benefits. This has certainly been of some importance in Denmark, as illustrated by the introduction of a voluntary, early retirement scheme in 1979. A main argument behind this scheme was to redistribute employment in a period of high unemployment in favour of the young. But a number of changes in eligibility criteria have also occurred, a notable example being the inclusion of independent business people, farmers and fishermen in the UI system. This leads us to the second issue that has attracted much interest in current political controversies – the decline in the average retirement age. In Denmark the old-age pension system has 67 years as pension age, but less than 20 per cent of the population are active up to the pension age and the actual average age of retirement is just above 61. As is seen from Figure 10.4 the change in retirement behaviour has been significant since 1970. In 1970 86 per cent of males active at the age of 50 and 47 per cent of females active at the age of 50 were still active at the age of 62. Today the corresponding figures are 47 and 29 per cent. In 1990 only 32 per cent of the population aged 60–66 participated in the labour force. 29 per cent received a disability pension, 28 per cent received a voluntary early retirement benefit (post-employment wage) while 11 per cent were living on private support including private pensions.4 Eligibility for a disability pension is different for persons aged 60–66 than for other age groups. Essentially, a person qualifies for a (small) pension if the probability of finding a job is considered very low and the total income of the family does
312 Figure 10.4
Søndergaard Labour market survival rates for active at age 50, 1970 and 1994
Source: Statistics Denmark, Statistisk Tabelværk 1974:VII, Statistiske Efterretninger 1995:13.
not exceed the level of voluntary, early retirement benefits. In particular many women have qualified for this type of disability pension. As mentioned already an enlargement of the voluntary early retirement scheme was in operation during a short period allowing unemployed aged 50–59 to retire. As seen from Figure 10.5 during eight quarters from 1994:Q1 to 1996:Q1, 8 per cent of the labour force in this age group took advantage of this offer and decided to retire. The scheme has been closed for new entrants since 1996:Q1. The decision to close was no doubt very much influenced by the heavy inflow to the scheme. Yet, there are still strong political controversies as to the importance and means of redressing the declining average age of retirement. Particularly, a strong focus has been on the voluntary early retirement scheme, but as mentioned this scheme accounts for less than half of total retirement in the age group 60–66. Apart from the strain on public finances the main issue in the controversy is the coming ageing of the population. Over the next three decades a 35 per cent increase in the 60+ age group is expected which, assuming unchanged retire-
The Welfare State and Economic Incentives
313
Figure 10.5 Retirement on early retirement scheme for unemployed aged 50–59, 1994:Q1–1996:Q4
Source: Statistics Denmark, Statistiske Efterretninger: Arbejdsmarked.
ment behaviour and unchanged average life-times, will increase the number of elderly on public support by some 300,000 persons. It attracted much attention from abroad when Denmark introduced its voluntary leave schemes in 1994. Three schemes were introduced: educational leave, parental leave for parents with children below the age of 9 and a sabbatical leave scheme, where an unemployed person must be employed as a substitute for the person on leave. The schemes that offered benefits at the same level as unemployment benefits soon became quite popular and some 3 per cent of the total labour force were on leave at the beginning of 1995 when participation in the leave schemes culminated (see Figure 10.6). The popularity was largely unexpected. A gradual cut back in benefits to people on parental and sabbatical leave ( but not educational leave) was decided in 1995 and from 1997 the benefits on parental and sabbatical leave was only 60 per cent of unemployment benefits. More than half of the persons on leave came from unemployment. For educational leave this was mainly due to active labour market policies that used educational leaves as a measure to improve qualifications among the unemployed. The use of parental leave among unemployed could be seen, on the other hand, as an indication of some degree of voluntary unemployment.
Søndergaard
314 Figure 10.6
Participation in leave schemes, 1994:Q1–1996:Q4
Source: Statistics Denmark, ABBA-database, AB711.
Human Capital Investments The growth rate of the stock of human capital in Denmark has been estimated at around 0.3 per cent p.a. during the last 50 years (see The Danish Economic Council, 1995). The measure used in that assessment is based on number of years of formal education of the labour force. Measuring the stock of human capital is not easy, however, and if relative wages are used as weights in aggregating years of education the growth rate is much smaller because of the compressed wage structure in Denmark. Yet, a growth contribution of some 0.3 per cent is consistent with similar estimates in growth accounting (see Maddison, 1991). To what extent has the development of human capital been influenced by the welfare state? No conclusive answer is available, but we can identify at least four different ways in which the welfare state may influence individual decisions on how much to invest in human capital. First, educational grants and state provision will increase participation in education and the length of study. This will in turn increase the market value of human capital provided people choose types of education that enhance productive qualifications that are of value in the market-place (education as investment) rather than education aiming at other types of competence (education as consumption). Educational grants and public provision carry a deadweight loss. The effect on human capital relative to the deadweight loss
The Welfare State and Economic Incentives
315
depends on how sensitive educational choice is to economic incentives – something that has not been carefully examined. Secondly, the welfare state may increase enrolment in education by alleviating poverty and improving living conditions among disadvantaged groups, since many studies have pointed to a significant influence from social and economic conditions of parents on the amount of education invested in the children. A less unequal distribution of income would therefore increase the average amount of education.5 However, a reduction of inequality would also have a negative incentive effect by reducing private returns on human capital investments. The relative importance of these two effects of reduced inequality is unknown. Nevertheless, one may conjecture that the relationship between equalization and incentives is non-monotone – that is, that the overall incentive effect will be positive up to some degree of equalization, but as private returns to human capital investments continue to decline as inequality decreases eventually the incentive effect from lower investment returns will come to dominate. The Danish Economic Council (1995) found private returns to be negative or close to zero for several types of education unless other factors such as lower risks of unemployment and disability when completing an education are included in the calculation of returns. Thirdly, a comprehensive safety net and redistribution of income will provide some insurance against the income uncertainty associated with individual investments in human capital. This risk is otherwise neither diversifiable nor insurable through market transactions. Therefore, the welfare state may enhance human capital investment. Eaton and Rosen (1980) show that income taxes may be superior to lump-sum taxes for this reason. Nielsen and Sørensen (1996) provide a different argument for progressive taxation of labour income when capital income is subject to a proportional tax. With no uncertainty a proportional income tax gives too strong incentives to human capital investment. Fourthly, it is an empirical fact that administration of welfare systems and public provision use more highly educated labour – in particular, labour with a university degree – than comparable private services. The increase in demand for labour with a higher education has not shown up in relative wages. On the contrary, relative wages of those with a university degree have been declining during the last 25 years, which suggests that the demand effect of the welfare state has been weaker than the supply effect.
Work Effort and Intensity of Search While participation and human capital may be seen as the nominal labour supply, effective labour supply depends on whether work efforts match
316
Søndergaard
individual capabilities and on how intensively individuals are searching for jobs. Disincentive effects from the welfare state may make effective labour supply smaller than nominal supply because the loss of net income from being laid off is reduced. It is a common feature in many different models of labour market behaviour that work efforts and intensity of search decline correspondingly with the level of the net replacement rate and the duration of benefits. Several reports have calculated net replacement rates for Denmark and much policy debate has been triggered off by the fact that net replacement rates sometimes exceed 100 per cent so that people suffer a net income loss from working. In their study Pedersen and Smith (1995a) found that some 5 per cent of all employed wage earners in Denmark would have an immediate economic gain from becoming unemployed when transportation costs, day care expenditures and taxation were taken into account. An additional 22 per cent would only suffer a minor loss of disposable income (less than 500 DKK per month) if they became unemployed. The fact that both groups were nevertheless working may be less of a paradox than many policy makers and observers seem to believe. For one thing this benchmarking at 100 per cent instant replacement rate may overlook the possibility of a decrease in life-time income if a person becomes long-term unemployed, since future wages are likely to be a declining function of individual duration of unemployment. Secondly – and more important – not everybody is suffering from disutility of work. A more plausible description of the attitudes of the labour force is to assume that utility or disutility from work has an individual component as well as a job-specific component. To keep the argument as simple as possible let the job-specific reservation wage of individual i for job j be wrij and the jobspecific individual utility from work (positive or negative) be uij. Let net income when unemployed, ri, be r i = b i − T (b i + I i )
(10.3)
where total direct taxes, T, vary with total income including unearned income, I, and b is gross benefit when not working. In the Danish system b is 90 per cent of previous gross wages up to a maximum of approximately 12,000 DKK per month. However, for very low wages (less than 133,000 DKK in annual income when full-time employed) b may exceed 90 per cent owing to a fixed floor in the system. Let cij denote fixed costs when working at job j, then the reservation wage of individual I for job j becomes: wrij = r i + c ij − u ij
(10.4)
The Welfare State and Economic Incentives
317
Except in a very few cases (for example, low-paid lone parents) r cannot exceed 100 per cent of previous net income. Effective marginal tax rates (including effects of means-tested housing benefits and day care payment) may rarely be close to 100 per cent (see Table 10.2), so that r is also close to 100 per cent for low-paid workers with a gross replacement rate of 90 per cent. Some unemployed on social assistance may actually get a net compensation of more than 100 per cent, depending on their housing costs and the number of children. Nevertheless, owing to fixed costs of work one should not be surprised to find results like those cited above that work is associated with a direct income loss in many circumstances. However, if u is positive, high replacement rates need not be a serious problem. It is the sign (and distribution) of u that is the core determinant of the disincentive effects from a generous social security system. The true incentive problem arises whenever u is negative and/or c is large so that the reservation wage of person i for job j exceeds the actual wage in job j – that is, wrij > wj. In such circumstances the individual has an incentive to cheat or display moral hazard. However, unless disutility or fixed costs from work are high for all available jobs this does not lead to a complete stop of search activities. Rather, the range of jobs being of potential interest to the worker is reduced, implying that job search is confined to the more pleasant, better-paid jobs closer to where the individual is living. Some information on job search of the unemployed is available based on their own reporting in surveys. Bach (1984) using survey data for 1979 found that 23 per cent of the unemployed (out of work for more than a month) living on unemployment benefits and 35 per cent of those living on social assistance did not fulfil the ILO criteria for being unemployed. Pedersen and Smith (1995b) found similar results for 1993–4 using a small sample of unemployed where 40 per cent did not fulfil the ILO criteria. Using a somewhat larger sample of unemployed with more than 3 months of unemployment and no job training or other active measures the Ministry of Finance (1995) reported that 70 per cent did not fulfil the ILO criteria. The main results from the latter study are shown in Table 10.5. One should bear in mind that unemployed with such long spells account for only around half of the total number of unemployed at any given point of time, so that the result is largely consistent with that found by Pedersen and Smith. To what extent this picture stems from a change in norms and habits (the Lindbeck conjecture) or rather destroyed incentives including weak conditions for obtaining and maintaining eligibility is an open question. Pedersen and Smith (1995b) also found that economic incentives (and age) were important determinants of search activity and willingness to accept commuting. All in all the available evidence seems to support the view that disincentive effects show up as reduced search intensity and in a more narrow range (geographically and otherwise) of jobs being of interest.
Søndergaard
318 Table 10.5
Reported attitudes and search activity among unemployed with at least 3 months of unemployment, 1994
Per cent A: B: C: D:
Want a job Are actively searching and ready to start within 2 weeks Willing to accept up to 1 hour total commuting time per day All the above
Yes
No
81 60 40 30
19 40 60 70
Source: Ministry of Finance (1995), based on a survey by the Danish National Institute of Social Research.
Job search may be also influenced by activation policies that have become increasingly important since the late 1980s. The expected effect from compulsory participation in, for example, job training or public work will, however, depend on the attitudes of the individual. If the individual reservation wage for the job training is lower than the pay (usually equal to unemployment benefits), activation may have a negative search effect. When the reservation wage is higher than the pay a positive search effect is possible, although other reactions may be possible as well like sick declaration or choice of educational activity. The impact of increased use of compulsory activation has not yet been fully investigated. The same holds for the effects from the recent cut in duration of benefits to young unskilled. There are some indications of a positive effect of the combined cut in duration and offer of a special education and training programme both in the search for jobs and on enrolment in education. Unemployment among 18–24-year-olds was just above 5 per cent (seasonally adjusted) in the first quarter of 1997.
Incentives to Firms The potential impact of the welfare state on firm behaviour has not attracted much interest so far, one exception being the incentive effects on temporary lay offs. The Danish unemployment insurance system allows temporary lay offs, but recently this disincentive has been slightly modified by the introduction of a 2-day waiting period for unemployment benefits to be fully compensated by the employer when a worker is laid off. Temporary lay offs account for as much as 1 percentage point of total unemployment and a much larger fraction of short spells of unemployment.6
The Welfare State and Economic Incentives
319
Incentives and Wage Bargaining In bargaining models as well as monopoly union right-to-manage and efficiency wage models wages are positively related to replacement rates, while tax progressivity has a negative wage impact. The impact of the level of taxation and other kinds of welfare expenditures are in general indeterminate. Atkinson and Micklewright (1991) in a survey of the empirical literature draw the conclusion that the disincentive effects from the benefit system are generally uncertain partly because the models of the systems are often incomprehensive. Pedersen and WestergårdNielsen (1993), in a survey of panel data analyses, found that surprisingly one gets larger effects on American data than using data from European countries. In Denmark, unemployment benefits were increased in the late 1960s when the average replacement rate increased from 45 per cent to around 75 per cent, but since then a slight decline has taken place to around 65 per cent (see Pedersen, 1993). Estimates of aggregate wage functions for Denmark have confirmed the expected positive effect from the replacement rate. The Secretariat of the Danish Economic Council (1994) estimated a long-run elasticity of the product wage with respect to the average replacement rate (calculated as the ratio of average benefits to average wages, hence ignoring the fact that a majority of the claimants of benefits have belowaverage wages) of approximately 0.4. Most econometric studies apart from a paper by Hansen et al. (1996) have not found statistically significant tax effects. Hansen et al. used a somewhat unconventional model of combined bargaining and neoclassical assumptions. They found significant negative effects from the degree of tax progression on wages of unskilled, but positive wage effects for high-income earners. The structure of replacement rates is likely also to influence the wage distribution. The Danish system has replacement rates of 90 per cent at the bottom of the wage distribution, declining to around 50 per cent at the average wage. This should produce larger effects on reservation wages among the low paid than among the better paid, and so contribute to a more compressed wage structure. Rowthorn (1992) in a comparison of wage dispersion across OECD countries concludes that Denmark has one of the most compressed wage distributions. In a neoclassical model unemployment caused by high replacement rates would be voluntary. Collective wage bargaining, however, may produce a different outcome because collective wage contracts, at least in Denmark, usually set minimum wages. Therefore an individual worker with a low reservation wage (positive utility/ low fixed costs from work) may become
320
Søndergaard
rationed if the minimum wage is set higher than the value of his or her productivity. Jensen et al. (1994) have estimated that some 0.7 per cent of the male labour force and 1.7 per cent of the female labour force may be unemployed due to this minimum wage effect.
Overall Labour Market Impact The probable impact from the welfare system on the functioning of the labour market may be summarized as follows. First, a generous social security system is likely to exert an upward pressure on the wage level and so increase unemployment. Progressive taxation may modify this effect if collective wage bargaining is important. Secondly, high replacement rates and long durations are likely to make wages less sensitive to demand. Macroeconomic shocks should therefore be expected to produce larger shifts in employment. Yet, this may be modified by larger automatic stabilizers due to high tax and benefit rates. Thirdly, the wage distribution is likely to be more equal due to high replacement rates among low-paid workers. The latter should be expected to produce an unequal distribution of unemployment and unemployment risks, something that has been demonstrated in many reports. This shows up also as very different durations of unemployment – that is, not only is the risk of becoming unemployed unequally distributed, but so is the expected duration. Therefore a fraction of the labour force has become ‘marginalized’ in the Danish labour market, in the sense that their spells of employment are very few and very short. This indicates that they either do not want to work or that the value of their productivity is below the ruling minimum wage in their part of the labour market. A report by Ingerslev and Pedersen (1996) found that as many as 6 per cent of the labour force were in a marginal position in 1994 (more than 20 months of unemployment within the last 3 years). The appearance of a large marginal group may indicate that unemployment has a self-increasing impact. The probability of getting a job decreases with the duration of unemployment and long-term unemployed lose their competitiveness in the labour market. In part this may be due to a ‘lemon effect’, if firms with vacant jobs due to screening costs and so on recruit somebody already employed elsewhere rather than a long-term unemployed. A combination of high replacement rates and a small wage dispersion weakens the incentives to enrol in training and education and to job mobility. Another effect of the welfare state that has not been fully analyzed so far is the consequences of the attempts to modify economic disincentives from high replacement rates through tighter control of availability of the unemployed and more extensive use of active measures as a condition for maintaining eligibility.
The Welfare State and Economic Incentives 10.5
321
SUMMARY AND DISCUSSION
We have found some indications that the welfare state may distort labour market incentives in Denmark in a way that contributes to high and persistent unemployment. In comparisons with the United States it is often said that the negative side of the much more vital US job growth is a large number of working poor and increasing inequality. Conversely, a positive side of the Danish combination of high minimum wages and high replacement rates for the low paid is that the distribution of income has not become more unequal during the long period of high unemployment (see Pedersen and Smith, 1995c). A very controversial issue is whether the welfare state has had a measurable impact on output and productivity. Figure 10.7 displays log GDP per capita in Denmark for the last 110 years and a trend with an average growth rate of 1.9 per cent per year.
Figure 10.7
Index (1900 = 0) of log GDP per capita, 1884–1994
Source: Pedersen and Søndergaard (1996).
322
Søndergaard
One striking observation is the impact of two world wars. Another is the difficulty of tracing any impact from the development of the welfare state. Post-war growth has been slightly above the trend, even when corrections for the loss of output during the war are taken into account. This is consistent with the findings of Andersen and Schmidt in Chapter 9 in this volume and Agell et al. (1997) who were not able to establish a firm link between welfare state variables and macroeconomic indicators. Many difficulties are inherent in empirical studies of growth and welfare state variables (see Atkinson, 1995). In particular, we are faced with a serious identification problem, because welfare expansions could be determinants of as well as a politically decided reaction to growing real incomes. It may also be that the loss of output due to higher unemployment is partly or fully offset by higher participation and higher average productivity. The latter may be brought about both directly by more investment in human capital and indirectly by changes in relative factor prices due to welfare state incentive effects. Until recently the level of household savings has been very low in Denmark. This may be the result of developing a welfare state that obviously reduces the need for private savings both for life-cycle consumption-smoothing and precautionary purposes. The magnitude of such disincentive effects is difficult to assess. Furthermore, at least in principle, a decline in household savings can be offset by additional savings in other sectors. Therefore it is largely a political choice whether the welfare state is allowed to reduce future incomes through a slower accumulation of wealth. Notes 1.
2.
3.
4.
Among other things this has been the subject of three government commissions in the 1990s – The Social Commission (1991–3), The Committee on Structural Problems in the Labour Market (1992) and the Commission on Future Employment and Business Opportunities (1994–5). The tax reform in 1994 introduced a flat rate tax on earned income. The revenue from this finances three ‘funds’ – an unemployment benefit fund, an activation fund and a sick pay fund. However, so far it is not clear that this will become a truly earmarked tax rather than just a formality. The opposite holds for labour supply to the shadow economy where income is not subject to taxation or means-testing for social security. Mogensen et al. (1995) provides estimates of the shadow economy in Denmark since the early 1980s. International comparisons do not suggest a simple relationship between taxes and welfare systems on the one hand and the magnitude of the shadow economy on the other. Pedersen and Smith (1995a) present an econometric test of the impact of individual tax rates in their estimation of a labour supply function for the shadow economy. They found no significant effect of tax rates and besides reported work in the shadow economy was positively – not negatively – correlated with working hours in the white economy. See Social Commission (1993a).
The Welfare State and Economic Incentives 5. 6.
323
Such intergenerational correlations have been demonstrated in several Danish investigations (see, for example, The Social Commission, 1992). The estimate was provided by the Committee on Structural Problems on the Labour Market (1992).
References Agell, J., T. Lindh and H. Ohlsson (1997) ‘Growth and the public sector: a critical review essay’, European Journal of Political Economy, 13, pp. 33–52. Atkinson, A.B. (1995) ‘The welfare state and economic performance’, National Tax Journal, 48, pp. 171–98. Atkinson, A.B. and J. Micklewright (1991) ‘Unemployment compensation and labour market transitions: a critical review’, Journal of Economic Literature, 29, pp. 1679–727. Bach, H. (1994) Job Search Behaviour of the Unemployed (Copenhagen: The Danish National Institute of Social Research) (in Danish). Bengtson, S. (1997) Handicap and Disability in the 1990s (Copenhagen: The Danish National Institute of Social Research) (in Danish). Blomquist, S. and V. Christiansen (1995) ‘Public provision of private goods as a redistributive device in an optimum income tax model’, Scandinavian Journal of Economics, 97, pp. 547–68. Committee on Structural Problems in the Labour Market (1992) Report from the Committee on Structural Problems in the Labour Market (Copenhagen: Udredningsudvalget, sekretariatet) (in Danish). Danish Economic Council (1995) Danish Economy Spring 1995 (Copenhagen: Det økonomiske Råds Formandskab) (in Danish). Eaton, J. and H.S. Rosen (1980) ‘Taxation, human capital, and uncertainty’, American Economic Review, pp. 705–15. Graversen, E.K. (1996) ‘Male and female labour supply in Denmark’, Working Paper, 96-15, Center for Labour Market and Social Research, Aarhus. Hansen, C.T., L.H. Petersen and T. Sløk (1996) ‘Danish results on the relationship between marginal tax rates and wages’, Nationaløkonomisk Tidsskrift, 134, pp. 153–74 (in Danish). Hansen, H. et al. (1995) Unemployment Benefits and Social Assistance in Seven European Countries (Copenhagen: Økonomiministeriet). Hoff, K. and A.B. Lyon (1995) ‘Non-leaky buckets: optimal redistributive taxation and agency costs’, Journal of Public Economics, 58, pp. 365–90. Ingerslev, O. and L. Pedersen (1996) Marginalization 1990–1994 (Copenhagen: The Danish National Institute of Social Research) (in Danish). Jensen, P. et al. (1994) ‘Unemployment and minimum wages – a microeconometric analysis, Working Paper 1994-08, Center for Labour Market and Social Research, Aarhus Business School and University of Aarhus. Lassen, D.D. and S.B. Nielsen (1996) ‘Is the tax burden higher in Denmark than in other European countries?’, Nationaløkonomisk Tidsskrift, 134, pp. 209–22 (in Danish). Lindbeck, A. (1995) ‘Welfare state disincentives with endogenous habits and norms’, Scandinavian Journal of Economics, 97, pp. 477–94. Maddison, A. (1991) Dynamic Forces in Capitalist Development (New York: Oxford University Press). Ministry of Finance (1995) Budget analysis 95 (Copenhagen: Finansministeriet) (in Danish).
324
Søndergaard
Mogensen, G.V. et al. (1995) The Shadow Economy in Denmark 1994. Measurement and Results (Copenhagen: Danmarks Statistik). Nielsen, S.B. and P.B. Sørensen (1997) ‘On the optimality of the Nordic system of dual income taxation’, Journal of Public Economics, 63, pp. 311–29. Okun, A.M. (1975) Equality and Efficiency: The Big Trade-off (Washington, DC: The Brookings Institution). Pedersen, P.J. (1993) ‘The welfare state and taxation in Denmark’, Chapter 7 in A.B. Atkinson and G.V. Mogensen (eds), Welfare and Work Incentives: A North European perspective (Oxford: Clarendon Press). Pedersen, P.J. and N. Smith (1995a) ‘Taxed and non-taxed labour supply: wages, taxes and costs of work’, Chapter 3 in G.V. Mogensen (ed.), Work Incentives in the Danish Welfare State (Aarhus: Aarhus Universitetsforlag). Pedersen, P.J. and N. Smith (1995b) ‘Unemployment and Incentives’, Chapter 7 in G.V. Mogensen (ed.), Work Incentives in the Danish Welfare State (Aarhus: Aarhus Universitetsforlag). Pedersen, P.J. and N. Smith (1995c) ‘Trends in the Danish income distribution, 1976–90’, Centre for Labour Market and Social Research, Aarhus, mimeo. Pedersen, P.J. and J. Søndergaard (1996) ‘Welfare state and growth’, Economic Essays in honour of Anders Ølgaard, Nationaløkonomisk Tidsskrift, pp. 317–28 (in Danish). Pedersen, P.J. and N. Westergård-Nielsen (1993) ‘Unemployment: a review of the evidence from panel data’, OECD Economic Studies, 20, pp. 65–114. Ploug, N. and J. Kvist (1997) ‘Transfer systems in Europe’, Social Security in Europe 3 (Copenhagen: The Danish National Institute of Social Research) (in Danish). Rowthorn, R. (1992) ‘Centralization, employment and wage dispersion’, The Economic Journal, 102, pp. 506–23. Sandmo, A. (1995) ‘Introduction: The welfare economics of the welfare state’, Scandinavian Journal of Economics, 97, pp. 469–76. Secretariat of the Danish Economic Council (1994) SMEC Documentation of Model and Simulated Effects of Economic Policy (Copenhagen: Det økonomiske Råds sekretariat (in Danish). Sinn, H.W. (1995) ‘A theory of the welfare state’, Scandinavian Journal of Economics, 97, pp. 495–526. Smith, N. (1995) ‘A panel study of labour supply and taxes in Denmark’, Applied Economics, 27, pp. 419–29. Social Commission (1992) The Young Generation in the Welfare Society (Copenhagen: Socialkommissionen) (in Danish). Social Commission (1993a) The Elderly – An Analysis of Provision of Support for the Elderly (Copenhagen: Socialkommissionen) (in Danish). Social Commission (1993b) Out of Work – Public Transfers to the Middle Age Group (Copenhagen: Socialkommissionen) (in Danish).
Comment Torsten Sløk Jørgen Søndergaard has written a thorough and detailed chapter on the experiences of the Danish welfare state. Chapter 10 goes into great detail describing the actual schemes and focuses on where in the Danish welfare state potential incentive problems arise. This is an ambitious task and the chapter balances elegantly between economic analysis and specifying the individual benefits and their eligibility requirements in the Danish welfare state. The chapter also raises the notion that the growth of the welfare state is due to a change in social norms. This change in social norms has been from an earlier view that the benefits of the welfare state are an offer to the current view that the benefits are a claim. I will argue that this change in social norms is the real source of the incentive problems in the welfare state. The existence of incentive problems does not necessarily mean that the size and scope of the welfare state should be reduced dramatically, but just that the benefit mechanisms and eligibility criteria should be better designed and coordinated. This is both to minimize the incentive problems and to control for a burgeoning expenditures. In any listing of national income per capita Denmark is always ranked as one of the richest countries in the world. In the 1996 World Development Report (World Bank Group, 1996) Denmark is listed as the third richest country in the world, surpassed only by Switzerland and Japan (measured by GNP in 1994 USD). Sweden and Norway have for a long time also been and still are ranked highly on the top 10 list of richest countries. The position as one of the richest countries has been maintained over a long period of time and in this period Denmark has, as Søndergaard says, experienced a drastic increase in the size of the welfare state: the welfare state in Denmark has grown at a greater rate than real national income. This can potentially in the long run be a problem, but not if there are other periods where the growth rate of the welfare state is lower than the growth rate of national income. Hence when evaluating the output performance of the welfare state a conclusion could be that the positive externalities connected with having a welfare state (such as the benefits of a more productive and better performing labour force) apparently exceed the negative externalities (such as the incentive problems and financing). However, the design and the functioning of the welfare state can and should continuously be reviewed and that is precisely what Søndergaard has done in Chapter 10. What is the optimal design of the welfare state? What is the optimal number of benefits, and what is the optimal level of individual benefits? These are the essential questions which we would like to answer. When talking about optimizing, we need to establish what the objectives of the welfare state are. The objectives are, (see also Barr, 1993): efficiency, the support of living standards, the reduction of inequality, social integration and administrative feasibility. Chapter 10 addresses the objective of efficiency which can be split up into (1) macro efficiency, (2) micro efficiency and (3) incentives. ‘Macro efficiency’ refers to the optimal fraction of GDP allocated to welfare state institutions. ‘Micro efficiency’ refers to policy designed to ensure an efficient division of total welfare state resources between different cash benefits, different types of medical treatment and different kinds of education. Finally, ‘incentives’ say that where institutions
325
326
Sløk
are publicly funded, their finance and the construction of benefits should minimize adverse effects on (1) labour supply and employment and (2) saving. Søndergaard addresses the first of these two incentive problems of the Danish welfare state. But what is an incentive problem? An incentive problem is essentially about mechanism design and designing institutions and rules so that the group who actually receives the benefit is also the same group for whom the benefit was intended. It is well known that in practice it can be very hard to design social benefit mechanisms that are not subject to abuse. Note further the fact that when eligibility is relaxed for individual schemes it automatically makes more people eligible to receive benefits from the programme and by definition this should make the programme more requested and more popular. It has been a feature of the Danish welfare state over the years that eligibility has been relaxed and thereby more people automatically receive benefits from different schemes. But in what sense is this an incentive problem? The basic problem with creating a social security benefit or a benefit in kind is that it should be used by the people it is intended for. Since a lot of benefits are not means tested this implies that people can get benefits from several schemes even though they already, in some survival sense, receive a sufficient income. This leads us to an important explanation why Denmark has experienced an increase in the size of the welfare state and why the number of incentive problems has grown – namely, that social norms and the attitude of the population (and also politicians) has changed along with a growth in the size of the welfare state. It is an often-voiced opinion from individuals that they are entitled to benefits with reference to the high amount of taxes they pay. This spiral in which the more you feel the tax pressure the more benefits you feel entitled to is a dangerous one for a welfare state to get caught in. The problem is that we are up against the attitudes and moral standards of people in the society, which is something that lies outside the economist’s toolbox. The reason economists care about incentive problems comes from the need to finance the welfare state. Since we have only limited resources available these resources should be used as efficiently as possible. A costly part of the Danish welfare state is the insurance part which deals with entering and leaving the labour market. To be eligible for these schemes one has to fulfil different criteria as described by Søndergaard in Chapter 10 (for example, Table 10.1). In recent years one of the most debated benefits has been the early retirement scheme. It is not because the early retirement scheme is particularly costly relative to other benefits but rather because of the incentive problems inherent in the early retirement scheme. Age is the only criterion for eligibility in the scheme. If all people who turn 60 years of age decide to retire early a massive increase in revenue would be required to finance these benefits. If, however, only a small fraction of the population who turns 60 retires early the revenue required to finance the benefits would be more manageable. But, as noted by Søndergaard, the fundamental problem is that in general the welfare state has experienced weaker conditions of eligibility so that a broader group of people are now included in the group of possible recipients of benefits. In this case what is the source of the incentive problems? Is it the new group of recipients that receives benefits or is it the politicians who have relaxed the eligibility criteria? Again this is a product of social norms. Another participation issue which also has received a lot of attention is the labour supply literature. Both compensated and uncompensated elasticities found for Denmark are pretty small and very close to zero in most studies. It is, however, a common problem in empirical studies of labour supply that there is only very little variation in the variable that we try to explain – namely, hours worked – since all observations are clustered around full-time work. A further problem is that the key
Comment
327
variables that we need to observe to understand how, for example, the tax system influences the number of overtime hours worked and the overtime wage rate are very hard to observe. Only very scarce information is available on overtime work while measures of overtime wage rates are relatively poor. The fact that it is so very difficult to explain hours worked leads Søndergaard to conclude that econometric studies give little knowledge about the causes of the decline in annual working hours. Yes, it is hard to explain the fall of annual working hours with individual (household) variables. The reason is that annual working hours are, for the majority of the labour market, not determined as an individual decision but rather by the institutions of the labour market – namely the labour unions who bargain wages and the weekly number of hours worked. Here again is an example of how important institutions and noneconomic factors can be. Even though in our theoretical models we say that trade unions have the same objectives as the aggregate of the individual members it is not obvious that this is the case when it comes to the actual behaviour of Danish labour unions. An important part of wage negotiations seems to be political factors and when hearing reports about what is actually happening during the negotiations of collective agreements there are many different factors considered by the labour union beyond those of maximizing wages and employment for its members. Hence the fall in annual working hours comes from labour unions believing that lower working hours is what their members want and it is hard to say on which variables this belief is based. The conclusion regarding labour supply seems then to be that in Denmark what determines an individual’s labour supply has only a little to do with economic incentives, or at least it is hard to find evidence of a significant importance of economic incentives. In general, as also discussed in Søndergaard’s chapter, there are serious identification problems involved in finding the incentive problems of social benefits. The main problem is that an individual who is actually eligible for the benefit will in that regard be similar to everyone else who also receives the benefit. And in this way identification problems arise that are hard to solve. In the international literature there has been some research on a related issue – namely, the question whether there is a trade-off between social protection and economic flexibility. In an NBER volume on this issue, Blank (1994) concludes that there is little evidence that labour market flexibility is substantially affected by the presence of social programmes, nor is there evidence that the speed of labour adjustment can be enhanced by limiting these programmes. This conclusion is based on studies for Sweden, Germany, France, Belgium, the Netherlands, the United Kingdom, Spain, the United States and Japan. The conclusion seems to be that countries with more extensive social protection systems find other ways by which adjustment can occur. In other words, there is some evidence that high social protection does not necessarily imply low economic flexibility and a high level of rigidity. Instead in countries with many social programmes labour market adjustment works through other channels. One example of this is that in comparison with the United States, Germany and Belgium rely more heavily on hours’ adjustment rather than employment adjustment when faced with an economic downturn. While in Japan even though there is less interregional mobility than in the United States there are high rates of interregional commuting. Another example is that, compared to the United States, Sweden and Japan rely more heavily on older workers leaving the labour market in an economic downturn to keep unemployment rates low. Studies for Sweden also indicate that while young Swedish women have labour force participation rates very similar to young American women a much higher fraction of young Swedish women take more than a year off work when they have a child. And finally in Denmark the leave schemes (parental, educational and sabbatical) have also reduced unemployment rates, and it is not clear as yet whether the leave schemes in a labour market view have been a failure or not.
328
Sløk
Seen in the perspective of incentive problems the early retirement scheme is a good example of the complexity involved in discussing such problems in the welfare state. The incentive problem of the early retirement scheme lies to some extent in the existence of the scheme. The only eligibility criterion is your age. Everyone who turns 60 weighs individually marginal cost against marginal benefits of retiring early. The marginal benefit is the amount of money (and the utility from leisure) you received when retiring early, but the marginal cost (or utility loss) depends on the current wage, as well as health conditions, happiness, job satisfaction and other such intangible elements of life. An important part of these intangible variables is the social norms of society. If the social norm says that it is acceptable to try and exploit as many of the benefits from the welfare state as possible then it is no surprise that the welfare state is being overburdened by an increasing number of people. The problem is that recipients and potential recipients tend to think of the welfare state as an additional economic source and this is precisely what is unsustainable in the longer run. In an ideal world individuals would perceive the benefits of the welfare state as an offer that is available if needed rather than viewing the benefit as a claim to which they have an automatic right. Imagine two polar cases. At the extreme, if there were no social security or benefits in kind at all, then there would be no incentive problems. At the other extreme is the case of no eligibility criteria for receiving money, so that each individual could just get the amount he or she requested from the government sector. In this second case if everyone behaved in a morally optimal way we would have no incentive problems (since people would request exactly the amount of help they need, which in many cases would involve no help). But since people seek to optimize their personal wealth we need to make eligibility criteria (or mechanism design) for the individual benefit. To implement such a criterion can be a complex task. In the real world, however, the bottom line is that it is the change in social norms between viewing benefits as an offer and a claim that is the real source of the incentive problems in the welfare state. This brings us to the question of whether we should in fact be concerned by the incentive problems that exist in the welfare state. I would say ‘not necessarily’, partly because it is impossible to eliminate all incentive problems in an administratively feasible way and partly because if the positive externalities of the welfare state are bigger than the negative externalities (like incentive problems) then we might still be better off just accepting them while acknowledging that it is very hard to design mechanisms or benefits that can and will not be capitalized on. Our ambition must then be to design the benefits as precisely as possible and in an administratively feasible way. If the welfare state is part of the explanation of why Denmark year after year is one of the richest countries in the world and if the adjustment of the economy takes place through other channels as the international evidence, for example for Sweden and the Netherlands shows, then it would be a bad strategy to try to reduce the welfare state dramatically in the belief that it would increase welfare.
References Barr, N. (1993) The Economics of the Welfare State (Stanford: Stanford University Press). Blank, R. (1994) ‘Social protection versus economic flexibility. Is there a trade-off?’, NBER Comparative Labour Market Series (Chicago: University of Chicago Press).
Comment
329
World Bank Group (1996) World Development Report 1996 (Washington, DC: World Bank) pp. 188–9.
Comment Thorvaldur Gylfason Søndergaard’s paper provides an informative account of economic developments in Denmark in a long-term perspective, with emphasis on various incentive problems arising from Danish welfare policies, especially in labour markets. Søndergaard’s conclusion is essentially: so far, so good. While I do not necessarily disagree with this assessment per se, I still see several important similarities between Denmark and Sweden (and, indeed, some other European countries) – similarities that seem likely to weaken Danish growth prospects in the future unless corrective action is taken. It seems to me that Søndergaard tends to make too little the looming danger of a gradual erosion of economic incentives in Denmark over the past 25 years. The erosion of incentives is like overfishing and other forms of environmental degradation: it can stimulate economic activity for quite a long time, before the damage becomes apparent. Therefore, the likely consequences of blunted incentives must not be judged solely on the basis of currently available empirical evidence. They must also be assessed in consideration of the experience of comparable countries as well as intuition and deductive analysis. I wish to organize my discussion of the paper around three main themes: (a) welfare, equality, and efficiency; (b) wages, education and growth; and (c) the size of government and growth.
WELFARE, EQUALITY, AND EFFICIENCY By a welfare state I mean a state where the allocation of responsibilities between the private sector and the public sector, and between individuals and organizations, is aimed at securing reasonable equality among the citizens and social justice without inhibiting economic growth. In my view, therefore, the linkages between equality and efficiency are at the heart of the welfare state.1 These linkages can perhaps most easily be summarized by a simple parabola (Figure C10.1). Gross inequalities of income and wealth have been a socially disruptive force in many developing countries, and appear thus to have hampered economic efficiency and growth. Brazil is an extreme case: more equality would almost surely be good for growth in Brazil. For another example, rapid growth in East Asia has generally been accompanied by income inequality that is no more pronounced than that observed in many OECD countries. My guess is that a large majority of the world’s countries belong on the lower segment of the parabola, and they would thus benefit from incentive-compatible government policies aimed at fostering social cohesion, not least through more high-quality public expenditure on education, health, and infrastructure. Denmark and Sweden, on the other hand, and a few other countries have probably expanded their public undertakings and their quest for equality a bit too far; if so, a smaller public sector and a little less equality would, in their case, probably be good for growth. Moreover, by thus stimulating growth, a little less equality today may actually help create conditions for higher living standards for those at or near the bottom of the distribution of income, education, and wealth.
330
Comment Figure C10.1
331
Equality and efficiency
Equality Sweden Denmark
Brazil
Efficiency
Mapping the various possibilities and quantifying the potential trade-offs involved – in short, bringing distribution and social-welfare issues to the forefront of macroeconomics on a par with unemployment, inflation, and growth – remains, in my view, one of the most important items on our profession’s agenda of unfinished business.
WAGES, EDUCATION, AND GROWTH In modern welfare states, the quest for greater equality has taken three main forms: through (a) public expenditure on education and health, (b) taxes and transfers, and (c) centralized wage bargaining. By design, the centralization of wage bargaining has led to a greater equalization of wages, both before and after tax, than would have occurred in a more flexible labour market. This wage compression, in turn, seems to have reduced the demand for education among young people – why bother to educate yourself if the private return to education does not seem to justify the effort? – and thus impeded growth. There is some evidence for this phenomenon in Sweden 2 (and also in Iceland). I fear that the recent emergence of this problem is not confined to those two countries. According to the World Bank, for example, school enrolment at the tertiary level in Denmark rose from 28 per cent of the relevant age group in 1980 to 41 per cent in 1993. Comparable figures for Sweden are 31 per cent and 38 per cent, and for the high-income countries as a whole 35 per cent and 56 per cent.3 Thus, both the level of
Gylfason
332
tertiary school enrolment in Denmark and its increase since 1980 are below the average of the high-income countries. These trends merit detailed empirical scrutiny in a macroeconomic context. The upshot of my argument, if further empirical research shows it to be valid, is as follows. Decentralization and liberalization of labour market arrangements in Denmark and, more generally, in Europe are necessary not only in order to reduce unemployment permanently to lower levels without reigniting inflation, but also to help create stronger private incentives for education and thereby also strengthen the foundations of high-quality growth. This is so simply because virtually anything that helps promote economic efficiency is good for growth, if not for all time, as endogenous growth theory would have it, then at least for periods long enough to be highly relevant from the point of view of public policy and welfare, according to the neoclassical growth model. My message is simply this. If the allocation and distribution of resources needs to be changed, then, as we know, it is almost always inefficient to try to do so by interfering with prices or wages on a grand (i.e., macroeconomic) scale. It is more efficient to use public expenditure, taxes, and transfers – within reasonable limits. This means that we should let the labour market alone. The message is not that the general aims of the welfare state should be abandoned or modified, for that is not our department, but rather that worthy political and social aims deserve to be pursued with the most efficient economic methods at our disposal. Before proceeding to my last point, let me consider two other aspects of efficiency and growth in Denmark, besides education: investment and external trade. The table below compares Denmark with Sweden and with the world at large. We see here that Denmark, like Sweden, invests too little: the investment ratios of the two countries lie about one-third below the world (and OECD) average, not to speak of East Asia, where saving and investment rates of 30 per cent to 40 per cent are common. More surprisingly, perhaps, Denmark, like Sweden, also seems to export and import too little. In these two countries, whom we are used to regarding as being at the forefront of world trade, exports and imports of goods and services are actually below average in the world economy as a whole. The point is that Danish and Swedish trade has not kept pace with the rapid growth of world trade in recent years. Insofar as trade expansion is good for growth, Denmark and Sweden seem to be lagging a bit behind in this respect (Table C10.1).
Table C10.1 Denmark
Denmark Sweden World *
Denmark, Sweden and the world: investment, trade, and growth Investment (1995, % of GNP) 16 14 23*
Trade (1995, % of GNP)
Growth per head (% p.a. 1985–95)
64 77 80**
1.5 –0.1 0.8*
Weighted by country size. Unweighted by country size. Source: World Bank Atlas 1997 and World Development Report 1997. **
Comment
333
THE SIZE OF GOVERNMENT AND GROWTH Is it true, as is sometimes claimed, that economic growth is essentially independent of the size of the public sector? This is clearly a crucial question in a country which, in 1995, according to the OECD, had the world’s largest public sector measured by total government revenue in proportion to gross domestic product (GDP).4 I think it is fair to say that, thus far at least, the empirical evidence is mixed. Cross-section data, where each country is represented by a single pair of averages spanning several decades, generally indicate a rather weak connection between the size of government and growth. For example, a recent study of OECD countries reported that an increase in the ratio of total government spending to GDP from, say, 30 per cent to 50 per cent is associated with a decline in GDP growth over the long run from about 3.2 per cent per year on average to about 2.6 per cent.5 But valuable information may be lost by taking such an aggregate approach. It would, I guess, be more informative to use panel data, where each country is represented by several pairs of averages, each spanning, say, a five-year interval, thus exploiting also the time properties of the data. Figure C10.2 illustrates my point. Imagine first a case where the cross-section observations are evenly spread out within the thick square: then no relationship between the size of government and growth can be discerned from the data. But if each dot within the square represents an average of several observations lying on axes from north-west to south-east in the figure, then the panel data fit in the box, not the square. If so, a panel regression may well produce evidence of a negative relationship between size and growth, even if a cross-section regression based on the same data fails to do so. Even so, the causation may run both ways: not only is big government likely to blunt incentives and reduce growth (for
Figure C10.2
Size of government and growth
Size
Growth
334
Gylfason
example, through rapid growth in benefit dependency as documented by Søndergaard; see his Figures 10.1 and 10.2), but slower growth tends to inflate demands for increased government expenditure on social security. There are other measures of government interference in economic activity that may also be relevant for growth. Recent work suggests that economic growth across countries is inversely related to the share of state enterprises in the labor force and in GDP.6 Perhaps it will turn out that the size of government is not as important as what it does and does not do.
Notes 1.
2. 3. 4. 5. 6.
Efficiency can for our purposes be measured, e.g., by the purchasing-powerparity-adjusted per capita gross national product in 1995 (USD 5.400, 21.200, and 18.500 in Brazil, Denmark, and Sweden, according to the World Development Report 1997). Equality can be measured by the ratio of the incomes of the lowest quintile of the population to the highest quintile (0.03, 0.14, and 0.22 in Brazil, Denmark, and Sweden in the 1990s). These numbers correspond roughly to the co-ordinates of the three hypothetical observations shown in Figure 1. See Gylfason, T., T.M. Andersen, S. Honkapohja, A.J. Isachsen, and J. Williamson, The Swedish Model under Stress (Stockholm: SNS Förlag 1997) ch. 5. See World Development Report 1997, table 7. Sweden had the largest public sector in 1995 measured by total government expenditure in proportion to GDP; Denmark was second. See Gylfason, Andersen, Honkapohja, Isachsen and Williamson, The Swedish Market, ch. 6. See T. Gylfason, T.T. Herbertsson and G. Zoega, ‘Ownership and growth’, CEPR Discussion Paper (forthcoming)
11 Effort Commitment in Active Labour Market Policy1 Claus Thustrup Hansen and Torben Tranæs
11.1
INTRODUCTION
The received perception of what is the right mixture of benefit reforms and active measures in the battle against unemployment varies. Despite the warnings from Scandinavian economists like Calmfors, Forslund, Nymoen and others, active measures were highly recommended some years ago with Sweden as the example to follow (see, for example, Jackman, 1990; Layard, Nickell and Jackman, 1991). After the major increase in Swedish unemployment after 1991, economists tended to be more sceptical and the extensive studies of active labour market programmes is now the foundation for serious doubts about the effectiveness of active labour market policy (hereafter, ALMP). The discussion of the appropriateness of ALMP programmes in the fight against unemployment has focused on labour supply and wage formation effects. The arguments in favour of active programmes stress that such programmes can help increase the effective labour supply: by training unemployed workers, or at least keeping them on their toes, these programmes prevent the ‘discouraged-worker effect’ of unemployment. Consequently, the jobless are more effective job seekers; they are more attractive to firms and – equally important – they believe so themselves. This tends to increase the effective labour supply which eventually transforms into additional employment (see for example Jackman, Pissarides and Savouri, 1990; Layard, Nickell and Jackman, 1991). The argument against ALMP points out that active programmes may ‘crowd-out’ regular employment. First, the programmes tend to reduce regular employment to the extent that unemployed workers in programmes do not search or search less than the open unemployed. This reduces the effective labour supply and hence employment. Second, ALMP tend to reduce employment by reducing the disutility of lay offs compared to the situation without these programmes. In imperfectly competitive labour 335
336
Hansen and Tranæs
markets, this will spill over to wages and thus eventually reduce employment (see Holmlund, 1990; Calmfors and Forslund, 1991; Holmlund and Linén, 1993; Calmfors and Lang, 1995). This argument relies on the assumption that the (discounted present) utility of entering a programme is higher than the utility of continuing as openly unemployed2 – for instance, because programme employment is compatible with regular jobs. These objections against ALMP are important but they rely on active programmes having a very specific design; small changes may mitigate or even reverse the crowding-out effect. What is needed is to make sure that workers in programmes do search and that utility does not increase when entering a programme from open unemployment, and neither is incompatible with ALMP. On the contrary, requiring participation in active programmes implies by itself that time (leisure) is appropriated in exchange for benefits (unemployment pay). This reduces the private utility of an ‘activated’ unemployed person and hence wage pressure is moderated or reversed depending on the income level during participation in programme and the required effort. Moreover, by varying the time or effort requirement, it is possible to increase the effective labour supply: some workers who are initially voluntary unemployed will wish to find a job and all involuntary unemployed will increase their search effort in order to ‘escape’ the programme faster. The labour market policy implemented in Denmark after 1993 has explicitly emphasized the effort–commitment aspect of ALMP. In general, effort is required in exchange for unemployment benefits after some time as open unemployed. The effort can be work (public or private), job training, or education. The important issue is how to determine the size of the effort requirement. According to the new Danish ALMP the basic principle is that: (1) a worker in a programme receives the hourly wage rate of his latest regular job, and (2) the total salary cannot exceed what the worker would get in unemployment benefits as open unemployed.3 Typically, the programme salary is equal to unemployment benefits. For (1) and (2) to be compatible, the number of hours’ effort required has to be the benefits divided by the wage rate. This means typically shorter hours compared to what is required in ordinary employment, but it makes sure that utility as activated unemployed is not higher than as open unemployed. In this chapter we study this effort–commitment principle theoretically in order to evaluate the prospects for the Danish reforms. The main results are as follows. First, effort–commitment is effective in general and the principle introduced in the new Danish legislation is particularly fortunate since it facilitates the coexistence of high social security and incentives for labour market participation. Stronger participation incentives reduce voluntary unemployment which may amount to as much as 40 per cent of total unemployment in Denmark.4 Second, in a unionized labour market the effort–commitment benefit system implies lower wages than a pure benefit
Effort Commitment in Active Labour Market Policy
337
system as long as unemployment is not too high. This wage-moderating effect, or ‘crowding-in’, is stronger the less correlated is the effort requirement of an unemployed worker and the wage negotiated by his union. This reduces involuntary unemployment. Thus, the reforms can potentially improve the functioning of the labour market. So how do they work? Unfortunately, it is too early to provide a proper empirical analysis of the impact of the reforms; the programmes are not fully implemented and on top of that, data always comes with a lag. Therefore, we confine ourselves to some casual evidence and leave comprehensive empirical investigations for future research. This evidence suggests, nevertheless, that the policy might have had a positive effect on the functioning of the labour market. The sharp drop in unemployment between 1994 and 1997 (from 12.4 per cent to 7.7 per cent) was not accompanied by accelerating wage inflation. Another related indication is the relative number of vacancies, which did not increase when unemployment dropped. The chapter is organized as follows. Section 11.2 reviews the labour market policy after 1973 in Denmark, with special focus on the main novel ingredients of the active policy implemented after 1993. Section 11.3 briefly discusses some evidence. Section 11.4 looks at the consequences for participation incentives and section 11.5 considers the impact on wage setting. Finally, section 11.6 draws some conclusions.
11.2
POLICY AGAINST UNEMPLOYMENT IN DENMARK
Policy makers have been focusing on two main strategies in the fight against unemployment over the last two decades, an incentive strategy and an educative strategy. The first emphasizes the need for stronger incentives for job search and wage moderation and suggests reductions in unemployment benefits. Beside reducing unemployment such measures may raise inequality. Advocates of the incentive strategy consider this to be unavoidable, but point out that inequality increases only on the margin since the main effect is that inequality with respect to unemployment risk is exchanged for wage inequality. The educative strategy renounces the use of stronger economic incentives precisely because of the adverse effect on inequality – someone will always remain unemployed and they will have to live on the reduced benefits. Instead, unemployed workers should be educated so that they neither have to be unemployed nor be forced to take up low-paid jobs. During the 1970s and 1980s the emphasis has been on the educative strategy although economic incentives were strengthened in the mid-1980s. The reforms of the 1990s represent a compromise between the two strategies. They provide stronger incentives but do so through effort–commitment
338
Hansen and Tranæs
rather than by cutting benefits. Furthermore, the stronger incentives are provided in such a way that it is possible to upgrade unemployed workers qualifications, which is the main objective of the educative strategy. Below we discuss this in more detail. Before 1993 The key element of the labour market policy in place until 1993 was implemented during the late 1960s and the early years of the 1970s. Here, a general replacement ratio of 0.9 combined with a relatively high benefit maximum was introduced. This nearly doubled the average replacement ratio for unemployed (from slightly above 0.40 in the mid-1960s to 0.75 10 years later (see Neumann, Pedersen and Westergård-Nielsen, 1991). For years after the sharp rise in unemployment, following the first oil price shock, no systematic initiatives were taken to reduce unemployment; it was seen as a temporary phenomenon – not a structural one. The primary concern was to prevent unemployed workers from losing their benefits. This 1 was done by expanding the duration of benefits to 2 2 years and making it 1 easier to qualify for another 2 2 years of benefits. This is likely to have increased structural unemployment. Different elements of ALMP came gradually into the legislation during the 1980s. These were the standard measures such as education, job training and relief work. An active programme then was something that an unemployed person could join voluntarily or was offered as a mean to be entitled 1 to yet another 2 2 years of benefits. Hence, the benefit system gave income security for a long period to all – formally 9 years including two mandatory 6month spells of relief work or ‘training’ arranged for by the public employment agency. However, in practice benefits were more or less open-ended. The relative income during unemployment varied from a replacement ratio at 0.9 for the low-earning groups to well below 0.5 for high earners (because of the benefit maximum). For high earners the replacement ratio became particular low because the maximum was not indexed during a period in the 1980s. Labour Market Policy After 1993 In 1993 and again in 1995 reforms were enacted to shift the emphasis in labour market policy. The new policy has increased the incentives for participation significantly, and this has been accomplished by only a minor fall in income security. The reforms contained some measures intended to improve the match making ability of the labour market in general, while others aimed at improving the participation capacity and incentives of the individual job seeker. The latter
Effort Commitment in Active Labour Market Policy
339
part was divided into a reform targeted at persons below 25 years and reforms targeted at adults. In this chapter we concentrate on the reforms for adults and comment only on the reform targeted at youngsters. The reforms for adults (persons above 25 years) contained the following main elements. Concerning benefits, the 9-year duration was reduced to 5 years and now a new period on benefits requires regular work, programme participation will no longer do. This is not such a big change, however, because after 5 years of unemployment you can still obtain social security from your local council and this is not that much lower than unemployment benefits. The most important change is that after 2 years, unemployed workers have to be activated, which means that they have to make an effort to continue being entitled to benefits. Thereby, the reforms have increased participation incentives quite significantly but maintained income security at a high level. The effort requirement can be met by participation in one of the following programmes. General Job Training This is short-term job introduction and training in the private as well as in the public sector. The employer is given a subsidy and if job training is for more than 6 months the employer has to hire the person in a regular job afterwards. The wage rate and working conditions follow the standard terms agreed upon in collective bargaining. After April 1996 a wage ceiling applies. Individual Job Training This is job introduction as above, but here the wage rate may be lower and the subsidy higher than under the general programme. This programme is meant to give the public employment agency additional means when it comes to unemployed workers that are particularly difficult to get reemployed. Individual job training is arranged for a particular individual, not for a group or an area. This signals to ‘insiders’ that it is not a short cut to lower standards in general. Pool Jobs This is relief work in the public sector of up to 3 years’ duration. An unemployed worker qualifies for this after 2 years of unemployment (within the last 3 years). The jobs can only be in certain sectors that are given special priority by Parliament. Education/Training This can either be in the ordinary education system or in special training programmes set up by the public employment agencies.
340
Hansen and Tranæs
An unemployed worker can choose from this menu after being unemployed for 1 year and is required to enter one of the programmes after being unemployed for 2 years; alternatively, benefits expire.5 Concerning the income during participation in a programme the general principle is that the income cannot exceed the unemployment benefits that the person can obtain as open unemployed (see The Danish Ministry of Labour, 1996). However, there are exceptions to this principle – for instance, it does not apply to job training in private firms. Here, the terms are negotiated between the firm and the public employment agency. The other part of the reforms decentralized the administration of the ALMP to a number of local Labour Market Councils (LMCs) with responsibility for their own region (14 regions cover Denmark). The LMCs were given discretionary power over important matters concerning implementation of the reforms. For instance, if a firm anticipates the need for certain skills in the near future, the local LMC can finance a specific training programme and recruit an appropriated number of unemployed workers to meet the firm’s need. The purpose of the decentralization is to improve the contact between local firms, local unions and local public employment agencies (the AF offices), in order to improve information and thereby facilitate better job matches without additional wage pressure.
11.3
SOME EVIDENCE
We start this section by a brief overview of some micro evidence of the effects of different active programmes. Only one study has looked at the period after the new reforms were implemented. Thus, it is basically impossible to judge whether programme participants have better employment chances after the reforms. Furthermore, we argue that existent evidence can neither verify nor falsify whether such programmes improve the overall performance of the labour market. Therefore, we turn to some macroeconomic indicators: a comparison of the boom of the 1980s (from 1984–7) with the one starting in 1993 suggests a structural improvement of the Danish labour market between the two booms. Weise (1997) surveys the evidence of employment effects at the micro level of different programmes before the new labour market reforms. The conclusions are mixed.6 Most studies on job training in the private and public sector find that it increases the participants’ employment probability but that the effects are lower in the public sector.7 The local authorities arrange relief work for unemployed that receive social security. Before the new reforms, these programmes qualified participants for unemployment benefits after the programme but the effects on employment were either no impact or a negative impact. No overall picture is clear from the evidence on education schemes.
Effort Commitment in Active Labour Market Policy
341
Langager (1997) analyzes the employment effects on insured unemployed on data from 1995 – that is, after implementation of the 1993 reform. The results indicate significant positive employment effects of general and individual job training in both private and public schemes and also positive employment effects of education except on voluntary educational leave schemes. None of the existent micro evidence enables us to evaluate the effects of the different programmes on the overall performance of the labour market. Positive employment effects on programme participants may have no macroeconomic impact because subsidized employment crowds out ordinary employment8 or even has negative effects because it increases average utility of unemployed leading to wage pressure in the overall economy. Non- or negative effects on programme participants may also coexist with positive effects in the aggregate because open unemployed search more intensively in order not to enter a (compulsory) programme in which the effort commitment reduces utility below the utility of open unemployment. Furthermore, if effort commitment reduces utility below the utility of open unemployment, it will tend to reduce wages and improve the employment probability of everybody but not necessarily improve the employment prospects of a programme participant relative to a nonparticipant. Unfortunately, it is too early to judge empirically whether there has been any macroeconomic impact of the new reforms. Some macroeconomic indicators do, however, point to an improvement in the Danish labour market between the boom of the 1980s and the boom that started in 1993. The first boom had an annual real GDP growth rate equal to 2.7 per cent and the second one has so far (1996) had 3.2 per cent growth. Thus, both booms are significant compared to an annual growth rate of 2.1 per cent from 1980 to 1996. We focus on two pieces of evidence for these two periods – wage inflation responses to unemployment and the shift in the Beveridge curve. We start with the Beveridge curve. The expected reaction to a boom is an increase in vacancies and a reduction in unemployment, just as happened in the boom of the 1980s (see Figure 11.1), not only in Denmark but generally throughout the OECD. Moreover, the anatomy of a boom seems to be that the vacancy rate responds first, and with the strongest rise early in the boom, while the drop in unemployment tends to come later. The development in Denmark between 1993 and 1996 has been remarkably different from this picture (see Figure 11.1): the big reduction in unemployment has not (yet) been accompanied by an increase in vacancies,9 thus, the Beveridge curve seems to turn west or even south west. Another indication of a structural improvement is the weak response of wage inflation to the drop in unemployment. The annual real wage growth of blue-collar workers was 2.2 per cent from 1984 to 1987, after which unemployed started to increase again.10 The annual real wage growth in the latest boom is so far only 1.1 per cent,
342
Figure 11.1
Beveridge curve for Denmark, 1980–96.
Vacancies in per cent of the labour force 0.1200
0.1000
0.0800
1993
1987
0.0600
1996 0.0400
1980 0.0200
0.0000 6.00
1984
7.00
8.00
9.00
10.00
Rate of unemployment (%)
11.00
12.00
13.00
Effort Commitment in Active Labour Market Policy
343
although the boom is at least as strong as the former one measured both by GDP growth and the reduction in unemployment.11 We are aware that this is casual evidence, indeed. For instance, the wage increase in 1987 should not be fully attributed to low unemployment. In the years before these wages were negotiated, oil prices took a big drop and profitability in general improved. With union–firm bargaining, such gains will be shared to some extent and thus, everything being equal, wages should go up for these reasons alone. 11.4
PARTICIPATION INCENTIVES
In this section we discuss the design of effort–commitment benefit systems and its implications for participation incentives. The Basic Incentive Problem The basic problem to be considered is a classical dilemma facing any welfare state government; it wishes to provide well for the unlucky workers and at the same time make sure that no one prefers to be ‘unlucky’ – that is, make sure that everybody has an incentive to participate in the labour market in order to be self-provided. Since the government (usually) cannot observe a worker’s disutility of work the benefits provided for jobless workers may have to be very low to ensure that all types of workers have incentives to participate. Alternatively, the government has to accept high expenditures on unemployment benefits and welfare. This is illustrated in Figure 11.2 where Y denotes income, l hours of work, and l maximal hours of work. Consider two workers both with hourly productivity w (the slope from the left of the budget line); they have preferences over income and leisure represented by v1 and v2. Now, a government that cannot tell who is who can guarantee an income at B only if it wants both types to have incentives for self-provision, which here means that they both prefer what they can achieve at the labour market represented by points 1 and 2, respectively. The highest level of social security that does not conflict with work incentives, B, will in general be low – in many countries much lower than the level that is considered to be the minimum acceptable income level. This is certainly the case for Denmark. So here is an obvious political constraint on the problem faced by the government – one of several constraints. Political Constraints A benefit level at zero would be the obvious solution if cost minimization was pursued by the government with no constraints. However, welfare state
344 Figure 11.2
Hansen and Tranæs The basic incentive problem
governments typically face several political constraints. First, the opportunity of a minimum level of income, B ≥ B, has to be available to everybody, so that it is by choice if they receive less than B. This constraint is the horizontal line in Figure 11.3. Second, it has to be possible to obtain B within a maximum number of hours; 24 hours per day are the physical maximum, but the politically relevant maximum is, of course, much lower. This constraint is the vertical line at ln. Finally, there might be an additional constraint saying that any work has to be paid at least some minimum wage, which also applies when the government requires work in exchange for benefits. The relevant
Effort Commitment in Active Labour Market Policy Figure 11.3 Y
345
The effort–benefit system v2
v1
v3
1
2 B
3
4 5
w l
– l
ln
l e(w)
0
constraint in Denmark is that each type of worker has to be paid the going wage rate – that is, a wage equal to the worker’s productivity. This is represented by the line of slope w (from right to left). The latter constraint is typically demanded by unions but can also be a more general electoral demand to remove any resemblance between ALMP and the erstwhile poorhouses. Taking all constraints into consideration leaves the shaded area of Figure 11.3. Only from this area can the government choose an alternative to a pure benefit system (point 3) so as to pursue its objectives.
346
Hansen and Tranæs
Government Information Our basic assumptions concerning information are that individual productivity is public information while disutility of work is private information. Thus, the authorities can observe w in Figure 11.3 but not preferences illustrated by the indifference curves v1, v2 and v3. This is the appropriate assumption for the Danish system since benefits are related to the last work record. The system rewards high types and recipients have to document their type by a past work record (employment contract, wage bill, etc.). Therefore, we cannot apply the frameworks of Besley and Coate (1992, 1995) who study a somewhat related problem under the assumption that productivity is unobservable while preferences are known. The Labour Supply Decision Consider a population consisting of individuals characterized by their preferences over income and leisure and their income-earning ability w, which is interpreted as the person’s productivity and the wage rate that she or he can obtain on the market. Preferences are assumed to be strictly convex over income, y, and leisure and are represented by the indirect utility function vi(y, l) where we for simplicity use ‘work’, l, as an argument rather than leisure, l – l. By ‘work’ we mean time spent on all activities that generate income. The Pure Benefit System As a benchmark case consider first a pure benefit system. Labour supply of individual i under a standard benefit B is l (w) if v i (wl i , l i ) ≥ v i (B, 0) l i (w, B) = i otherwise 0
(11.1)
where li(w) is the amount of labour supplied to the market in the absence of any benefit system. The objectives of the government are to provide income security B at minimum costs, which is achieved if all types have an incentive to participate and become self-provided. The two objectives obviously conflict under a pure benefit system. Consider a group with income-generating ability w. Depending on the distribution of vi(wl*i , l*i ), where l*i = li (w), it is possible to guarantee only a certain low level of income, if the government wants all types to participate. Thus, the government has to sacrifice either its target level for income opportunities or its ambition of inducing incentives for self-provision to all types of workers (that is, accepting some voluntary
Effort Commitment in Active Labour Market Policy
347
unemployment). This problem can, nevertheless, be overcome by a simple effort–benefit system. The Effort–Benefit System The task is to design a general effort–benefit system that uses only available information and, nevertheless, generates participation incentives for all types of workers – that is, for any w. This section shows that the two objectives of the government do not necessarily conflict. Even with a simple effort–benefit system, it is possible to provide high income security and induce participation incentives at the same time as long as workers do not have too low productivity (that is, they must have productivity w ≥ B/ln). An effort–benefit package, {B, le}, consists of a money transfer, B, and an effort requirement in terms of hours, le, which has to be met before B can be collected. Suppose w ≥ B/ln, then knowing a worker’s productivity is sufficient to create participation incentives at (almost) any level of income security as long as preferences are strictly convex over income and work. Knowledge of the disutility from work is no longer needed. The argument for this is simple. Suppose that benefits are only offered as part of the effort–benefit package {B, le(w)} (see point 4 in Figure 11.3) consisting of the uniform type-independent monetary transfer B, and an individual effort commitment which is inversely related to the productivity of the workers in the following way, le(w) = B/w – that is, all programme work is paid the going wage rate as the 1993–5 reforms prescribe. Under this system type i’s labour supply is l ( w) if vi ( wli , li ) ≥ vi ( B, B/ w) li ( w, B, l e ) = i otherwise 0
(11.2)
Consider a worker with productivity w and let l*i = li(w, B, le). Now for any level of benefit B ≠ wl*i and independently of the worker’s disutility from effort, participation is strictly preferred to the effort–benefit package {B, B/w} – that is, vi(wl*i , l*i ) > vi(B, B/w). This is illustrated in Figure 11.3 where types with v1, v2, and v3 choose point 1, 2 and 5, respectively, instead of point 4. In conclusion, the simple effort–benefit package {B, B/w} is incentivecompatible for all types (w ≥ B/ln); it reinforces the efficient labour supply and removes voluntary unemployment altogether. Workers with Productivity below the Threshold Thus, in general, it is possible to provide participation incentives without sacrificing income security or the principle of paying all programme
348
Hansen and Tranæs
activities the going wage rate. This may, however, imply unrealistically long hours for certain low-productivity types. In practice there will be a political constraint on hours as mentioned above – that is, a maximum number of hours during which a person has to be able to provide for herself. Alternatively, the government will offer social assistance. Suppose that ln is the standard work hours to serve as the government’s norm. Then what we did above applies to all types of workers with productivity w ≥ wn ≡ B/ln. For workers with productivity below this threshold, the effort–benefit package above requires a number of hours l > ln, which is not possible given the political constraint l ≤ ln. This is illustrated in Figure 11.4 where point 4 lies to the left of point 5. If types with productivity below the threshold are to be given incentives to participate and become self-provided – at least partially – the task becomes harder and participation can no longer be without cost to the government. Partial self-provision is the best one can hope for concerning the low-productivity types. Obviously, the cost-minimizing effort–benefit package is {B, ln}. But if this is offered in isolation some workers with productivity w < wn may claim the package, because self-provision is not as good as the minimum acceptable level defined by the government (that is, the opportunities on the market for these types are no better). In Figure 11.4, this is the case for the v1 type, as point 5 lies in the preferred set compared to point 1. This is not the case, however, for the v2 type who prefers point 2 to point 5. Assuming that the government wishes to provide incentives to any ‘possible’ type of worker, it has to improve the alternative to the package {B, ln} – that is, make self-provision more attractive. This can be done by subsidizing market activities. The government can offer the choice between a wage subsidy s and the package {B, ln}. If the subsidy compensates for the worker’s lack of productivity such that w + s ≥ wn, all workers with productivity w prefer partial self-provision (that is, they accept s and supply labour at the rate w + s) to the package {B, B/w} for the very same reason that workers with productivity above wn prefer self-provision. The new choice of the worker is a point like 6 in Figure 11.4 from which it is clear that it is no longer possible to reinforce an efficient individual labour supply; some may supply more than what is efficient (type 1) and some may supply less (type 2).12 Caveats The simplicity of the model implies that a number of questions arise when the results are compared to reality:
Effort Commitment in Active Labour Market Policy Figure 11.4
349
Productivity below the threshold
Why are so Many People Still in Programmes? We have ignored any restrictions on the individual choice of work hours. In reality, some workers’ choice of labour supply is restricted to the standard work hours, say ln. These workers may prefer the package to a regular job given the restriction on hours. Furthermore, many unemployed workers are involuntary unemployed (see section 11.5), whereas in the model above unemployment is purely voluntary. Involuntary unemployed cannot just find
Hansen and Tranæs
350
a job that match their qualifications; they have to be in a programme or take a job for which they are overqualified. Why Acquire Skills? An obvious drawback of the subsidy system on p. 347 is that some lowproductivity types no longer have any incentives to acquire skills. This problem will accompany any attempt to bring about an income distribution that differs from the distributions of ability, and an important part of the new labour market policy (that we have ignored here) deals explicitly with this problem – namely, the reform for workers under the age of 25 years which was part of the second reform enacted in December 1995. This reform reduced unemployment benefit for youngsters (age under 25 years) by 50 per cent and introduced mandatory activation after 6 months (as opposed to after 2 years for adults). Before this reform, an unemployed person under the age of 25 years received from the government twice the amount in welfare of an ordinary student. After the reform they receive about the same, but the general conditions for students are slightly better. A year later (January 1997) unemployment had dropped by 50 per cent for the young workers targeted by the reform. This should be compared to the fall in overall unemployment from 10 to 8.8 per cent.
11.5
WAGE SETTING
So far, we have analyzed only individual incentive effects for a given wage. Of course, the aggregate outcome depends on the institutional settings in the labour market. As a benchmark case, consider first price taking behaviour. Unemployment benefits will in this case reduce the aggregate labour supply and thus reduce employment (below the efficient level). Unemployment occurring in the statistics is then purely voluntary and due to the benefit system alone. Section 11.4 showed that the introduction of an effort–benefit system increases the labour supply of those receiving benefits and leaves unchanged the labour supply of those already working. Hence, such a system reduces voluntary unemployment. Furthermore, if everyone is capable of earning the benefit level in the market (that is, no one has a productivity below the threshold Wn) then the system removes voluntary unemployment entirely and reinforces an efficient supply of labour even though unemployment benefits exist. Yet, the actual wage formation in most (European) labour markets is not well described by a Walrasian market. Some wages are more or less set by trade unions, others are determined by bargaining between groups of workers and groups of firms, and others again are set by firms taking efficiency wage
Effort Commitment in Active Labour Market Policy
351
considerations. These wage setting institutions lead to rationing in the labour market. The high unemployment in many European countries is probably a mixture of both voluntary and involuntary unemployment, although it is difficult to distinguish between the two.13 The unemployment benefit system has also an important role in theories explaining involuntary unemployment; as in the case of voluntary unemployment, this occurs because unemployment benefits raise the opportunity costs of employment. The purpose of this section is to explore the effects on wage setting and involuntary unemployment of introducing an effort–benefit system as discussed above. In doing so, we concentrate on a simple theory of trade unionism – namely, the monopoly union model of Dunlop (1944). However, it is easy to show that the qualitative results carry over to more complicated trade union models, for example the ‘right-to-manage’ wage bargaining model. We end by a brief discussion of the corresponding results in an efficiency wage analysis. A Simple Unionized Labour Market Section 11.4 analyzed the labour supply responses of different types of workers to an effort–benefit system. In this section, we disregard these effects in order to isolate the direct wage setting effects of introducing an effort– benefit system. More specifically, we consider only a particular type of worker, all organized in a trade union. There are N members of the trade union and each member supplies inelastically 1 unit of labour. Firms demand labour according to a downward-sloping demand curve L(W), where W is the wage. The wage is set unilaterally by the trade union whereas aggregate employment is determined entirely by the firms. Let B denote unemployment benefits, the fraction of unemployed in programmes, and ew and ep the disutility from effort in regular employment and in a programme, respectively, both measured in money terms. The objectives of the trade union are to maximize expected utility of a representative member or, identically in our setting (see, for example, Oswald, 1985), the aggregate utility of the members. Assuming that union members are risk neutral, the objectives are to WV (W , e p ) = ( N − U (W ))(W − e w ) + U (W ) B − U (W ) e p
(11.3 )
where U(W) ≡ N – L(W) defines the number of unemployed members which is increasing in the wage set by the union. The first term denotes aggregate utility of employed members: the number of employed members times the utility of an employed, which equals the going wage less the disutility of work. The second term denotes the utility that unemployed members obtain from receiving benefits. The third term is the disutility of unemployed members in a programme: the number of unemployed times the fraction in a programme times the disutility from effort in a programme.
Hansen and Tranæs
352
The result depends on whether the required effort in a programme is calculated using the wage set by the union. This is, for instance, the case if union members in a programme obtain the same pay per unit of effort as union members in regular employment such that programme participants have to deliver the effort ep = ewB/W. Below we consider two extremes where the effort requirement is either independent of the action of the union or based on the above formula. Thus, the effort requirement is non-manipulable by the union in the first case and (fully) manipulable in the second. Non-manipulable Effort Requirement In this case, the union has to balance a higher wage for employed members against the utility loss of members that become unemployed if the wage is raised. Maximization of (11.3) yields the wage equation14 W* =
e w + B − re p 1+ 1 / ε (W*)
, ε (W ) ≡ −
L ′( W ) W L(W )
=
U ′( W ) W N − U (W )
(11.4 )
where ε(W) is the wage elasticity of labour demand. Equation (11.4) states that the wage is set as a mark up on the net opportunity costs of having a regular job equal to the disutility of working in a regular job plus foregone unemployment benefits when working but less the disutility of effort in a programme. A more sensitive labour demand (larger value of ε(W*)) improves the employment–wage trade-off which reduces the mark up and moderates the wage claim. The outcome is illustrated in Figure 11.5 which is similar to a graphical analysis of a monopoly. The curve L(W) denotes the labour–demand schedule, the horizontal curve at ew + B – ep denotes the wage a worker would require to take regular employment and the vertical line is the number of union members. Without a union the equilibrium is at point 1 where supply equals demand and everybody is employed. When the union sets the wage, it takes into consideration that 1 more employed worker requires a wage reduction for all which is why the marginal revenue of employment (MR in Figure 11.5) lies to the left of the labour demand schedule. The optimal employment is determined where the MR curve intersect with the opportunity costs of employment, giving outcome 2 in Figure 11.5. All unemployed workers, the horizontal distance from point 2 to the vertical line, are involuntary unemployed as they all strictly prefer to work at the going wage. It follows from Figure 11.5 that an effort–benefit system influences the wage claim of the union both through the presence of unemployment benefits and through the effort requirement of the unemployed. Unemployment benefits contribute to involuntary unemployment by increasing the opportunity costs of employment, leading to larger wage claims, while the effort requirement moderates the wage claims by reducing the opportunity costs.
Effort Commitment in Active Labour Market Policy Figure 11.5
353
Wage setting
We previously asked whether it was possible to remove voluntary unemployment caused by a pure benefit system by introducing an effort–benefit system. Similarly we now ask whether it is possible to remove the part of involuntary unemployment that is due to the existence of unemployment benefits. Without any benefits or active programmes, the wage is set as a mark up only on the disutility of work illustrated by point 3 in Figure 11.5. This point is attainable only by the effort–benefit system if the effort requirement of unemployed is set such that the aggregate disutility of effort in programme, ep, is exactly equal to the benefit level, B. However, this is not compatible with the previous assumption that unemployed have to deliver the effort that corresponds to their productivity and the current benefit level – that is ep = Bew/W. It is clear that this restriction implies that ep is always less than B as w is always larger than the disutility of regular work (see Figure 11.5). Therefore, certainly ep < B as ≤ 1. To conclude, it is possible to reduce involuntary unemployment created by the benefit system through an effort requirement but not possible to remove
Hansen and Tranæs
354
it entirely unless we require the unemployed to deliver more effort than that which corresponds to their productivity and the current benefit level (that is, unless programme participants are paid less than the going wage). Manipulable Effort Requirement We now look at the other extreme case where the effort requirement in a programme is linked purely to the wage set by the union according to the rule ep = Bew/W, and where the union know this relationship. The union has the same objectives as given by (11.3) above. The only new element is that the union can manipulate the effort requirement of its unemployed members. The union has to take into consideration that a higher wage reduces the amount of effort that authorities require of unemployed members in order to obtain a given benefit level. In isolation, this effect tends to increase the wage claims of the union. It is possible to show (see Hansen and Tranæs, 1997) that the effort–benefit system reduces wages if the elasticity of unemployment with respect to the wage is larger than 1: U ′(W *)W * U (W *)
>1
This condition is quite intuitive when looking at the union’s aggregate disutility from unemployed in programme: U(W)Bew/W. The left-hand side of the above inequality represents the percentage increase in the number of unemployed workers who have to deliver effort in a programme when the wage is raised by 1 per cent. This effect was also present under the nonmanipulable effort requirement. The right-hand side stems from the new rule manipulation effect: a 1 per cent increase in the wage reduces disutility of effort in a programme by 1 per cent (because ep = Bew/W). Thus, the effort requirement tends to moderate wages if unemployment is relatively low and if unemployment is relatively sensitive to the wage. Note that this conclusion is independent of how strongly the instrument is used as the fraction of unemployed in a programme, , does not enter in the above formula. Efficiency Wages The effects of an effort–benefit system in an efficiency wage framework are found in Hansen and Tranæs (1997). If the effort requirement is nonmanipulable, the results are similar to those of the union model: the effort requirement reduces the opportunity costs of employment, which increases effort in regular jobs such that firms need less high wages to elicit effort from their employees.
Effort Commitment in Active Labour Market Policy
355
The effect of rule manipulation changes a lot when considering an efficiency wage analysis because it is the other side – the firms – that determine wages. Rule manipulation gives the firm less incentive to offer high wages, as this reduces the effort requirement in a programme. This reduces the utility loss of being fired, implying that effort in regular jobs decreases. Thus, the rule-manipulation effect enhances the direct beneficial effects on wages and employment in an efficiency wage framework.
11.6
CONCLUSION
We have seen that the hybrid strategy of incentives and educative measures that the new Danish ALMP represents can potentially go a long way in solving the unemployment problem without creating major new problems such as increasing inequality. The reason why the new ALMP looks particular promising is that it reverses the crowding-out effect associated with standard ALMP. It is obvious to try upgrade workers’ skills while they are unemployed, or at least make sure they do not lose their skills. This has motivated many countries to experiment with ALMP. The problem that may arise administratively when putting unemployed workers to work is a fall in regular employment: the programmes can have a positive effect on the effective labour supply but at the same time crowd out regular employment, making the overall effect ambiguous. The particular design underlying the new Danish ALMP can potentially avoid the negative crowding-out effect; it may even produce a ‘crowding-in’ effect if implemented systematically. This will then reinforce the positive effect of the activities themselves on the effective labour supply. Effort commitment reduces the opportunity costs of unemployment which, ceteris paribus, reduces both voluntary and involuntary unemployment. The effect on involuntary unemployment is derived in a union model but is also present in an efficiency wage model. One has to be careful, however, when implementing such a system. Rule-manipulation effects may arise when the effort requirement in a programme is linked to a wage measure that is influenced by wage setters. Rule-manipulation effects counteract and may overturn the beneficial effects in a union model. This is because higher wages reduce the effort requirement of the unemployed and thus increase the utility of unemployed workers. In an efficiency wage model this is not the case; here the rule-manipulation effect works in the opposite direction. Rule manipulation enhances the direct beneficial effects in this setting because lower wages increase the effort requirement of the unemployed; this reduces the opportunity costs of employment and thereby elicits more effort from employed workers. Thus, firms have
356
Hansen and Tranæs
less incentive to offer higher wages in an efficiency wage model when the effort requirement is manipulable. The scarce evidence available indicates that the Danish labour market may have improved structurally after the reforms. The Beveridge curve has moved inwards and real wage inflation has increased less than expected during the latest boom. The question is whether this is enough to solve the Danish unemployment problem. Probably not, although it is to soon for a final verdict. There are still features of the Danish labour market structures that are likely to cause problems. First, effort commitment is not applied in all programmes, and it is highly recommendable that it should be. Second, the very high replacement ratio for low-paid workers makes it very difficult to create participation incentives for them; for some workers the compensation is close to 100 per cent (after tax and subsidies). Notes 1. 2. 3. 4. 5. 6. 7. 8. 9.
10.
We wish to thank Torben M. Andersen, Niels Blomgren-Hansen, Jørgen Elmeskov and Donald Parsons for comments and discussion, and Rasmus Lentz for competent research assistance. Supposedly, this was the case for many programmes in Sweden (see Calmfors and Nymoen, 1990). It remains a puzzle, though, why many of the Swedish programmes were compulsory if they increased the utility of the unemployed. The first active programmes introduced in Denmark in the 1980s did experiment with similar principles. The idea of paying the going wage and reducing the work hours in programmes is also endorsed by the OECD (1994). For example, Pedersen and Smith (1995) find in a survey of 666 unemployed that 40 per cent are not unemployed according to ILO criteria. In the future programme participation is required after only one year of unemployment as decided by the Danish parliament in December 1998. Not only do the studies obtain different effects of different programmes, they also obtain different effects of the same programmes owing to the use of different statistical methods – for example, duration models or fixed-effect models. Furthermore, Langager and Madsen (1992) find that 33 per cent of the participants in private schemes are still at the same firm 1 year after the scheme ended whereas the figure for public schemes is 17 per cent. One study of this deadweight loss exists for Denmark. It shows that a quarter to a third of subsidized jobs would have existed without subsidy. Foreign evidence indicates deadweight losses between 60 per cent to 85 per cent (see Weise, 1997). A significant part of the reduction in unemployment has been due to reductions in labour supply rather than more jobs. But this does not explain away the puzzle, because employment did increase, private as well as public; for example, between 1994 and 1995 almost 50,000 extra came into private employment. The real wage is measured as average hourly nominal wage divided by the consumer price index. Data on nominal hourly wages are from the statistics of the Confederation of Danish Employers’ Associations, the consumer price index is published by Statistics Denmark.
Effort Commitment in Active Labour Market Policy 11. 12. 13. 14.
357
It has to be mentioned that wage inflation might be just about to take off. For instance, in the local negotiations of Spring 1997, some workers in construction obtained a nominal wage increase of some 8 per cent. Note that s = w – wn is not likely to be the cost-minimizing subsidy. If everyone is capable of earning B in the market then an initial direct effect from introducing the effort–benefit system would be to visualize the amount of involuntary unemployment by removing the voluntary part. For all technical details concerning second-order conditions, etc. the reader is referred to Hansen and Tranæs (1997).
References Besley, T. and S. Coate (1992) ‘Workfare versus welfare: incentives arguments for work requirements in poverty-alleviation programs’, American Economic Review, 82, pp. 249–61. Besley, T. and S. Coate (1995) ‘The design of income maintenance programmes’, Review of Economic Studies, 62, pp. 187–221. Calmfors, L. and A. Forslund (1991) ‘Real-age determination and labour market policies: the Swedish experience’, Economic Journal, 101, pp. 1130–48. Calmfors, L. and H. Lang (1995) ‘Macroeconomic effects of active labour market programmes in a union wage-setting model’, Economic Journal, 105, pp. 601–19. Calmfors, L. and R. Nymoen (1990) ‘Real wage adjustment and employment policies in the Nordic countries’, Economic Policy, 11, pp. 398–448. Danish Ministry of Labour (1996) Tilbud til ledige og orlovsordninger (Copenhagen). Dunlop, J.T. (1944) Wage Determination Under Trade Unions (New York: Macmillan). Hansen, C.T. and T. Tranæs (1997) ‘Effort commitment in active labour market programmes: consequences for participation incentives and wage formation’, Discussion Papers, University of Copenhagen. Holmlund, B. (1990) Svensk lønebilding – teori, empiri, politik, bilaga 24 till Långtidsutredningen, Finansdepartementet, Stockholm. Holmlund, B. and Johan Linén (1993) ‘Job matching, temporary public employment, and equilibrium unemployment’, Journal of Public Economics, 51, pp. 329–43. Jackman, R. (1990) ‘Wage formation in the Nordic countries viewed from an international perspective’, in L. Calmfors (ed.), Wage Formation and Macroeconomic Policy in the Nordic Countries (Oxford: SNS and Oxford University Press). Jackman, R., C. Pissarides and S. Savouri (1990) ‘Labour market policies and unemployment in the OECD’, Economic Policy, 5(11), pp. 449–90. Langager, K. (1997) Indsatsen over for de forsikrede ledige, Socialforskningsinstitutet, 97:20. Langager, K. and P.K. Madsen (1992) Forløbsanalyse af ATB-modtagere, bilag 9 i rapport om arbejdsmarkedets strukturproblemer, del II. Finansministeriet, Udredningsudvalget (Copenhagen). Layard, R., S. Nickell and R. Jackman (1991) Unemployment: Macroeconomic Performance and the Labour Market (Oxford: Oxford University Press). Neumann, G., P.J. Pedersen and N. Westergård-Nielsen (1994) ‘Long-run international trends in aggregate unionization’, European Journal of Political Economy, 7, pp. 249–74. OECD (1994) The OECD Jobs Study (Paris: OECD). Oswald, A.J. (1985) ‘The economic theory of trade unions: an introductory survey’, Scandinavian Journal of Economics, 87, pp. 160–93.
358
Hansen and Tranæs
Pedersen, P.J. and N. Smith (1995) ‘Search activity among employed and unemployed members of the workforce’, in G.V. Mogensen (ed.), Work Incentives in the Danish Welfare State (Aarhus: Aarhus Universitetsforlag). Solow, R.M. (1979) ‘Another possible source of wage stickiness’, Journal of Macroeconomics, 1, pp. 79–82. Weise, H. (1997) ‘Erfaringer med aktivering’, in L. Pedersen (ed.), Er der veje til fuld beskœftigelse?, Socialforskningsinstitutet, 97:13.
Comment Niels Blomgren-Hansen When in January 1993 the newly formed coalition government under the stewardship of the Social Democratic leader Poul Nyrup Rasmussen announced as its primary priority to ‘break the unemployment curve’ and reduce unemployment from 11 per cent in 1992 to less than 7 per cent by the turn of the century (Danish Government, 1993), most economists considered the goal overly ambitious, apparently with good reasons. Unemployment had been growing steadily since 1987; if anything, the pace was accelerating in the second half of 1992. The prospects for international economic development were grim. Forecasts of economic growth in the EU had recently been adjusted down to 34 of a percentage point; in the important Swedish market, growth was negative for the third year in a row. The competitiveness of Danish manufacturing industry had been seriously hurt by the devaluations of the Swedish krona and pound sterling. Above all, the experience in 1986–7 when a reduction in unemployment (down to 8 per cent from 10 12 per cent in 1983) resulted in a dramatically increasing wage inflation (up to 10 per cent in 1987) had made most analysts believe that the ‘natural rate of unemployment’ (NAIRU) was considerably higher than 8 per cent – and growing. For the years 1993 and 1994 the Ministry of Finance estimated the NAIRU at above 10 per cent, only 2 percentage points below the actual unemployment rate (see Ministry of Finance, 1996). Despite these odds, the government has succeeded in realizing its ambitious employment goals. By Fall 1997 unemployment had come down to 7 12 per cent (a little above 6 per cent by EU standards) without sign of accelerating wage inflation. From being a case illustrating the impossibility of combining egalitarian social policies and high employment within the framework of a market economy, Denmark has become a model of unemployment-combating policies for the EU. For quite a while the Danish ‘miracle’ was explained as a result of (a) traditional expansionary policies and (b) labour supply-reducing measures. And, clearly, these traditional policies were part of the government’s strategy to ‘break the unemployment curve’. Public investments were sped up, an underfinanced tax reform was enacted and elderly long-term unemployed wage-earners (over 50 years) were offered early retirement pensions. More unconventionally, employed as well as unemployed wage-earners were offered a government paid leave of absence for taking care of their children, for educational purposes or – at a reduced rate – as a ‘sabbatical’. Numerically the measures were ‘successes’. In 1995 almost 3 per cent of the labour force was on leave of absence and in 1996 almost 2 per cent had made use of the early retirement scheme for long-term unemployed (see Ministry of Economic Affairs, 1997). No wonder unemployment dropped. However, that is only part of the story. Labour supply did not shrink correspondingly, mainly owing to substitution between various labour market ‘arrangements’ and a flexible demand for public support from people previously outside the labour force. In fact, from 1992 to 1996 the labour force shrank by only about 1 12 per cent, or about one-third of the number of people making use of the new schemes. More remarkably, the reduction in the labour force due to leaves of absence and accelerated retirement went pari passu with an increase in ordinary employment,
359
360
Blomgren-Hansen
in particular in the private sector. From a low point in 1993 until now, employment in the ordinary labour market has increased by more than 150,000 or about 5 12 per cent, of which only about 1 percentage point can be explained by population growth (see Ministry of Economic Affairs, 1997). Employment expanded despite a gradual tightening of fiscal and labour market policies. So, leaving the ‘gimmicks’ aside, there still is a ‘miracle’ to be explained. Claus Thustrup Hansen and Torben Tranæs suggest in Chapter 11 that the explanation is a labour market reform enacted in 1994 and a subsequent (but not yet fully implemented) tightening of unemployment insurance eligibility rules agreed by the government and the opposition Conservative Party in December 1995 – the so-called ‘overhaul’. Most labour market analysts agree that the 1994 reform and the ‘overhaul’ – surveyed in Hansen and Tranæs’ chapter – contain elements that have improved the functioning of the labour market. However, they did not constitute a radical shift. On the contrary, they were minor revisions and it is in not conceivable that they should have caused the drastic turnaround in unemployment illustrated in Figure 11.1 (p. 342). Nor has anybody suggested that the similarly drastic increase in unemployment during the previous period reflected ill designed changes in labour market policies during that period. A number of observations support this scepticism: (1) neither the government nor independent economists predicted that the reform would have such large impact on the functioning of the labour market – in fact, at the time, economists focused as much on the negative effects on the effective labour supply of the paid leave of absence schemes as on the positive effects of the introduction of individual plans for activation and the tightening of eligibility rules (see, for example, Economic Council 1994), and (2) the implementation of the labour market reform was seriously delayed because of the administrative difficulties of drawing up individual activation plans – the number of unemployed during activation or education actually dropped in 1994 and 1995 and remains below the 1992 figures (Ministry of Economic Affairs, 1997). To the extent that labour market reform has been significant, its influence has probably been mainly indirect: it ‘sent the message’ that living inactively ‘on welfare’ (regardless of the name of the scheme) was no longer socially acceptable. The stigma attached to unemployment was increased. The theoretical part of the chapter falls into two – not fully integrated – parts. In the first part the authors illustrate, using a very simple model, that the disincentive effects of unemployment benefits are neutralized if the unemployment benefit is conditioned on the obligation of the unemployed to work as many hours within a ‘programme’ (at the going rate) as they would have had to do to earn the same amount of money in the ordinary labour market. There are two interesting things about this ‘Danish model’: first, in theory the effort commitment eliminates voluntary unemployment. Second, the policy is feasible in the sense that the construction of individual ‘disincentive-free’ action plans does not require knowledge of the unemployed’s preferences. Of course, it is no art to make a gift unattractive by conditioning it on onerous features. The ‘political elegance’ of the Danish model is that the disutilitycreating feature – work at the going rate – is such that neither the unemployed nor the unions can reasonably object. In practice, the model may be less than perfect. It is a crucial assumption that the unemployed person recently has had a job and that the wage rates he earned in that job is a good indicator of how much (potential employers think) s/he is worth in the ordinary labour market. That might not be the case. Correctly or incorrectly potential employers may think that an unemployed worker lost her job because s/he was worth less than the going wage rate, or that s/he has become worth less than her previous wage rate because of unemployment; even in a ‘programme’ outside her field, she
Comment
361
may lose touch with her profession. In addition, the disincentive-neutralization property requires that the unemployed can freely choose the number of hours s/he works in the ordinary labour market. So even though the ‘Danish model’ undoubtedly reduces the disincentive effects of unemployment benefits, we should still expect to see voluntary unemployed in the programme. In the second part of Chapter 11 Hansen and Tranæs discuss the macroeconomic implications of the work requirement. In this part the authors reduce the choices of the wage earners to being full-time employed in the ordinary labour market or parttime employed in a programme. The basic logic is the same, however. Being subject to a work requirement, any wage earner prefers working in the ordinary labour market to being unemployed (except those few who might be indifferent). The work requirement makes the unemployed put pressure on their union to keep wages down and, thereby, create more jobs. In a competitive labour market model, the work requirement would make the unemployed more efficient searchers and in an efficiency wage model it would make them less tempted to shirk. The result is the same: lower wages and more ordinary jobs. In short, workers would be crowded-in rather than crowdedout, as would be the case if the unemployed were given a full-time subsidized job at the going rate. Or so it is if the work requirement is not ‘manipulable’ – that is if, in the monopolyunion model, the union setting the ordinary wage rate does not believe that by raising the wage rate it reduces the work requirement and increases the utility of the unemployed. If the union is ‘sophisticated’, then the work requirement will reduce wages only if the elasticity of unemployment with respect to the wage rate is larger than 1. Frankly, I find this elaboration of the analysis a little ‘academic’. First, a comprehensive monopoly union will take into account the fact that the government is likely to adjust the unemployment benefit to the development in the wage level; consequently, it considers the ‘replacement ratio’ – and, hence, the required workload constant. Second, even if it considers the work load ‘manipulable’, it is unlikely that the elasticity of unemployment with respect to the wage rate should be less than 1: a rentmaximizing union will choose a point at the labour demand curve where the elasticity of employment with respect to the wage rate is larger than 1. The elasticity of unemployment equals the elasticity of employment multiplied by the ratio of the number of employed to that of unemployed. A necessary condition for the elasticity of unemployment to be less than 1 is that unemployment exceeds 50 per cent. One of the costs of formal policy analysis is that one has to simplify and disregard certain aspects. In my view, a description of the Danish model should include two features in addition to those described: (1) work commitment helps to maintain and develop the human capital of the unemployed, and (2) most programme jobs are in the public sector. From a structural and efficiency point of view, the unemployed should be engaged in on-the-job training in the growing sectors. That would ensure that the accumulated human capital is most valuable to the unemployed and to society. The fact that most programme jobs are in the public sector may have a negative efficiency effect on that sector. Even though Hansen and Tranæs are probably right in claiming that the Danish model has a crowding-in effect rather than a crowding-out effect for the economy as a whole, there is little doubt that it has a crowding-out effect in the public sector. Jobs as professionals are partly substituted by temporary jobs for less motivated people with little relevant experience. The deprofessionalization of the public sector is a cost that must be allowed for. In theory, this deficiency is readily remedied. What is required is that the ‘burden’ of creating programme jobs for the unemployed be shifted from the public to the private sector. However, so far this possibility has been rejected by firms and unions. The political and professional
362
Blomgren-Hansen
challenge is to design large-scale creation of programme jobs in the private sector on a non-discriminatory basis acceptable to the major interest groups and thereby improve the efficiency of the Danish model. My scepticism about the claim that labour market reforms explain the surprisingly drastic turnaround in unemployment in Denmark in recent years and my focus on the limitations of the theoretical analysis should not conceal the fact that I have found the chapter by Hansen and Tranæs stimulating and useful, for two reasons. First, it gives a simple theoretical description of important elements of Danish labour market policy. Second, Hansen and Tranæs show that elementary economic theory is useful in analyzing economic policy and – equally important – remind us that human beings are motivated by non-pecuniary factors as well as by money, that the negative incentive effects of a ‘too generous’ public support may be neutralized by other motivation factors or administrative rules and that the optimal policy towards unemployment is likely to be one that combines these elements.
References Danish Government (Regeringen) (1993) Ny Kurs Mod Bedre Tider. Economic Council (Det Økonomiske Råd) (1994) Dansk Økonomi (May). Ministry of Economic Affairs (Økonomiministeriet) (1997) Økonomisk Oversigt (October). Ministry of Finance (Finansministeriet) (1993 and 1996), Finansredegørelse.
Comment Jørgen Elmeskov There is something intuitively appealing about the results of Chapter 11. The starting point is that generous unemployment benefits imply work disincentives, that may lead to higher unemployment. But if the unemployed are ‘harassed’ through what the chapter euphemistically calls ‘effort commitment’, work disincentives are reduced and a generous benefit system may no longer lead to higher unemployment. In practice, what the chapter proposes seems to be a system which comes close to what is frequently referred to as ‘workfare’. At this basic level, it is difficult to disagree with the tenor of the chapter. Nevertheless, it is probably worth emphasizing some of the assumptions since these influence the rather upbeat results as concerns the effectiveness of an effort commitment system. First, it is costless to extract the effort commitment. In practice, this assumption clearly does not hold. Indeed, governments across OECD countries spend large amounts of money on active labour market policies (ALMPs). It is difficult to isolate the expenditure purely associated with the running of these programmes (that is, removing the part of expenses which constitute transfers to individuals), but that such expenditure is high may be conjectured from the fact that it is not uncommon for countries to spend what corresponds to 1–2 per cent of GDP on ALMPs. Clearly, the taxes needed to pay for these programmes have labour market ramifications but these are not considered in the chapter. Second, there is no positive effect for the individual of committing effort. This is, of course, rather different from the way many ALMPs are perceived. Typically, they aim to provide training, job experience and so on to the unemployed. However, in the chapter, such effects would increase the utility of being unemployed, because they would raise the future earnings capacity of workers, and thereby raise unemployment. This is clearly the case in a model with union wage setting, such as the one considered in section 11.5 (p. 350). In the model with different types of workers, considered in section 11.4 (p. 343), one might consider that training and work experience could lead to a worker shifting from one group to another, but this is not a track that is pursued in the chapter. Third, in the model with different types of workers, unemployment is entirely voluntary (as illustrated by Figure 11.3) (p. 345). In this case, the effort commitment system effectively boils down to presenting the unemployed with a benefit 2–1 effort package which is already available in the ordinary labour market. This evidently begs the question why the government would first add to the utility of the unemployed by giving them high benefits and subsequently reduce the utility by requiring effort commitment. The argument presented in the chapter is that this will lead to a more narrow income distribution, which in itself is positive. Nevertheless, this begs the question why a government would be more concerned with people’s income than their welfare. The answer could, of course, be a concern for the dependent families of benefit recipients. But the unemployed are presumably also concerned about their families and would already have factored family well-being into their own utility. Thus, while the chapter spells out clearly the constraints that policy is assumed to operate under, it could say more about the rationale for these constraints.
363
364
Elmeskov
Fourth, the model with endogenous wages considers only one category of workers. In practice, of course, workers differ significantly including, in the case of Denmark, in the following respect: the effective replacement rate of low-paid workers is much higher than for better-paid workers. By the logic of the model, this feature of the Danish benefit system presumably would imply that policies to extract effort commitment should be focused in particular on low-paid workers. Again, this is an issue which is not pursued in the chapter. Finally, one empirical point may be worth noting. The chapter argues that there are signs that the Beveridge curve has shifted inwards over the period 1993–6. Data on vacancies are clearly not very robust, but leaving that aside most of what appears like an inward shift in the Beveridge curve over this period seems to have taken place in 1996. That year was a rather extreme case in one respect: the government announced that it was going to discontinue the advanced early retirement scheme available to long-term unemployed above 50 years of age. The result of this announcement was that a large group of older unemployed decided to take advantage of the scheme while it was still in force. The removal of a large group of relatively marginalized people from the labour force would be expected to lead to an inward shift of the Beveridge curve. Thus, it may be somewhat hazardous to interpret the seeming shift in the Beveridge curve as evidence that the reorientation of labour market policies in the direction of effort commitment (to the extent that there is an appropriate description of the policy) has lead to an improved functioning of the labour market. This is the more so since other policy shifts without any direct link to effort commitment, such as the cut in effective duration of unemployment benefit also took effect in this period. In summing up, it is worth repeating that the basic thrust of the chapter seems fairly sensible. However, some of the assumptions made to demonstrate the effectiveness of an effort commitment system are extreme. As a result, the chapter probably overstates its case. Some further modelling work may present a more nuanced picture.
Index active labour market policy (ALMP) 300–1, 335–64 after 1993 338–0 before 1993 338 costs 363 evidence of effects 340–3 participation incentives and effort– benefit system 343–50, 360–1 wage setting 350–5, 361 adjustment mechanism 63 ageing of population 224 and intergenerational redistribution 32–4 see also generational accounting Agell, J. 322 Alesina, A. 25, 207, 258, 262, 279, 281 EFC and substantial cuts 233 fiscal consolidations 232, 239, 240–1 Allowance for Corporate Equity (ACE) 110 Andersen, D. 30 Andersen, T.M. 14, 71, 240, 252, 280 Appledorn, A. 30 Aschauer, D.A. 276 asset prices see consumption-CAPM associate status in EMU 48–51, 58–9 asymmetry 44–5 Atkinson, A.B. 131, 319, 322 Auerbach, A. 198 Austria 51 automatic stabilizers 266–72, 285–6, 290–2 Bach, H. 317 Balino, T.J.T. 58 bargaining 168 decentralized see decentralized bargaining theory and involuntary unemployment 166–7 see also unions Barro, R. 209
Barry, F.G. 231, 232, 241, 259 basic income, citizens’ 305 Basle-Nyborg Agreement 47 Baumol’s Law 274 Bawa, V.S. 99 Belgium 42, 50, 51 Bengtson, S. 310 Bergman, U.M. 250–3 Berndt, E.R. 276 Bertola, G. 231–2, 239, 257 Besley, T. 346 Beveridge curve 341, 342, 364 bilateralism 44–5 Bingley, P. 11 Bjørn, N. 11 Blanchard, O.J. 140, 230–1, 248, 257 equity premium 92–3 Blank, R. 327 Blomquist, S. 295–6 bond investments 66–105 consumption-CAPM 79–87, 94–6 Danish and Swedish returns compared 99–101 equity premium 76–9, 92–4 policy implications 87–8, 96–7 portfolio strategies 75–6, 101–3 return-risk characteristics 72–6 Bovenberg, L. 207 Bradford, D.F. 109 Brazil 330 Bretton Woods system (BWS) 63 bridging policy 6 Brown, S.J. 99 budget deficits 248–9, 278–81 Burnside, C. 94 business capital 142–3, 146, 152, 155, 156–7 business cycles current account and 18–20 macroeconomic trends and 1–6 public sector and macroeconomic volatility 266–72, 285–6, 290–2
365
366
Index
Caballero, R.J. 163 Callesen, P. 31 Campbell, J.Y. 67, 96 capital business capital 142–3, 146, 152, 155, 156–7 free international mobility 64 housing capital 143–4, 146, 153, 155 transitional tax relief 114–15, 124 capital income taxation 106–7, 116, 131–2, 163–4 Tax Reform Act 1993 136–9, 147 effects 148–54, 156–8 capital markets EPRU model 142–4, 146 imperfections of international 20–1 liberalization 14–16 care 295–6, 302 cash benefits 30 cash flow 110–11 computation of cash-flow tax rates 111–14 cheating 310–11 child care 302 Christensen, A.M. 240, 253 Christiansen, J. 66, 89 Christiansen, V. 295–6 Christophersen, H. 262 citizens’ basic income 305 Coate, S. 346 competitive devaluations 45 composition 269–1 Comprehensive Business Income Tax (CBIT) 109–10 comprehensive (global) income tax (GIT) 108 Confederation of Danish Employers’ Associations (DA) 169–70, 173 Confederation of Trade Unions (LO) 169–70 constant relative risk aversion (CRRA) 83–4, 94 consumer price index (CPI) 176, 178, 183
consumption 143–4 EFC and 226, 228–2; boom in Denmark 249–53, 258; international evidence 233–8 effects of debt elimination 209–10 effects of Tax Reform Act 151, 155, 156, 157 consumption-CAPM (C-CAPM) 67–8, 78–87, 93, 94–6 estimated risk aversion and IMRS 83–4 Grossman–Shiller variance bounds methodology 80–1, 95–6 Hansen–Jagannathan bounds for IMRS 81–3, 94–5 level and variability of stock prices 85–7 convergence criteria 16, 41–2, 43, 44, 50–1, 225 Copenhagen Stock Exchange 71 core currency 63 corporation taxes 111–14 cost competitiveness 258–1 countercyclicality 269–71 credibility of a currency board 49–50, 53–5, 58 cross-country comparisons EFC hypothesis 232–41 welfare state 266–72, 274–5, 280–1, 287–8, 290–2 Currency Board Arrangement (CBA) 49–51 suitability for Denmark 59 suitability for Greece 58–9 technical conditions for viability 53–5 current account 4–5 saving, investment, foreign debt and 17–21 cycles see business cycles Darenhill, C. 102 Daugaard, S. 177 debt foreign 17–21 government see government debt decentralized bargaining 170–7, 191 determinants of the wage rate 176–7 determination of the wage 175–6
Index minimum wage contracts and productivity levels 174 outcome of bargaining over minimum wages 175 deficits, budget 248–9, 278–81 Delors Report 61–2 dependence on social security 304–6 Devereux, M.B. 231, 232, 241, 259 differential tax treatment 131–2 direct tax systems 108–10 discount rate 85–7 disinflationary policy 7 Disney, R. 162 disposable income 120–4 distribution 164 flat tax and 115–24 welfare state and redistribution 294–6 dividends 137 Dogonowski, R.R. 280 Domar–Musgrave effect 271–2 Dornbusch, R. 241 double ageing 32–3 Drazen, A. 231–2, 239, 257, 281 DREAM model 159 dual income tax (DIT) 106, 109 optimality of 131 Dunlop, J.T. 351 early retirement scheme 29–30, 310, 359, 364 enlarged scheme 297–300, 312 entitlement and 197–8 incentive problems 311–13, 326, 328 Eaton, J. 315 Economic Council 246–8, 249 economic expansions 225–63 Danish fiscal reform 241–8; evaluation 248–50; structural VAR evidence 250–3 EFC hypothesis 228–2 international evidence 232–41 economic flexibility 327–8 economic growth see growth economic incentives see incentives economic policy 2–3, 35 general strategies 6–8 Edinburgh Agreement 16
367 education 9, 10, 13, 35, 192 spillover effects 277 welfare state 296, 301; human capital investments 314–15; wages, growth and education 331–2 see also training educational leave 313–14 educative strategy 337–8, 339–40 efficiency flat tax and 114–15 welfare state and 294–6, 325–6, 330–1 efficiency wages 166, 354–5 effort commitment in ALMP 335–64 work effort and intensity of search 315–8 effort–benefit system 347–9, 360 wage-setting 350–5, 361 effort requirement manipulable 354, 361 non-manipulable 352–4, 361 eligibility criteria 298–9, 326, 328 employment 34, 145, 232, 288 debt reduction and 209–10 effects of Tax Reform Act 147, 148, 157 impact of training schemes 13, 340–1 relative wage and 9, 10 trends 3–4, 28–9, 359–60 see also labour market; unemployment employment tax cut proposal 167–8 Engen, E.M. 276 Englander, A.S. 277 Engsted, T. 67, 78, 92, 96 Enoch, C. 58 entitlement 197–8, 326 EPRU model 140–6, 162–4, 208 business and housing capital markets 142–4, 146 calibration 144–5 economic distortions 145–6 general assumptions 140–1 labour market 141–2, 162–3 equality 278, 330–1 equity premium 66–7, 76–9, 92–4
368
Index
equity premium puzzle 94–6 Eskesen, L. 71 estimation risk 99, 104 Estonia 58 euro: linkage to 16–17, 48–51 European Central Bank (ECB) 40–1, 45, 46–8, 49, 50 European Monetary Institute (EMI) 45–6 European Monetary System (EMS) 14, 63, 261–2 European Monetary Union (EMU) 14, 16, 40–65, 193 associate status 48–51, 58–9 causes of macroeconomic instabilities 61–2 Denmark and CBA 59 ‘EMU debt’ 199–201 ERM 2 see exchange rate mechanism Greece and CBA 58–9 suggested taxonomy 56–7 variable geometry 40–3 European Union (EU) 1–2, 24, 25 deepening 58–9 eurosclerosis 1 exchange rate mechanism (ERM) 14 ERM 2 for outsiders to EMU 43–8, 51–2, 59, 62–4; flexibility 57; fulfilment of neccessary macroeconomic conditions 63–4 exchange rates 14–17 associate membership of EMU 48–51 fixed exchange rate policy 14, 261–3 expansionary fiscal contraction (EFC) 226–7, 257–9, 261 Danish consumption boom 249–53, 258 hypothesis 228–2; expectations approach to fiscal policy 228–30; extensions 230–2 international evidence 232–41 expectations approach to fiscal policy 228–30 of higher permanent disposable income 248–9
expenditure tax 108 experience 178, 183 expenditures, government see government expenditures financial market volatility 42 firms: incentives to 318 fiscal balance rule 205 fiscal contractions 225–63 Danish fiscal reform see fiscal reform EFC hypothesis 228–2 international evidence 232–41 fiscal policy 193–224 comparison of Denmark and Netherlands 221 dealing with ageing 224 expectations approach 228–30 intergenerational distribution of fiscal burdens 32–4, 203–7 limitations of generational accounting 223 macroeconomic and welfare effects of fiscal adjustments 207–12, 223–4 measures in fiscal reform 246 sustainability 198–203, 221–3 trends in fiscal performance 194–8 fiscal reform 1982 6–7, 226–7, 241–53, 257–9, 261–3 EFC hypothesis and 250–3 evaluation 248–50 fiscal policy measures 246 income policy 242–6, 261–3 official estimates of impact 246–8 fixed exchange rate policy 14, 261–3 flat tax (FT) 27–8, 106–33 computation of cash-flow tax rates 111–14 distributional considerations 115–24 dynamic and efficiency effects 114–15 measuring wage and cash-flow bases 110–11 flat tax income 117, 119–22 Flavin, M.A. 86, 95
Index flexibility, economic 327–8 foreign debt 17–21 foreign exchange tax 64 fragmentation 279 Frankel, J.A. 60 Frederiksen, N.K. 114–15 Freeman, R. 188 Frennberg, P. 78, 102 Gammie, M. 110 GDP growth after fiscal reform 249 growth of real GDP 1–3 impact of debt elimination 209–10 log GDP per capita 321 gender 185–6, 191–2 general equilibrium analysis 207–12, 223–4 see also EPRU model general job training 339 generational accounting 34, 216–17, 222 intergenerational distribution of fiscal burdens 203–7 limitations of 223–4 welfare effects of debt reduction 211–12 Gentry, W.M. 109 German Council of Economic Experts 226 Germany 47–8 Giavazzi, F. 210, 240, 241, 257–8 EFC 226, 229–30, 233, 239, 257 global income tax (GIT) 108 Gordon, R.H. 110, 120 government debt 4, 24–6 debt–GDP ratio trends 196, 197 reduction of 193–224; macroeconomic effects 207–10; sustainability 198–203, 214–15, 221–3; welfare effects 211–12 and Tax Reform Act 159 welfare state and 278–1, 288–9, 292 government expenditures and consumption boom 250–3 cross-country comparisons 265, 266–7
369 Danish fiscal reform 246, 248 and EFC 231–2, 233–8, 239 old-age expenditures 33, 299, 201, 206 and taxation 21–8 trends 194–6 welfare state 285–6, 303–4; government debt and 280–1, 288–9; growth and 276–8, 286–8, 292 government revenues 23, 196 cross-country comparisons 265, 266–7 effects of Tax Reform Act 137–40 rate of flat tax needed 111–14 government solvency 193–224 Graversen, E.K. 186, 307 Greece 45, 51, 57 suitability of a CBA 58–9 ‘green’ taxes 27, 134, 137 Gros, D. 42, 48 gross government debt 199–201 gross income 117–19, 122–4 Grossman, S.J. 67 Grossman–Shiller variance bounds methodology 80–1, 95–6 growth trends 1–3, 35, 196 welfare state and 272–8; debt accumulation 292; expenditure and taxes 286–8; size of welfare state 274–6, 333–4; wages, education and growth 331–2 see also GDP growth-adjusted real interest rate 197, 203 Gurney, A. 277 habit persistence 83–4, 95 Hagen, J. von 59 Hagen, K. 22, 197 Hall, R.E. 109 Hammour, M.L. 163 Hansen, C.T. 144, 168, 169, 186, 319, 354 Hansen, H. 302 Hansen, K. 70
370
Index
Hansen, L.P. 67 Hansen–Jagannathan volatility bounds 81–3, 94–5 Hansson, B. 78, 102, 276 health 296, 301 Heinesen, K. 261 Henrekson, M. 276 Hoff, K. 296 Hoffmeyer, E. 71 Holmlund, B. 144 housing capital 143–4, 146, 153, 155 ‘hub and spokes’ 44–5 Hubbard, R.G. 109 human capital ALMP and 361–2 investment in 224, 314–15 Tax Reform Act and human wealth 149, 155, 156 welfare state and 296, 314–15 see also education; training Hutchison, M.M. 250–3 implicit assets/liabilities 207 incentive strategy 337–8 incentives, economic 31–2 and the labour market 306–20; human capital investments 314–15; incentives to firms 318; labour supply 306–8; overall impact 320; participation 308–14; wage bargaining 319–20; work effect and intensity of search 315–18 participation incentives 343–50, 360–1; basic incentive roblem 343, 344; caveats 349–50; effort–benefit system 347–9; government information 346; labour supply decision 346; political constraints 343–5; pure benefit system 346–7 and the welfare state 278, 294–334 income tax 111 alternative direct tax systems 108–10
capital income see capital income taxation dual (DIT) 106, 109, 131 labour income see labour income taxation non-constant marginal tax rates 132 reform 26–7 see also flat tax; Tax Reform Act incomes policy 242–6, 261–3 incorporated firms 118, 120 indexation of transfers 197–8, 199, 202–3 individual job training 339 inflation 4, 5, 341–3 information government information and effort commitment 346 taxation and incomplete 132 infrastructure 276–7 Ingerslev, O. 31, 320 instabilities, macroeconomic see stability institutional investors 87–8, 96–7, 104 insurance, social 277–8, 279–1, 295 intensity of search 315–18 interest rates 42, 50–1 and consumption 240, 261–3 exchange rates, prices and 14–17 growth-adjusted real interest rate 197, 203 stock prices and real interest rates 86–7 intergenerational redistribution 32–4, 203–7 intergenerational welfare effects 154, 155–6, 157–8 international capital mobility 64 international shocks 2–3 international tax relations 124–5 intertemporal budget constraint government 198, 204, 214–15, 222, 229 households 20, 228–9 intertemporal marginal rate of substitution (IMRS) 81–6, 93 estimated risk aversion and 83–4
Index Hansen–Jagannathan bounds 81–3, 94–5 interventions by central banks 46–8 intramarginal interventions 46–7 investment booms and fiscal contraction 239, 249 public and growth 276–7, 332 saving, current account, foreign debt and 17–21 involuntary unemployment 165–7 effort–benefit system and 349–50, 351, 352–4 Ireland 240–1, 257–8 Italy 42, 51, 53–4 Jagannathan, R. 67, 81–3, 94–5 Jennergren, L.P. 86 Jensen, A.M. 13 Jensen, J. 27 Jensen, P. 10–11, 13, 320 Jensen, S.E.H. 34, 140, 144, 159, 189, 204, 240 job search, intensity of 315–8 job training 339–40, 340–1, 361–2 Kleidon, A.W. 86, 89 Klein, R.W. 99 Knudsen, M.B. 159 Kolm, A.-S. 144 Kotlikoff, L. 205 Kvist, J. 294 labour income taxation 106, 131–2 debt reduction 205–6, 207–12 flat tax and 110–11, 119 marginal tax rates and welfare state 302–3 Tax Reform Act 27, 135–6, 137; effects of labour tax reform 147–56, 158 and wage formation 176, 178, 185–6, 191–2 labour market 165–8 active labour market policy see active labour market policy EPRU model 141–2, 162–3 incentives and see incentives reforms 8–9, 35
371 structure of 169–70 wages and structural shifts in 8–13 see also employment; unemployment Labour Market Councils (LMCs) 340 labour supply 3–4, 162 econometrics 306–8, 326–7 effort–benefit system 346–50 marginal tax rates and wages 176, 186, 191–2 Langager, K. 340–1 Lange, K. 139–40, 158 Lassen, D.D. 302 Law Model (LM) 115, 116–17 Layard, R. 163 leave schemes 30, 313–14, 359 Liang, Y. 99 Lindbeck, A. 132, 313 Lockwood, B. 176, 186 long-term investments 72–6 see also bond investments; stock investments Lund, J. 67, 92, 95, 96 Lykketoft, M. 193 Lyon, A.B. 296 Lystbæk, B. 66, 89 Maastricht Treaty: convergence criteria 16, 41–2, 43, 44, 50–1, 225 macroeconomic volatility 266–72, 285–6, 290–2 Maddison, A. 314 Malcolmson, J.M. 168 mandatory interventions 46–8 Manning, A. 176 ‘March package’ 241 marginal tax rates and incentives 302–3 labour income and wage rate 176, 178, 185–6, 191–2 non-constant 132 marginalization 28–32 Martins, J.O. 163 Masson, P. 22, 197 mean reversion 76 Medoff, J. 189 Mehra, R. 67, 77–8, 94
372
Index
Micklewright, J. 319 Miller, M. 257 minimum payment contracts 170 minimum wage contracts 13, 165–94 and different productivity levels 174 outcome of bargaining 175 Mirrlees, J.A. 132 moral hazard 310–11 multinational corporations 124–5 Mundell, R. 56 Mussa, M. 22, 197 Myer, F.C.N. 99 natural rate of unemployment (NAIRU) 359 Netherlands 221, 222–3 Neumann, G. 338 ‘new view’ 20–1 Newey–West estimator 93 Nielsen, S.B. 18, 78, 110, 120, 131, 240, 302, 315 non-business income 117 non-constant marginal tax rates 132 non-human wealth 150, 155, 156 norms, social 326, 328 Norway 58 OECD 12 evidence for EFC in OECD countries 232–9 Okun, A.M. 294 old-age expenditures 33, 199, 201, 206 Ølgaard, A. 30, 196 Olsen, E. 71 opportunity cost of work 175 Pagano, M. 210, 240, 241, 257–8 EFC 226, 229–30, 233, 239, 257 parental leave 313–14 Parsons, D.O. 188 partial gift exchange 166 participation incentive effects on 308–14 incentives and effort–benefit system 343–50, 360–1
payroll taxes 27, 135 Pedersen, L.H. 32, 144, 170, 320 Tax Reform Act 139–40, 158 taxes and wages 168, 169, 186 Pedersen, P.J. 166, 295, 316, 317, 319, 321, 338 labour supply 306–7, 308 pension funds 87–8, 96–7, 104 pensions 300 debt elimination and cuts in pension benefits 207–12 performance, macroeconomic after Danish fiscal reform 249 fiscal consolidation and 232–9 permanent income, expected 248–9, 253 permanent tax ratios 198–203, 221–2 Perotti, R. 207, 233, 258, 262, 279 fiscal consolidations 232, 239, 240–1 personal allowances 112–14 personal income taxes 111–14 personal (unincorporated) firms 117–18, 119–20 Petersen, C. 207 Pilat, D. 163 Pisani-Ferry, J. 41 Pissarides, C.A. 166 Ploug, N. 31, 294 Pöhl, K.O. 61–2 political constraints 343–5 pool jobs 339 population ageing see ageing of population portfolio strategies 75–6, 101–3 Poterba, J.M. 76 Prescott, E. 67, 77–8, 94 prices 4, 5 asset see consumption-CAPM CPI 176, 178, 183 exchange rates, interest rates and 14–17 private saving 18, 19 private sector macroeconomic volatility 269–72, 273, 285–6, 291–2 programme jobs 361–2 productivity 288
Index effort–benefit system 347; productivity below the threshold 248–9 minimum wage contracts: different productivity levels 174; and wage rate 175, 176, 178–9, 191 public expenditure and total factor productivity 276 programme jobs 339–40, 340–1, 361–2 public debt see government debt public expenditures see government expenditures public investments 276–7, 332 public sector programme jobs 361–2 wages 13 welfare state see welfare state pure benefit system 346–7 R-base tax 110–11 Rabushka, A. 109 Raffelhüschen, B. 34, 159, 204 Rasmussen, P.N. 359 real interest tax 106–7, 111 realignments 46–8 recipient ratio 30 redistribution 294–6 see also distribution ‘registered government debt’ 199–201 regulation 87–8, 96–7 replacement rates 316–17, 319, 320, 338 reservation wage 316–8 retirement 310 age 311–13 early see early retirement scheme return–risk characteristics 66–7, 68–79 equity premium 76–9 long-term investments 72–6 short-term investments 69–72 Sweden and Denmark compared 99–101 revenue, government see government revenues Risager, O. 9, 14, 71, 78, 240
373 risk estimation risk 99, 104 return–risk characteristics see return–risk characteristics risk aversion 67 estimated and IMRS 83–4 ‘Robin Hood’ effect 167, 191–2 Rosen, H.S. 315 Rosenholm, M. 10–11 Roubini, N. 25 Rowthorn, R. 319 rule manipulation 354, 355 sabbatical leave 313–14 Sadka, E. 130 Sachs, J. 25 Sandmo, A. 131, 264, 295 Sartor, N. 168 saving 322 EPRU model 143–4, 146 impact of Tax Reform Act 150, 155, 157 investment, current account, foreign debt and 17–21 public and ageing 207–12, 224 Scarpetta, S. 163 Schluter, P. 262 search, intensity of 315–8 search theory 166 Shapiro, C. 166 Shiller, R. 67, 80–1, 95–6 shirking model 166 short-term investments 69–72 see also bond investments; stock investments Siegel, J. 67, 89, 92–3 Sinn, H.W. 296 size of welfare state 303–6, 325 growth and 274–6, 333–4 and macroeconomic volatility 267–9 Skidelsky, R. 257 skills, incentives and 350 Skinner, J. 276 Slemrod, J. 110, 132 Sløk, T. 144, 168, 169, 186 Smith, N. 32, 170, 186, 321 labour supply 307, 308
374
Index
Smith, N. – continued unemployment 10–11, 166, 316, 317 working hours 144 social insurance 277–8, 279–81, 295 social norms 326, 328 social protection 327–8 social security 297–301, 339 dependence on 304–6 government solvency and debt reduction 193–224 Søndergaard, J. 18, 240, 295 Sørensen, P.B. 106, 131, 134, 135, 139–40, 158, 315 spillover effects 276–8 stability, macroeconomic causes of instabilities and EMU 61–2 and welfare state 266–72, 285–6, 290–2 Stability and Growth Pact 40, 52 stabilization Danish fiscal reform see fiscal reform Ireland 240–1 standard wage contracts 170, 174 Stephensen, P. 32, 170 Stiglitz, J. 166 stock investments 66–105 consumption-CAPM 79–87, 94–6 Danish and Swedish returns compared 99–101 equity premium 76–9, 92–4 level and variability of stock prices 85–7 long-term investments 72–6 policy implications 87–8, 96–7 portfolio strategies 75–6, 101–3 short-term stock returns 69–72 structural VAR model 250–3 subsidies 137–40, 163–4, 348 Summers, L.H. 76 Supplementary Pension Fund (ATP) 87, 300 sustainability, fiscal 198–203, 221–3 intertemporal budget constraint and 214–15 Sutherland, A. 257 Sweden 262, 327–8, 330, 332, 335 and EMU 45, 46, 48, 51, 57
stock and bond investments 99–105 switching strategy 6
78,
Tabellini, G. 25 tax competition 125 Tax Reform Act 1993 26–7, 34, 126, 134–64, 196 EPRU model 140–6, 162–4 impact on tax rates 135–7, 138 macroeconomic and welfare effects 146–58; capital tax reform 148–54, 156–8; decomposition of tax reform 146–7; effects of total tax reform 158; labour tax reform 147–56 revenue effects 137–40 taxation 196, 239 alternative direct tax systems 108–10 changes in structure and unemployment 167–8 Danish fiscal reform 246, 248 differential tax treatment of different tax bases 131–2 expenditure tax 108 flat tax see flat tax ‘green’ tax 27, 134, 137 income tax see capital income taxation; income tax; labour income taxation intergenerational distribution of fiscal burdens 33–4, 203–7 payroll taxes 27, 135 permanent tax ratios 198–203, 221–2 public expenditures and 21–8 real interest tax 106–7, 111 and wage determination 175–6 welfare state 302–3; government debt 288–9, 292; growth 276, 286–8, 292; macroeconomic volatility 269–70, 272, 273, 285–6, 292 see also marginal tax rates; Tax Reform Act 1993 temporary lay offs 318 temporary leave 310 leave schemes 30, 313–14, 359
Index Thygesen, N. 17, 42, 48, 56 Toft-Nielsen, P. 86 trade 332 trade unions see unions training 13 ALMP 339–40, 340–1, 361–2, 363 Tranæs, T. 186, 354 transfer payments 21–3 cuts and evidence of EFC 233–8 Danish welfare system 297–301 debt reduction and cuts in 207–12 dependence on 304–6 indexation 197–8, 199, 202–3 macroeconomic volatility 269–70, 272, 273 marginalization 28–32 public debt and 280–1, 288 social insurance and incentives 277–8 see also welfare state transitional tax relief 114–15, 124 trends 1–6 unemployment 3–4, 7, 29, 30–1, 337 ALMP 337–40, 359, 360 education and 9, 10 impact of training programmes 13, 340–1 involuntary see involuntary unemployment NAIRU 359 and public finances 196, 206–7 voluntary 165–7, 336, 350, 363 wage formation 165–7, 175, 176, 178, 179, 183–5 work effort and intensity of search 315–18 see also employment; labour market unemployment benefits 9, 300–1 ALMP 300–1, 336, 338, 339, 360; youth unemployment 350; see also effort–benefit system incentives to firms 318 reductions 167 and wages 11–12, 176, 178, 185, 189–90
375 unemployment insurance (UI) funds 177 unincorporated (personal) firms 117–18, 119–20 unions EPRU model 141–2, 163 LO 169–70 minimum wage contracts and wage formation 170–5, 183–6, 188–90, 191–2 wage-setting and effort–benefit system 351–4, 361 United Kingdom (UK) 78 and EMU 45, 46, 48, 51, 57 United States (USA) 1–2, 77–8, 188, 321 taxation 24, 25 variable geometry 40–3 variance bounds 80–1, 95–6 variance ratio test 76 vector autoregression (VAR) 96 EFC and Danish fiscal reform 250–3 Very Short-Term Financing Facility (VSTFF) 47 volatility macroeconomic 266–72, 285–6, 290–2 stock market 80–3, 85–7, 94–6 voluntary unemployment 165–7, 336, 350, 363 Wage Earners’ Fund 87 wages base for flat tax 110–11 debt elimination 209–10 education, growth and 331–2 EPRU model 141–2, 144, 145, 163 incentives and wage bargaining 319–20 inflation 4, 5, 341–3 minimum wage contracts and individual wage formation 13,165–94; decentralized bargaining model 170–7; determinants of wage rate 176–7; determination of the
376
Index
wages : minimum wage ... – continued wage 175–6; outcome of bargaining over minimum wages 175 reservation wage 316–18 setting and effort–benefit system 350–5, 361; efficiency wages 354–5; manipulable effort requirement 354, 361; non-manipulable effort requirement 352–4, 361; simple unionized labour market 351–4 and structural shifts in the labour market 8–13 Wall Street crash 71, 92–3 wealth effects of short-term investments 69–72 effects of Tax Reform Act 150, 155–6, 157 Webb, J.R. 99 Weise, H. 30, 340 welfare 145–6 effects of fiscal adjustments 207–12 effects of tax reform 146–58 equality, efficiency and 330–1
welfare state 22–3, 197, 264–93 Danish welfare system 297–303; care 302; education and health 301; social security 297–301; taxation 302–3 and debt 278–81, 288–9, 292 dimensions of 295–6 and economic incentives 278, 294–334 and growth see growth macroeconomic volatility 264–72, 285–6, 290–2 and marginalization 28–32 size of see size of welfare state wages, education and growth 331–2 see also transfer payments Weller, P. 257 Westergård-Nielsen, N. 11, 319, 338 Whelan, K. 258 work effort see effort; effort–benefit system; effort requirement working-age population 28–9 working hours 288, 308, 309, 327, 349 youth unemployment
350
